(12) United States Patent (10) Patent No.: US 9.260,522 B2 Kufer Et Al

Home , Arunachal macaque, Assam macaque, Black colobus, Bonnet macaque, Booted macaque, Capped langur

US009260522B2

(12) United States Patent (10) Patent No.: US 9.260,522 B2 Kufer et al. (45) Date of Patent: Feb. 16, 2016

(54) BISPECIFIC SINGLE CHAIN ANTIBODIES WO WO 2008, 119565 A2 10/2008 WITH SPECIFICITY FOR HIGH WO WO 2008, 119566 A2 10/2008 MOLECULAR WEIGHT TARGET ANTIGENS WO WO 20089567 A2 102008 OTHER PUBLICATIONS (75) Inventors: Peter Kufer, Munich (DE); Claudia Blimel, Munich (DE); Roman Kischel, Sist etal (r. NA i. S. 2. Munich (DE) 139-159).*ariuZZa et al. . eV. O phyS. O phyS. e. : (73) Assignee: AMGEN RESEARCH (MUNICH) syst al. (Proc. Natl. Acad. Sci. USA. May 1987; 84 (9): 2926 GMBH, Munich (DE) Chien et al. (Proc. Natl. Acad. Sci. USA. Jul. 1989: 86 (14): 5532 5536).* (*) Notice: Subject to any disclaimer, the term of this Caldas et al. (Mol. Immunol. May 2003; 39 (15): 941-952).* patent is extended or adjusted under 35 Wils a systs, lig,...si:18): U.S.C. 154(b) by 553 days. 5.adoSeal. Elia? J. VTOl. (J. Immunol.S1Ol. Jul. 2002;, 169 (6): 3076-3084).*: (21) Appl. No.: 13/122,271 WuCasset et al. et (J.t Mol.(Biochem. Biol. Nov.Biophys. 19, 1999;Res. &R294 (1): 151-162).*Jul. . 2003; 307 (1): 198-205).* (22) PCT Filed: Oct. 1, 2009 MacCallum et al. (J. Mol. Biol. Oct. 11, 1996; 262 (5): 732-745).* Holmetal. (Mol. Immunol. Feb. 2007; 44 (6): 1075-1084).* (86) PCT NO.: PCT/EP2009/062794 ClinicalTrials.gov archive, "Phase II Study of the BiTE(R) Blinatumomab (MT103) in Patients With Minimal Residual Disease S371 (c)(1), of B-Precursor Acute ALL.” View of NCT00560794 on Aug. 11, (2), (4) Date: Jul. 14, 2011 2008, pp. 1-3, DXP-002572438. Maletz, K., et al., “Bispecific Single-Chain Antibodies as Effective (87) PCT Pub. No.: WO2010/037837 Tools for Eliminating Epithelial Cancer Cells From Human Stem Cell Preparations by Redirected Cell Cytotoxicity.” Int. J. Cancer, PCT Pub. Date: Apr. 8, 2010 (2001), pp. 409-416, vol. 93 DXP-00230 1955. Wolf, E., et al., “BiTEs: bispecific antibody constructs with unique (65) Prior Publication Data anti-tumor activity.” DDT. (2005), pp. 1237-1244, vol. 10 n. 18. www.drugdiscoverytoday.com. US 2011 FO262439 A1 Oct. 27, 2011 Bihler, et al. "A bispecific diabody directed against prostate-specific membrane antigen and CD3 induces T-cell mediated lysis of prostate cancer cells.” Cancer Immunol Immunother 57:43-52 (2008). Related U.S. ApplicationO O Data Kipriyanov,E. et MairAal., "Bispecific Human CD3xCD19 B Cells.” Int. Diabody J. C for 77,763-771T Cell-Me (60) Provisional application No. 61/101,933, filed on Oct. (1998). 1, 2008. Dreier et al., T cell costimulus-independent and very efficacious s inhibition of tumor growth in mice bearing Subcutaneous or leukemic (51) Int. Cl. human B cell lymphoma xenografts by a CD19-/CD3-bispecific C07K 6/30 (2006.01) isshain antibody construct, J. Immunol. 170(8):4397-402 C07K 6/28 (2006.01) Kipriyanov et al., Bispecific tandem diabody for tumor therapy with A61 K39/00 (2006.01) improved antigen binding and pharmacokinetics, J. Molec. Biol. (52) U.S. Cl. 293:41-56 (1999). CPC ...... C07K 16/2809 (2013.01); C07K 16/3069 (Continued) (2013.01); A61 K 2039/505 (2013.01); C07K 2317/31 (2013.01); C07K 23.17/34 (2013.01) Primary Examiner — Stephen Rawlings (58) Field of Classification Search (74) Attorney, Agent, or Firm — Marshall, Gerstein & Borun CPC. G01N33/505; C12N5/0636; C12N 5/0693; LLP C12N5/0093: C07K 16/28: C07K 16/2809; C07K 16/30; A61K 2039/5152 (57) ABSTRACT USPC ...... 435/4, 7.1, 325, 352,366,372.3: The present invention provides a method for the selection of 436/501; 530/387.1, 387.3, 388.2, bispecific single chain antibodies comprising a first binding 530/388.22,388.75 domain capable of binding to an epitope of CD3 and a second See application file for complete search history. binding domain capable of binding to the extracellular domain cell Surface antigens with a high molecular weight (56) References Cited extracellular domain. Moreover, the invention provides bispecific single chain antibodies produced by the use of the U.S. PATENT DOCUMENTS method of the invention, nucleic acid molecules encoding these antibodies, vectors comprising Such nucleic acid mol 5,726,044 A 3, 1998 Lo et al. ecules and methods for the production of the antibodies. Furthermore, the invention provides pharmaceutical compo FOREIGN PATENT DOCUMENTS sitions comprising bispecific single chain antibodies of the EP 1293514 A1 3, 2003 invention, medical uses of the same and methods for the WO WO-99.54440 A1 10, 1999 treatment of diseases comprising the administration of bispe WO WO 2004/106381 A1 12/2004 cific single chain antibodies of the invention. WO WO-2006, 125481 A1 11, 2006 WO WO 2007/042261 A2 4, 2007 4 Claims, 36 Drawing Sheets US 9.260,522 B2 Page 2

(56) References Cited Loeffler et al. Effect elimination of chronic lymphocyte leukaemia B cells by autologous T cells with a bispecific anti-CD 19, anti-CD3 OTHER PUBLICATIONS single-chain antibody construct, Leukemia 17(5):900-9 (2003). Kipriyanov et al., Effect of domain order on the activity ofbacterially International Search Report and Written Opinion of the International produced bispecific single-chain Fv antibodies, J. Molec. Biol. Searching Authority, European Patent Office, PCT/EP2004/005685, 330(1):99-111 (2003). dated Nov. 12, 2004. Loeffler et al. A recombinant bispecific single-chain antibody, CD19xCD3, induces rapid and highlymphoma-directed cytotoxicity by unstimulated T lymphocytes, Blood 95(6):2098-103 (2000). * cited by examiner U.S. Patent Feb. 16, 2016 Sheet 1 of 36 US 9.260,522 B2 Figure 1

MAb 5-10

EpCAM-D1-hNG2-CHO

EpCAM-D3-hNG2-CHO

EpCAM-D1D3-hNG2-CHO

EpCAM-D1D2-hNG2-CHO

EpCAM-hNG2-CHO

EpCAM-CHO

Fluorescence intensity U.S. Patent Feb. 16, 2016 Sheet 2 of 36 US 9.260,522 B2

Figure 2

0 EpCAM-CHO o EpCAM-D1-hNG2-CHO - - - O EpCAM-D3-hNG2-CHO x EpCAM-D1D3-hNG2-CHO

110- -D EpCAM-D1D2-hNG2-CHO 100 EpCAM-hNG2-CHO 90 80 ess 70 60 50 40 30 20 10

10-4 10-3 10-2 10-1 100 1 0 1 102 dilution of supernatant (%) U.S. Patent Feb. 16, 2016 Sheet 3 of 36 US 9.260,522 B2 U.S. Patent Feb. 16, 2016 Sheet 4 of 36 US 9.260,522 B2

Figure 4

human mutated unmutated PSMA rat PSMA rat PSMA La P1 x 2C La W W P2 x 2C La W A P3 x 2C La W A P4 x 12C La W W P5 x 12C

La La W D1 X 2C La F-1 D2 X 2C amanmmanummbFluorescence intensity U.S. Patent Feb. 16, 2016 Sheet 5 of 36 US 9.260,522 B2

Figure 5

90 80 as pse - - - P1 x 2C 60 Arsyfait ksoal P2 x 2C s 4. M 4 - - - - P3 x 2C S. 50 % / ^ P4 x 12C 1. ? M g 40 27 A P5 x 2C 27 A eam D1 x 12C 30 --Y-42 / M M 20 traig2-1 - ? a o o D2x2C 1. -Negative control 10 ------

10-1 100 101 102 dilution of supernatant % U.S. Patent Feb. 16, 2016 Sheet 6 of 36 US 9.260,522 B2

Figure 6

human mutated unmutaed PSMA rat PSMA rat PSMA C O r P6 X 2C U y n t S D3 x 12C

Fluorescence Intensity

Figure 7

100 a 90 fi 80 -21 f - 70 fa SS y - - - A D3 x 12C d 60 f M / - - - - P6 X 2C g 50 M A. negative control - 40 A 30 20 10 O 10-4 10-3 10-2 10-1 100 101 102 dilution of supernatant % U.S. Patent Feb. 16, 2016 Sheet 7 of 36 US 9.260,522 B2

Sunod

8eun61-I KOZIX7d U.S. Patent Feb. 16, 2016 Sheet 8 of 36 US 9.260,522 B2

Z[9TOiz LTi790Z U.S. Patent Feb. 16, 2016 Sheet 9 of 36 US 9.260,522 B2

Anti-FLAG

CHO transfected with a mutated human FAPA antigen with murine membrane-distal epitopes

CHO transfected With murine FAPA

Figure 10 U.S. Patent Feb. 16, 2016 Sheet 10 of 36 US 9.260,522 B2

O O S C -- to U.S. Patent US 9.260,522 B2

U.S. Patent Feb. 16, 2016 Sheet 12 of 36 US 9.260,522 B2

Effector T cells: stimulated human CD4/CD56 depleted human PBMC Target cells: CHO transfected with human FAPA E.T ratio: 10:1 FA2OH3HLx2CHL 9 8 FA22A9HL)d2CHL 8.- : RA22C11HL}{2CHL 70 x FA22D8HLd2CHL SO 5 A. 3. 2

-1) 10-8 - 1-8 1-5 10-4 1-3 1-2 1-1 dilution of supernatant

Effector T cells: stimulated human CD4/CD56 depleted human PBMC Target cells: CHO transfected with human FAPA E:T ratio: 10:1

g)- 8 FA 19D9HL}d2CHL B x FA22E8HLd2CHL & 70 A 50 " A.

------1-I - 10-G 1-5 1-4 - 1-2 1-1 dilution of supernatant

Figure 12a U.S. Patent Feb. 16, 2016 Sheet 13 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with a mutated human FAPA antigen with murine membrane-distal epitopes E.T ratio: 10:1

110 1OO 90 8O TO SO SO 40

O o FA22E8HLX2CHL O ------mm O-3 10-7 10-6 10-5 10-4 10-3 10-2 10-1 dilution of supernatant 6

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with a mutated human FAPA antigen with murine membrane-distal epitopes E:T ratio: 10:1

110 TF 1OO

5 O

3 O FA2CHL2CHL

10-8 O-7 O-6 10-5 10-4 10-3 10-2 10-1 dilution of supernatant 6

Figure 12b U.S. Patent Feb. 16, 2016 Sheet 14 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with murine FAPA E:T ratio: 10:1 FA22A9HLxi2CHL 'i . FA19D12HLxi2CHL 8O 70 x FA19D9HLX2CHL SO FA22E8HLX2CHL 50 40 3O 2O 10 {} O 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 dilution of supernatant %

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with murine FAPA E.T ratio: 10:1

OO 90 O FA22C11HLX2CHL 8O FA228HLX2CHL FO "r FA2H3HLXCHL SO SO

3O d a

1O 'r f O 10-8 10-7 10-6 C-5 10-4 10-3 10-2 10-1 dilution of supernatant %

Figure 12c U.S. Patent Feb. 16, 2016 Sheet 15 of 36 US 9.260,522 B2

Human C-MET transfected CHO HPBALL ME62A4HLX2CHL ME62A4H x2CHL 3

100 101 102 103 104 00 101 102 103 10? F 2-H FL2-H Macaque c-MET transfected CHO 4119LnPx

W Wm.WM Wu Sm sum

o0 101 102 103 104 o0 101 102 103 104 FL2. F2H Human C-MET transfected CHO HPBALL ME62A12HLX2CHL ME62A12HLX2CHL

A 100 101 102 103 104 00 101 102 103 104 F 2-H F 2. Macaque c-MET transfected CHO 4119LnPx ME62A12HLX2CHL

A-A O 0 101 102 103 104 oo 101 102 103 104 F 2-H FL2-H Figure 13a U.S. Patent Feb. 16, 2016 Sheet 16 of 36 US 9.260,522 B2

Human C-MET transfected CHO HPB-AL ME62C1 OHLX2CHL ME62C1 OHLX2OHL c

cd 00A-r 101 102 103 104 100 101 102 103 104 i FL2-H FL2-H Macaque c-MET transfected CHO 4119LnPx ME62C1 OHLX2CHL ME62C1 OHLX2CHL c o n d C co

C w C CN C 1 00 101 102 103 104 100 101 102 103 104 F2H F 2-H Human C-MET transfected CHO HPB-ALL

100 101 102 103 104 00 101 102 103 104 F2H FL2-H Macaque c-MET transfected CHO 4119LnPx ME62D11 HLXI2CHL 8.

00 101 102 103 104 100 101 102 103 10? FL2-H FL2-H Figure 13b U.S. Patent Feb. 16, 2016 Sheet 17 of 36 US 9.260,522 B2

Human C-MET transfected CHO HPBALL ME63F2HLX2CHL ME63F2HXI2CHL 3- 8. -- OOs o 2 cld 2 soo 35O 35C CN N A cd100 101-- 102 103 104 d100 101: MM 102 103 10 F 2-H F 2-H Macaque c-MET transfected CHO 4119LnPx ME63F2HLX2CHL ME63F2HLX2CHL c C o r C " v-v- WWW-mm------w- ---

o OO 2 of 92 of 3Sso 3SSo di o 100 101 102 103 104 100 101 102 103 10? F 2-H F2H Human C-MET transfected CHO HPB-ALL ME86H11 HLX2CHL ME86H11 HLX2OHL 9O ------C2 ------d g E 8 E 3 SR R. S. O d! C) o

dNA : is: f - CD a s 100 101 102 103 104 100 101 102 103 10? F2H FL2-H Macaque c-MET transfected CHO 4119LnPx ME86H11HLXI2CHL ME86H11HLXI2CHL d c 2 9 r - 9i g o o

2 c 2 o 3 COce. 39 cco S. S Od100 101 102 103 104 100 101 102 103 10 F 2-H FL2-H Figure 13c U.S. Patent Feb. 16, 2016 Sheet 18 of 36 US 9.260,522 B2

Oologog or oz o SunOO

: :

Oolog ogovo o O008s Og Ov SunOO SunOO U.S. Patent Feb. 16, 2016 Sheet 19 of 36 US 9.260,522 B2

Oolosogovo Sunoo

y0109%„01.gôl.20k H-ZT

OOO8 O9 O7 OCZ O008 Og Ov Og O sunoo SunOO

„01.g01.201101001 H-ZT

Oologogov o 0 SunOC U.S. Patent Feb. 16, 2016 Sheet 20 of 36 US 9.260,522 B2

THOZIXTHZ-199E|W ~~~~…,,……….…….……….……….~~~~);C)******************~*~~~~*~*~~~~~~~~~~ 01001„01.g01z01|

U.S. Patent Feb. 16, 2016 Sheet 21 of 36 US 9.260,522 B2

z0)101001„01.g0, H-ZT-|

Oologog or ozo SunOO U.S. Patent Feb. 16, 2016 Sheet 22 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO tr-ansfected with human C-MET

OO ME86H11 HLX2CHL 8 O -- ME62A12HLX2CHL

O-101-9 O-8 O-7 10-6 10-5 10-4 10-3 10-2 10-1 100 Supernatant dilution 6

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells. CHO transfected with human C-MET

1OO • ME82A4HLX2CHL - s 80- -- ME63F2HLXI2CHL o S. SO E 40 3. 20

10-1010-9 10-8 10-7 O-6 10-5 O-4 10-3 10-2 10-1 100 Supernatant dilution (%) Figure 15a U.S. Patent Feb. 16, 2016 Sheet 23 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells. CHO transfected with murine c-MET

OO 8 ME86H11HL)d2CHL SS H ME62A, 12H x2CHL g SO E 40

10-100-9 10-8 O-7 10-6 10-5 10-4 10-3 10-2 10-1 100 supernatant dilution o

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with murine c-MET

1OO 8O - ME83F2HLX2CHL S. MES2A4H x2CHL is 60 E 4D

O-i (10-91O-3 O-7 O-6 10-5 10-4 10-3 O-2 10-1 100 supernatant dilution yo) Figure 15b U.S. Patent Feb. 16, 2016 Sheet 24 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells: CHO transfected with the mutated murine C-MET antigen with human membrane-distal epitopes OO H ME62A12H (CHL s 80- . . ME86H11HLx)2CHL g 60 E 40 3. on 20

1O-101-9 10-8 10-7 10-6 10-5 O-4 10-3 10-2 10-1 100 supernatant dilution yo)

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells. CHO transfected with the mutated murine C-MET antigen with human membrane-distal epitopes 1OO -- ME63F2HLX2CHL S$ 80 ME62A4HLX2CHL

S. SO E 40 3. in 20

10-1010-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 C-1 100 Supernatant dilution a Figure 15c U.S. Patent Feb. 16, 2016 Sheet 25 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine CMET transfected CHO ME75H6HL ME75H6HL

stC. ------O 2 co, & C 9. 100 101 102 103 104 100 101 102 103 104 FL2-H FL2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes

100 101 102 103 104 100 101 102 103 F2-H F2-H Human CMET transfected CHO Murine CMET transfected CHO MES9B1HL ME99B1HL C s S. 102 103 104 100 101 102 103 104 FL2-H FL2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes ME99B1HL ME99B1HL O co co on

E E D s O O 8 . C Al-waaMik --- 100 101 102 103 104 100 101 102 103 104 FL2.H FL2-H Figure 16a U.S. Patent Feb. 16, 2016 Sheet 26 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine CMET transfected CHO ME05B7HL MEO5B7HL

o100 E 101 102 103 104 100 101 102 103 104 FL2-H FL2-H CHO cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human C-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO5B7HL

9. co co 2

d CD s

Od i s: 100 101 102 100 101 102 103 104 F2H F2H Human CMET transfected CHO Murine CMET transfected CHO MEO5F6HL MEO5F6HL

o i : CD 100 101 102 103 104 100 101 102 103 104 FL2-H FL2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human C-MET antigen with murine membrane-distal epitopes membrane-distal epitopes

100 101 102 103 104 101 102 103 104 Figure 16b U.S. Patent Feb. 16, 2016 Sheet 27 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine CMET transfected CHO MEO6B7HL MEO6B7HL

00 101 102 103 104 100 101 102 103 104 FL2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6B7HL

o w i 100 101 102 103 104 FL2-H Human CMET transfected CHO Murine CMET transfected CHO MEO6C6HL

S------& O 9 ob R. is 100 101 102 103 104 100 101 102 103 104 FL2-H FL2. CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6C6HL MEO6C6HL

O KO on an i s c) O CD C) ck. o: 100 101 102 103 104 100 101 102 103 104 FL2H FL2-H Figure 16c U.S. Patent Feb. 16, 2016 Sheet 28 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine cMET transfected CHO MEO6C7HL MEO6C7HL

Co. ter ge of&5 ?59 at "New 100 101 102 103 104 100 101 102 103 104 F2H F 2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human C-MET antigen with murine membrane-distal epitopes membrane-distal epitopes ME06C7HL

o O100 101 102 103 104 100 101 102 103 104 F2H FL2-H Human CMET transfected CHO Murine CMET transfected CHO MEO6D1 HL MEO6D1HL

d cd s

o 101 102 103 104 OW 10 102 103 10 F2H FL2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6D1 HL

co o v. A g s s d c CD () c 100 101 102 103 104 O100 101 102 103 104 FL2-H Figure 16d U.S. Patent Feb. 16, 2016 Sheet 29 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine CMET transfected CHO MEO6D2HL g co 2 s ; : O 100 101 102 104 100 101 102 103 104 FL2. FL2. CHO Cells transfected With a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human c-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6D2HL 8. K

o 100 101 102 103 104 100 101 102 103 10? F2H FL2-H Human CMET transfected CHO Murine CMET transfected CHO MEO6E1 OHL MEO6E1 OHL

3-r

00 101 102 103 104 100 101 102 103 104 F2 F2-H CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human C-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6E1 OHL MEO6E1 OHL

co st E C : C d O O als, ka. 100 101 102 103 104 100 101 102 103 104 F 2-H Figure 16e U.S. Patent Feb. 16, 2016 Sheet 30 of 36 US 9.260,522 B2

Human CMET transfected CHO Murine CMET transfected CHO MEO6F2H MEO6F2HL S. 8 --- 2 cdcyd 2 & a O 9 C 100 101 102 103 104 100 101 102 103 104 FL2H FL2-H

CHO Cells transfected with a mutated CHO Cells transfected with a mutated murine c-MET antigen with human human C-MET antigen with murine membrane-distal epitopes membrane-distal epitopes MEO6F2HL ME06F2HL 9 cur. co t a. o O () o100 O 1011 102 103 104 100 8.8101 102 103 10? FL2-H FL2-H

Figure 16f U.S. Patent Feb. 16, 2016 Sheet 31 of 36 US 9.260,522 B2

Human IGF-1R transfected CHO HPBALL IGF1R12HLX2CHL IGF1R12HLX2CHL c 9o :

2s & O C O o s st o: 100 0 1 102 103 104 100 101 102 103 104 F 2-H F2-H Macaque IGF-1R transfected CHO 4119LnPx IGF1R12HLX2CHL IGF1R12HLX2CHL C Sid -v-W- m - S ------8 3. a di 88: O CO: s c 100 101 102 103 104 100 101 102 103 104 F 2-H F 2-H Human IGF-1R transfected CHO HPBALL IGF1R24HLX2CHL IGF1R24HLX2CHL Cd o gun-unro-mem s c

i O ()

00 101 102 103 104 100 101 102 103 10? s F 2-H FL2H Macaque IGF-1R transfected CHO 4119LnPX GF1R24HLX2CHL IGF1R24HLX2CHL

-o-o-eso-wowroovywevorowaw O (NO ------:

on s O

CD al& 3. 00 to 102 103 104 100 101 102 103 104 F 2-H FL2-H Figure 17 U.S. Patent Feb. 16, 2016 Sheet 32 of 36 US 9.260,522 B2

„01.g01.20ky0k00 H-ZT-3

SunOO 8I,9.Infil U.S. Patent Feb. 16, 2016 Sheet 33 of 36 US 9.260,522 B2

Effector T cells: stimulated CD4/CD56 depleted human PBMC Target cells. CHO transfected with human IGF-1R

g -- GF1R24H x2CHL 80 v. IGF1R12HLXI2CHL 70 60 ...Y-sy-s---

SO X. 4. 3.

O O 1O-7 10-6 10-5 10-4 10-3 10-2 10-1 100 dilution

Figure 19 U.S. Patent Feb. 16, 2016 Sheet 34 of 36 US 9.260,522 B2

hUPSMArat140-169 in CHO huPSMArat 140-169 in CHO PSMA-P7 HL X 2C HL PSMA-D4HL X 2C H

oi 100 10 104 F 2-H

huPSMAalt281-284 in CHO huPSMArat281-284 in CHO PSMA-P7 HL X 2CHL PSMA-D4. HL X 2C HL

N go-oo-ooroo --- CNgorococorroo

o 100 104 FL2-H F 2-H

huPSMArat3OO-344 in CHO huPSMArat300-344 in CHO PSMA-P7 HL X 2CHL PSMA-D4HL X 2CHL

100 F 2

huPSMArat598-617 in CHO huPSMAratb98-617 in CHO PSMA-P7 HL x 2C HL PSMA-D4 H x 2CHL

Figure 20a U.S. Patent Feb. 16, 2016 Sheet 35 of 36 US 9.260,522 B2

huPSMArató83-690 in CHO huPSMArató83-690 in CHO PSMA-P7 HL X 2CHL PSMA-D4 HL X 2C HL

hUPSMArat/16-750 in CHO hUPSMArat/16-750 in CHO PSMA-P7 HL x 2CHL PSMA-D4HL x 2CHL

Figure 20b U.S. Patent Feb. 16, 2016 Sheet 36 of 36 US 9.260,522 B2

PSMA-D4 HL X 2CHL 1500

1 OOO

500

Figure 21

1OO - . PSMA-D4 HL x 2C HL PSMA-P7 HL x 2CHL Š, 80 g SO E 40 b 5, 20

concentration pg/ml)

Figure 22 US 9,260,522 B2 1. 2 BSPECIFIC SINGLE CHAN ANTIBODES sequence encoding the four V-domains, two linkers and one WITH SPECIFICITY FOR HIGH spacer can be incorporated into a Suitable host expression MOLECULARWEIGHT TARGET ANTIGENS organism under the control of a single promoter. This increases the flexibility with which these constructs can be CROSS-REFERENCE TO RELATED PATENT 5 designed as well as the degree of experimenter control during APPLICATIONS their production. Remarkable experimental results have been obtained using This application is the U.S. National Phase of PCT/ Such bispecific single chain antibodies designed for the treat EP2009/062794, filed Oct. 1, 2009, which claims priority ment of malignancies (Mack, J. Immunol. (1997) 158,3965 from U.S. Provisional Application No. 61/101,933 filed Oct. 10 1, 2008, all which are incorporated herein by reference in 70; Mack, PNAS (1995) 92,7021-5; Kufer, Cancer Immunol. entirety. Immunother. (1997) 45, 193-7: Löffler, Blood (2000) 95, The present invention provides a method for the selection 2098-103) and non-malignant diseases (Brühl, J. Immunol. of bispecific single chain antibodies comprising a first bind (2001) 166,2420-6); Brischwein et al. J. Immunother. (2007) ing domain capable of binding to an epitope of CD3 and a 15 30(8), 798-807; Bargou, et al. (2008) Science 321,974). If In second binding domain capable of binding to the extracellular Such bispecific single chain antibodies, one sclv unit is domain cell Surface antigens with a high molecular weight capable of activating cytotoxic cells, for example cytotoxic T extracellular domain. Moreover, the invention provides cells, within the immune system by specifically binding to an bispecific single chain antibodies produced by the use of the antigen on the cytotoxic cells, while the other ScFv unit spe method of the invention, nucleic acid molecules encoding cifically binds an antigen on a malignant cell intended for these antibodies, vectors comprising Such nucleic acid mol destruction. In this way, such bispecific single chain antibod ecules and methods for the production of the antibodies. ies have been shown to activate and redirect the immune Furthermore, the invention provides pharmaceutical compo system's cytotoxic potential to the destruction of pathologi sitions comprising bispecific single chain antibodies of the cal, especially malignant cells. In the absence of such a bispe invention, medical uses of the same and methods for the 25 cific single chain antibody construct, malignant cells would treatment of diseases comprising the administration of bispe otherwise proliferate uninhibited. cific single chain antibodies of the invention. When designing a new bispecific single chain antibodies Unifying two antigen binding sites of different specificity comprising one sclv unit is capable of recruiting cytotoxic into a single construct, bispecific antibodies have the ability cells, for example cytotoxic T cells, while the other scFv unit to bring together two discrete antigens with exquisite speci 30 specifically binds an antigen on a target cell to be eliminated ficity and therefore have great potential as therapeutic agents. by the recruited cytotoxic cell, it has been obsereved that This potential was recognized early on, leading to a number different combination of scFv's in the bispecific single chain of approaches for obtaining Such bispecific antibodies. Bispe antibodies show different effectivity in the elimination of the cific antibodies were originally made by fusing two hybrido target cells. The election of a promising candidate is an inten mas, each capable of producing a different immunoglobulin. 35 sive and time consuming procedure. The resulting hybrid-hybridoma, or quadroma, was capable The present invention provides means and methods for the of producing antibodies bearing the antigen specificity of the solution of this problem for a bispecific single chain antibod first parent hybridoma as well as that of the second parent ies binding with one domain to cytotoxic cells, i.e. cytotoxic hybridoma (Milstein et al., (1983) Nature 305, 537). How T cells, and with the second binding domainto target antigens ever, the antibodies resulting from quadromas often exhibited 40 with a high molecular weight extracellular domain. undesired properties due to the presence of an Fc antibody Accordingly, the present invention provides in a first portion. embodiment a method for the selection of bispecific single Largely due to Such difficulties, attempts later focused on chain antibodies comprising a first binding domain capable of creating antibody constructs resulting from joining two scEv binding to an epitope of CD3 and a second binding domain antibody fragments while omitting the Fc portion present in 45 capable of binding to the extracellular domain cell surface full immunoglobulins. Each Schv unit in Such constructs was antigens with a high molecular weight extracellular domain. made up of one variable domain from each of the heavy (VH) Different binding domains, which may be used as first bind and light (VL) antibody chains, joined with one another via a ing domain, are described in the art and in the appended synthetic polypeptide linker, the latter often being genetically sequence listing. As apparent from the above, the election of engineered so as to be minimally immunogenic while remain 50 an antigenic domain on a target cell for of preparation of a ing maximally resistant to proteolysis. Respective Sclv units target cell binding domain of a bispecific single chain anti were joined by a number of techniques including incorpora body is the critical step for the provision of new bispecific tion of a short (usually less than 10 amino acids) polypeptide single chain antibodies which allow for an efficient elimina spacer bridging the two scEv units, thereby creating a bispe tion of target cells via redirected T cell lysis. A first choice for cific single chain antibody. The resulting bispecific single 55 the election of an antigenic domain on a target cell for of chain antibody is therefore a species containing two VH/VL preparation of a target cell binding domain of a bispecific pairs of different specificity on a single polypeptide chain, single chain antibody might be a domain, which is easily wherein the VHand VL domains in a respective scFv unit are accessable from a steric point of view. Accordingly, the per separated by a polypeptide linker long enough to allow son skilled in the art would elect in the case cell surface intramolecular association between these two domains, and 60 proteins on target cells with a high molecular weight extra wherein the thusly formed schv units are contiguously teth cellular domain epitopes which are most distant from the ered to one another through a polypeptide spacer kept short target cell Surface are most exposed, therefore best accessible enough to prevent unwanted association between, for for T cells and thus particularly potent in redirecting T cell example, the VH domain of one schv unit and the VL of the cytotoxicity. However, it has been Surprisingly found that other scEw unit. 65 membrane distant epitopes of target cell Surface antigens with Bispecific single chain antibodies of the general form a high molecular weight extracellular domain show a poor described above have the advantage that the nucleotide potency of redirecting T cell cytotoxicity. US 9,260,522 B2 3 4 The method of the invention provides guidance for the 140, 169, 191, 308,334,339, 344, 624, 626, 716, 717 election of antigenic regions of cell Surface antigens with a and 721 are mutated to the corresponding amino acid high molecular weight extracellular domain which allow for residues of the wt rodent PSMA; and the selection of bispecific single chain antibodies with a high (iii) the rodent wt extracellular domain of PSMA on the cell potency for redirected T cell cytotoxicity. These cell surface 5 Surface; antigens with a high molecular weight extracellular domain (b) contacting each type of host cells (i), (ii) and (iii) of step are type I or type II integral membrane proteins with an (a) with the bispecific single chain antibodies and effector extracellular portion of >640 amino acids. The extracellular T cells; and portion of this group of membrane proteins is independently (c) identifying and isolating the bispecific single chain anti folded, thus formed by a single continuous stretch of extra 10 bodies that mediate the lysis of host cells expressing wt cellular amino acids adjacent to the transmembrane region in human extracellular domain of PSMA on the cell surface the primary protein sequence. In order to fulfil the require according to (b)(i) and of host cells expressing mutated ment of a high molecular weight extracellular domain in the form of the wt human PSMA on the cell surface according context of the invention, the extracellular domain essentially to (b)(ii) but not of host cells expressing the rodent wt comprises more than 640 amino acids. Optionally the extra 15 extracellular domain of PSMA on the cell surface accord cellular domain is characterized by at least one functionally ing to b(iii). and/or structurally defined subdomain formed by discontinu As noted above, prostate-specific membrane antigen ous stretches of extracellular amino acids within the primary (PSMA: PSM) is a large antigen falling under the provided protein sequence. Examples for Such cell Surface antigens definition of cell Surface antigens with a high molecular comprise prostate-specific membrane antigen (PSMA), weight extracellular domain. Israeli et al. (Cancer Res. 53: fibroblast activation protein C. (FAPC), Hepatocyte Growth 227-230, 1993) cloned a 2.65-kb cDNA for a prostate-spe Factor Receptor (c-MET), endosialin (TEM1 or CD248) and cific membrane antigen detected with a monoclonal antibody type 1 insulin-like growth factor receptor (IGF-1R). raised against the human prostatic carcinoma cell line PSMA and FAPO. are cell surface molecules for which the LNCaP. The PSMA gene encodes a 750-amino acid protein crystal structure and, thus, the three dimensional structure of 25 that has an apparent molecular weight of 100 kD (due to the extracellular domain are known in the art. These antigens posttranslational modification) and is expressed by normal show a compact discontinuous domain composition of the and neoplastic prostate cells. PSMA was originally defined extracellular domain. It has been Surprisingly found that by the monoclonal antibody (MAb) 7E11 derived from bispecific single chain antibodies binding to epitopes with a immunization with a partially purified membrane preparation distance of up to 60 A from the alpha C-atom of the thirteenth 30 from the lymph node prostatic adenocarcinoma (LNCaP) cell extracellular amino acid as counted from the junction of line (Horoszewiczetal. Anticancer Res. 7 (1987),927-35). A transmembrane and extracellular region (reference C-atom) 2.65-kb cDNA fragment encoding the PSMA protein was show a significant high efficiency in the redirected T cell lysis cloned and Subsequently mapped to chromosome 11 p11.2 of target cells. In contrast thereto, the efficiency in the redi (Israeli et al., loc. cit.; O'Keefeet al., Biochem. Biophys. Acta rected T cell lysis of target cells of bispecific single chain 35 1443 (1998), 113-127). Initial analysis of PSMA demon antibodies binding only to epitopes with a distance of more strated widespread expression within the cells of the prostatic than 60 A from the reference C-atom is reduced and thus secretory epithelium. Immunohistochemical staining demon renders such bispecific antibodies unattractive for a clinical strated that PSMA was absent to moderately expressed in development. hyperplastic and benign tissues, while malignant tissues c-MET, TEM1 and IGF-1 Rare cell surface molecules hav 40 stained with the greatest intensity (Horoszewicz et al., loc. ing a consecutive sequence of independently folded extracel cit.). Subsequent investigations have recapitulated these lular domains is formed by a corresponding sequence of results and evinced PSMA expression as a universal feature in continuous stretches of extracellular amino acids within the practically every prostatic tissue examined to date. These primary protein sequence. It has been Surprisingly found that reports further demonstrate that expression of PSMA bispecific single chain antibodies binding to epitopes within 45 increases precipitously proportional to tumor aggressiveness the first 640 amino acid residues counted from the junction of (Burger et al., Int. J. Cancer 100 (2002), 228-237; Chang et transmembrane and extracellular region show a significant al., Cancer Res. 59 (1999),3192-98: Changet al., Urology 57 high efficiency in the redirected T cell lysis of target cells. In (2001), 1179-83), Kawakami and Nakayama, Cancer Res. 57 contrast thereto, the efficiency in the redirected T cell lysis of (1997), 2321-24; Liu et al., Cancer Res. 57 (1997), 3629-34; target cells of bispecific single chain antibodies binding only 50 Lopes et al., Cancer Res. 50 (1990), 6423–29: Silver et al., to epitopes within the amino acid residues above the 640" Clin. Cancer Res. 9 (2003), 6357-62; Sweat et al., Urology 52 amino acid residue counted from the junction of transmem (1998), 637-40; Troyer et al., Int. J. Cancer 62 (1995), 552 brane and extracellular region is reduced and thus renders 558: Wright et al., Urology 48 (1996), 326-334). Consistent such bispecific antibodies unattractive for a clinical develop with the correlation between PSMA expression and tumor ment. 55 stage, increased levels of PSMA are associated with andro Based on these findings the invention relates in one gen-independent prostate cancer (PCa). Analysis of tissue embodiment to a method for the selection of bispecific single samples from patients with prostate cancer has demonstrated chain antibodies comprising a first binding domain capable of elevated PSMA levels after physical castration or androgen binding to an epitope of CD3 and a second binding domain deprivation therapy. Unlike expression of prostate specific capable of binding to the extracellular domain of prostate 60 antigen, which is downregulated after androgen ablation, specific membrane antigen (PSMA), the method comprising PSMA expression is significantly increased in both primary the steps of: and metastatic tumor specimens (Kawakami et al., Wright et (a) providing at least three types of host cells expressing al., loc. cit.). Consistent with the elevated expression in (i) the wt human extracellular domain of PSMA (SEQID androgen-independent tumors, PSMA transcription is also NO: 447) on the cell surface; 65 known to be downregulated by Steroids, and administration of (ii) a mutated form of the wt human PSMA on the cell testosterone mediates a dramatic reduction in PSMA protein Surface, wherein the amino acid residues at positions and mRNA levels (Israeli et al., Cancer Res. 54 (1994), 1807 US 9,260,522 B2 5 6 11; Wright et al., loc. cit.). PSMA is also highly expressed in potency without requiring compensation for the negative secondary prostatic tumors and occult metastatic disease. influence of more membrane-distance by other properties of Immunohistochemical analysis has revealed relatively the bScAb Such as a very high affinity to the target antigen. intense and homogeneous expression of PSMA within meta Moreover, the cytotxic potency relative to the distance of the static lesions localized to lymph nodes, bone, Soft tissue, and 5 epitope bound by a bispecific single chain antibody is dem lungs compared with benign prostatic tissues (Chang et al. onstrated in examples 3 and 4. (2001), loc. cit.; Murphy et al., Cancer 78 (1996), 809-818; Examples for assays for performing the steps of the method Sweat et al., loc. cit.). Some reports have also indicated lim according to the invention are described in the appended ited PSMA expression in extraprostatic tissues, including a examples. subset of renal proximal tubules, some cells of the intestinal 10 The invention further relates to a method for the selection brush-border membrane, and rare cells in the colonic crypts of bispecific single chain antibodies comprising a first bind (Chang et al. (1999), Horoszewicz et al., Israeli et al. (1994), ing domain capable of binding to an epitope of CD3 and a Lopes et al., Troyer et al., loc. cit.). However, the levels of second binding domain capable of binding to the extracellular PSMA in these tissues are generally two to three orders of domain offibroblast activation protein C. (FAPC), the method magnitude less than those observed in the prostate (Sokoloff 15 comprising the steps of et al., Prostate 43 (2000), 150-157). PSMA is also expressed (a) providing at least three types of host cells expressing in the tumor-associated neovasculature of most solid cancers (i) the wt human extracellular domain of FAPC. (SEQ ID examined yet is absent in the normal vascular endothelium NO: 448) on the cell surface; (Chang et al. (1999), Liu et al., Silver et al., loc. cit.). (ii) a mutated form of the wt human FAPC. on the cell Although the significance of PSMA expression within the 20 Surface, wherein the amino acid residues at positions vasculature is unknown, the specificity for tumor-associated 144, 185, 186, 229, 267, 273, 274, 278, 284, 301,328, endothelium makes PSMA a potential target for the treatment 329, 331, 335 and 362 are mutated to the corresponding of many forms of malignancy. amino acid residues of the wt rodent FAPC.; and As apparent from SEQ ID NO: 447 the extracellular (iii) the rodent wt extracellular domain of FAPC. on the cell domain of PSMA comprises 707 amino acid residues. The 25 Surface; 13" aa as counted from the junction of transmembrane and (b) contacting each type of host cells (i), (ii) and (iii) of step extracellular region (reference C-atom) is a histidine. The (a) with the bispecific single chain antibodies and effector identification of the amino acid residues to be mutated for the T cells; and mutant human PSMA is described in detail in appended (c) identifying and isolating the bispecific single chain anti example 2. According to the method of the invention all 30 bodies that mediate the lysis of host cells expressing wt amino acid residues which do not match between the mouse human extracellular domain of FAPO. on the cell surface and the rodent extracellular domain of PSMA and which have according to (b)(i) and of host cells expressing mutated a distance of more than 60 A from the reference C-atom are form of the wt human FAPO. on the cell surface according mutatet from the human sequence to the rodent sequence. to (b)(ii) but not of host cells expressing the rodent wt This mutation results in a transformation of all antigenic 35 extracellular domain of FAPC. on the cell surface according regions with a distance of more than 60 A from the reference to b(iii). C-atom from the human specific form to the rodent specific Another large antigen according to the above definition is form. Antibodies, e.g. bispecificantibodies which are specific the cell surface protease fibroblast activation protein alpha for human epitopes comprising antigenic regions with a dis (FAP alpha). In epithelial cancer, invasion and metastasis of tance of more than 60 A from the reference C-atom (specific 40 malignant epithelial cells into normal tissues is accompanied for membrane distal epitopes) do not bind to the mutant by adaptive changes in the mesenchyme-derived supporting human PSMA and the rodent PSMA. Accordingly, the stroma of the target organs. Altered gene expression in these method of the invention allows for a discrimination of anti non-transformed Stromal cells has been discussed to provide bodies which bind to epitopes comprising antigenic regions potential targets for therapy. FAP alpha is such an example for with a distance of more than 60 A from the reference C-atom 45 a target of activated tumor fibroblasts in tumor stroma. Fibro (antibodies specific for membrane distal epitopes) and a posi blast activation protein alpha is an inducible cell Surface tive identification and isolation of antibodies specific for glycoprotein that has originally been identified in cultured epitopes of the human PSMA within a distance of less than 60 fibroblasts using monoclonal antibody F19. Immunohis A from the reference C-atom (antibodies specific for mem tochemical studies have shown that FAP alpha is transiently brane proximal epitopes). 50 expressed in certain normal fetal mesenchymal tissues but As apparent from the appended examples it has been Sur that normal adult tissues as well as malignant epithelial, neu prisingly observed that the distance of the epitope from the ral, and hematopoietic cells are generally FAP alpha-nega cell membrane is a critical factor for the cytotoxic potency of tive. However, most of the common types of epithelial can a bispecific single chain antibody which engages effector T cers contain abundant FAP alpha-reactive stromal fibroblasts. cells and target cells, such as PSMA" cells. The general effect 55 FAP alpha cDNA was cloned and published in Gen Bank underlying the distance between the epitope, which is bound (Accession number NM 004460). The predicted human by an according bispecific single chain antibody, from the cell FAP alpha protein is a type II integral membrane protein with membrane of a target cell is exemplified in a model in the a large C-terminal extracellular domain, which contains 6 appended example 1. What came as a Surprise according to potential N-glycosylation sites, 13 cysteine residues, and 3 this example, however, was the large extent of the loss in 60 segments that correspond to highly conserved catalytic target cell lysis observed between a target size of 640 aa (D1) domains of serine proteases; a hydrophobic transmembrane and 679 aa (D3). Despite this small difference in target size segment; and a short cytoplasmic tail. FAP-alpha shows 48% there was more loss in target cell lysis than from 679 aa (D3) amino acid identity with dipeptidyl peptidase IV (DPP4) and to 1319 aa (D1+D3). Thus, a target size of 640 aa was unex 30% identity with DPP4-related protein (DPPX). Northern pectedly found as upper threshold for the membrane-distant 65 blot analysis detected a 2.8-kb FAP alpha mRNA in fibro epitopes of bispecific single chain antibodies, still capable of blasts. Seprase is a 170-kD integral membrane gelatinase inducing redirected T cell cytotoxicity with reasonable whose expression correlates with the invasiveness of human US 9,260,522 B2 7 8 melanoma and carcinoma cells. Goldstein et al. (Biochim. (d) identifying and isolating the bispecific single chain anti Biophys. Acta 1361: 11-19, 1997) cloned and characterized bodies that mediate the lysis of host cells according to (b)(i) the corresponding seprase cDNA. The authors found that and (b)(ii) but not of host cells according to b(iii). seprase and FAP alpha are the same protein and products of As described herein above, c-MET, TEM1 and IGF-1R are the same gene. Pineiro-Sanchez et al. (J. Biol. Chem. 272: 5 type I or type II integral membrane proteins with an extracel 7595-7601, 1997) isolated seprase/FAP alpha protein from lular portion of more than 640 amino acids. The consecutive the cell membranes and shed vesicles of human melanoma sequence of the extracellular domain comprises continuous LOX cells. Serine protease inhibitors blocked the gelatinase stretches of extracellular amino acids within the primary pro activity of seprase/FAP alpha, Suggesting that seprase/FAP tein sequence and independently folded extracellular subdo 10 main formed by a single continuous stretches. Hepatocyte alpha contains a catalytically active serine residue(s). The growth factor receptor MET (C-MET) is involved in the pro authors found that seprase/FAP alpha is composed of mono gression and spread of numerous human cancer types. The meric, N-glycosylated 97-kD subunits that are proteolytically MET oncogene, encoding the receptor tyrosin kinase (RTK) inactive. They concluded that seprase/FAP alpha is similar to for hepatocyte growth factor (HGF) and Scatter Factor (SF), DPP4 in that their proteolytic activities are dependent upon 15 controls genetic programs leading to cell growth, invasion, Subunit association. Due to its degrading activity of gelatine and protection from apoptosis. Deregulated activation of and heat-denatured type-I and type-IV collagen, a role for MET is critical not only for the acquisition of tumorigenic seprase/FAP alpha in extracellular matrix remodeling, tumor properties but also for the achievement of the invasive phe growth, and metastasis of cancers has been suggested. More notype (Trusolino, L. & Comoglio, P. M. (2002) Nat. Rev. over, seprase/FAP alpha shows a restricted expression pattern Cancer 2, 289-300). The role of MET in human tumors in normal tissues and a uniform expression in the Supporting emerged from several experimental approaches and was stroma of many malignant tumors. Therefore, seprase/FAP unequivocally proven by the discovery of MET-activating alpha may be used as a target for exploring the concept of mutations in inherited forms of carcinomas (Schmidt et al., tumor stroma targeting for immunotherapy of human epithe Nat. Genet. 16 (1997), 68-73; Kim et al., J. Med. Genet. 40 lial cancer. However, though several clinical trials have been 25 (2003), e97). MET constitutive activation is frequent in spo initiated to investigate seprases/FAP alpha's role as a tumor radic cancers, and several studies have shown that the MET antigen target, conventional immunotherapy approaches or oncogene is overexpressed in tumors of specific histotypes or inhibition of seprase/FAP alpha enzymatic activity so far did is activated through autocrine mechanisms. Besides, the MET not yet result in therapeutic efficacy (see e.g. Welt et al., J. gene is amplified in hematogenous metastases of colorectal 30 carcinomas (Di Renzo et al., Clin. Cancer Res. 1 (1995), Clin. Oncol. 12:1193-203, 1994: Narra et al., Cancer Biol. 147-154). The Scatter Factor (SF) secreted in culture by fibro Ther. 6, 1691-9, 2007; Henry et al., Clinical Cancer Research blasts, that have the ability to induce intercellular dissociation 13, 1736-1741, 2007). As apparent from SEQID NO. 448 the of epithelial cells, and the Hepatocyte Growth Factor (HGF), extracellular domain of FAPC. comprises 734 amino acid a potent mitogen for hepatocytes in culture derived from residues. The 13" aa as counted from the junction of trans 35 platelets or from blood of patients with acute liver failure, membrane and extracellular region (reference C-atom) is a independently identified as Met ligands turned out to be the methionine. The identification of the amino acid residues to same molecule. Met and SF/HGF are widely expressed in a be mutated for the mutant human FAPO is described in detail variety of tissues. The expression of Met (the receptor) is inappended example 5. According to the method of the inven normally confined to cells of epithelial origin, while the tion all amino acid residues which do not match between the 40 expression of SF/HGF (the ligand) is restricted to cells of mouse and the rodent extracellular domain of FAPC. and mesenchymal origin. Met is a transmembrane protein pro which have a distance of more than 60 A from the reference duced as a single-chain precursor. The precursor is proteolyti C-atom are mutatet from the human sequence to the rodent cally cleaved at a furin site to produce a highly glycosylated Sequence. and entirely extracellular a-subunit of 50 kd and a B-subunit Moreover, the invention relates to a method for the selec 45 of 145 kd with a large extracellular region (involved in bind tion of bispecific single chain antibodies comprising a first ing the ligand), a membrane spanning segment, and an intra binding domain capable of binding to an epitope of CD3 and cellular region (containing the catalytic activity) (Giordano a second binding domain capable of binding to the extracel (1989) 339: 155-156). The C. and B chains are disulphide lular domain of Hepatocyte Growth Factor Receptor linked. The extracellular portion of Met contains a region of (c-MET), endosialin (TEM1) and type 1 insulin-like growth 50 homology to Semaphorins (Sema domain, which includes the factor receptor (IGF-1R), the method comprising the steps of: full C. chain and the N-terminal part of the B chain of Met), a (a) identifying the membrane proximal 640 amino acid resi cysteine-rich Met Related Sequence (MRS) followed by gly dues of the human and the rodent homolog of the extracel cineproline-rich (G-P) repeats, and four Immunoglobuline lular domain of c-MET, TEM1 or IGF-1R: like structures (Birchmeier et al., Nature Rev. 4 (2003), 915 (b) providing host cells expressing 55 25). The intracellular region of Met contains three regions: (1) (i) the human wt of the extracellular domain of the extra a juxtamembrane segment that contains: (a) a serine residue cellular domain of c-MET (SEQ ID NO: 439), TEM1 (Ser985) that, when phosphorylated by protein kinase Corby (SEQID NO:443) or IGF-1R (SEQID NO:446) on the Ca" calmodulin-dependent kinases downregulates the cell surface; receptor kinase activity Gandino et al., J. Biol. Chem. 269 (ii) a fusion protein comprising the human membrane 60 (1994), 1815-20); and (b) a tyrosine (Tyr 1003) that binds the proximal 640 amino acid residues identified in step (a) ubiquitin ligase Cbl responsible for Met polyubiquitination, and the rodent amino acid residues >640 of c-MET, endocytosis and degradation (Peschard et al., Mal. Cell 8 TEM1 or IGF-1R; and (2001), 995-1004); (2) the tyrosine kinase domain that, upon (iii) the rodent wt extracellular domain of c-MET, TEM1 or receptor activation, undergoes transphosphorylation on IGF-1R: 65 Tyr1234 and Tyr1235; (3) the C-terminal region, which com (c) contacting the host cells according to step (b) with the prises two crucial tyrosines (Tyr1349 and Tyr1356) inserted bispecific single chain antibodies and effector T cells; and in a degenerate motif that represents a multisubstrate docking US 9,260,522 B2 10 site capable of recruiting several downstream adaptors con affinity. The tyrosine kinase domain of IGF-IR and of IR has taining Src homology-2 (SH2) domains Met receptor, as most a very high sequence homology although the Zones of weaker Receptor Tyrosine Kinases (RTKs) use different tyrosines to homology respectively concern the cysteine-rich region situ bind specific signaling molecules. The two tyrosines of the ated on the alpha-subunit and the C-terminal part of the docking sites have been demonstrated to be necessary and beta-subunit. The sequence differences observed in the sufficient for the signal transduction both in vitro and in vivo a-Subunit are situated in the binding Zone of the ligands and (Maina et al., Cell 87 (1996), 531-542; Ponzetto et al., Cell 77 are therefore at the origin of the relative affinities of IGF-IR (1994), 261-71). and of IR for the IGFs and insulin respectively. The differ A further example for a molecule having a large extracel ences in the C-terminal part of the beta-subunit result in a lular domain is the tumor endothelial marker (TEM) Endo 10 divergence in the signalling pathways of the two receptors; sialin (=TEM1). TEMs are overexpressed during tumor IGF-IR mediating mitogenic, differentiation and antiapopto angiogenesis (St. Croix et al., Science 289 (2000), 1197– sis effects, while the activation of the IR principally involves 1202). Despite the fact that their functions have not been effects at the level of the metabolic pathways (Baserga et al., characterized in detailso far, it is well established that they are Biochim. Biophys. Acta, 1332: F105-126, 1997: Baserga R., strongly expressed on vascular endothelial cells in develop 15 Exp. Cell. Res., 253: 1-6, 1999). The cytoplasmic tyrosine ing embryos and tumors studies (Carson-Walter et al., Cancer kinase proteins are activated by the binding of the ligand to Res.61: 6649-6655, 2001). Accordingly, Endosialin, a 165 the extracellular domain of the receptor. The activation of the kDa type I transmembrane protein, is expressed on the cell kinases in its turn involves the stimulation of different intra surface of tumor blood vessel endothelium in abroad range of cellular substrates, including IRS-1, IRS-2, Shc and Grb 10 human cancers but not detected in blood vessels or other cell (Peruzzi F. et al., J. Cancer Res. Clin. Oncol., 125:166-173, types in many normal tissues. It is a C-type lectin-like mol 1999). The two major substrates of IGF-IR are IRS and Shc ecule of 757 amino acids composed of a signal leader peptide, which mediate, by the activation of numerous effectors down five globular extracellular domains (including a C-type lectin stream, the majority of the growth and differentiation effects domain, one domain with similarity to the Sushi/ccp/scrpat connected with the attachment of the IGFs to this receptor. tern, and three EGF repeats), followed by a mucin like region, 25 The availability of substrates can consequently dictate the a transmembrane segment, and a short cytoplasmic tail final biological effect connected with the activation of the (Christian et al., J. Biol. Chem. 276: 7408-7414, 2001). The IGF-IR. When IRS-1 predominates, the cells tend to prolif Endosialin core protein carries abundantly sialylated, erate and to transform. When Shc dominates, the cells tend to O-linked oligosaccharides and is sensitive to O-Sialoglyco differentiate (Valentinis B. et al.; J. Biol. Chem. 274:12423 protein endopeptidase, placing it in the group of Sialomucin 30 12430, 1999). It seems that the route principally involved for like molecules. The N-terminal 360 amino acids of Endosia the effects of protection againstapoptosis is the phosphatidyl lin show homology to thrombomodulin, a receptor involved inositol 3-kinases (PI 3-kinases) route (Prisco M. et al., in regulating blood coagulation, and to complement receptor Norm. Metab. Res., 31:80-89, 1999; Peruzzi F. et al., J. Can C1qRp. This structural relationship indicates a function for cer Res. Clin. Oncol., 125:166-173, 1999). The role of the Endosialin as a tumor endothelial receptor. Although Endo 35 IGF system in carcinogenesis has become the Subject of sialin mRNA is ubiquitously expressed on endothelial cells in intensive research in the last ten years. This interest followed normal human and murine Somatic tissues, Endosialin pro the discovery of the fact that in addition to its mitogenic and tein is largely restricted to the corpus luteum and highly antiapoptosis properties, IGF-IR seems to be required for the angiogenic tissues such as the granular tissue of healing establishment and the maintenance of a transformed pheno wounds or tumors (Opaysky et al., J. Biol. Chem. 276 (2001, 40 type. In fact, it has been well established that an overexpres 38795-38807: Rettig et al., PNAS 89 (1992), 10832-36). sion or a constitutive activation of IGF-IR leads, in a great Endosialin protein expression is upregulated on tumor endot variety of cells, to a growth of the cells independent of the helial cells of carcinomas (breast, kidney, lung, colorectal, support in media devoid of fetal calf serum, and to the for colon, pancreas mesothelioma), sarcomas, and neuroectoder mation of tumors in nude mice. This in itself is not a unique mal tumors (melanoma, glioma, neuroblastoma) (Rettig et 45 property since a great variety of products of overexpressed al., loc. cit.). In addition, Endosialin is expressed at a low level genes can transform cells, including a good number of recep on a subset of tumor stroma fibroblasts (Brady et al., J. Neu tors of growth factors. However, the crucial discovery which ropathol. Exp. Neurol. 63 (2004), 1274-83; Opaysky et al., has clearly demonstrated the major role played by, IGF-IR in loc. cit.). Because of its restricted normal tissue distribution the transformation has been the demonstration that the and abundant expression on tumor endothelial cells of many 50 R-cells, in which the gene coding for IGF-IR has been inac different types of solid tumors, Endosialin has been discussed tivated, are totally refractory to transformation by different as a target for antibody-based antiangiogenic treatment strat agents which are usually capable of transforming the cells, egies of cancer. However, so far, there are no effective thera such as the E5 protein of bovine papilloma virus, an overex peutic approaches using Endosialin as a tumor endothelial pression of EGFR or of PDGFR, the T antigen of SV 40, target. 55 activated ras or the combination of these two last factors (Sell A still further example for a large antigen is the insulin-like C. et al., Proc. Natl. Acad. Sci., USA,90: 11217-11221, 1993; growth factor I receptor (IGF-IR or IGF-1R). IGF-IR is a Sell C. et al., Mol. Cell. Biol., 14:3604-3612, 1994; Morrione receptor with tyrosine kinase activity having 70% homology A. J., Virol. 69:5300-5303, 1995; Coppola D. et al., Mol. with the insulin receptor IR. IGF-IR is a glycoprotein of Cell. Biol., 14:458a-4595, 1994; DeAngelis Tet al., J. Cell. molecular weight approximately 350,000. It is a hetero-tet 60 Physiol. 164:214-221, 1995). IGF-IR is expressed in a great rameric receptor of which each half-linked by disulfide variety of tumors and of tumor lines and the IGFs amplify the bridges-is composed of an extracellular a-subunit and of a tumor growth via their attachment to IGF-IR. Other argu transmembrane beta-subunit. IGF-IR binds IGFI and IGF II ments in favor of the role of IGF-IR in carcinogenesis come with a very high affinity but is equally capable of binding to from studies using murine monoclonal antibodies directed insulin with an affinity 100 to 1000 times less. Conversely, the 65 against the receptor or using negative dominants of IGF-IR. IR binds insulin with a very high affinity although the ICFs In effect, murine monoclonal antibodies directed against only bind to the insulin receptor with a 100 times lower IGF-IR inhibit the proliferation of numerous cell lines in US 9,260,522 B2 11 12 culture and the growth of tumor cells in vivo (Arteaga C. et al., other state-of-the-art domain prediction method. Indepen Cancer Res., 49:6237-6241, 1989 Li et al., Biochem. Bio dently folded domains of the same protein chain are often phys. Res. Com., 196:92-98, 1993; Zia Fet al., J. Cell. Biol. joined by a flexible segment of amino acids, with each half of 24:269-275, 1996; Scotlandi Ket al., Cancer Res., 58:4127 the flexible segment counting to its adjacent independently 4131, 1998). It has likewise been shown in the works of Jiang folded protein domain. Independently folded domains of the et al. (Oncogene, 18:6071-6077, 1999) that a negative domi same protein may be connected in a precursor molecule by a nant of IGF-IR is capable of inhibiting tumor proliferation. protease cleavage site and after proteolytical processing may The term "cell surface antigen” as used herein denotes a lie on two different connected protein chains in the mature molecule, which is displayed on the Surface of a cell. In most molecule. Independently folded protein domains may com cases, this molecule will be located in or on the plasma 10 prise functionally and/or structurally defined subdomains membrane of the cell such that at least part of this molecule which do not take their correct conformation without requir remains accessible from outside the cell in tertiary form. A ing Support by other portions of the protein because they are non-limiting example of a cell Surface molecule, which is formed by discontinuous stretches of extracellular amino located in the plasma membrane is a transmembrane protein acids within the primary protein sequence or kept in their comprising, in its tertiary conformation, regions of hydrophi 15 correct conformation by adjacent or other portions of the licity and hydrophobicity. Here, at least one hydrophobic protein. region allows the cell surface molecule to be embedded, or Therterm “method for the selection', respectively the term inserted in the hydrophobic plasma membrane of the cell “selecting denotes in the context of the present invention the while the hydrophilic regions extend on either side of the identification and isolation of one or more bispecific single plasma membrane into the cytoplasm and extracellular space, chain antibodies from a population of candidate antibodies. In respectively. Non-limiting examples of cell Surface mol particular, the candidate antibodies are tested in separate set ecules which are located on the plasma membrane are pro tings for the binding and the mediation of cytotoxicity for teins which have been modified at a cysteine residue to bear a each of the three different host cell populations. Populations palmitoyl group, proteins modified at a C-terminal cysteine of bispecific single chain antibodies to be tested and methods residue to bear a farnesyl group or proteins which have been 25 for the generation of Such populations are described in the modified at the C-terminus to bear a glycosyl phosphatidyl appended examples. Since the method of the invention allows inositol (“GPI) anchor. These groups allow covalent attach for the isolation of one or more bispecific single chain anti ment of proteins to the outer Surface of the plasma membrane, bodies the method is also understood as a method for the where they remain accessible for recognition by extracellular production of bispecific single chain antibodies of the inven molecules such as antibodies. Examples of cell Surface anti 30 tion. Of course, such method for the production involves the gens are CD3 (in particular CD3e), PSMA, FAPC, c-MET, production of the population of bispecific single chain anti endosialin and IGF-IR. As described herein above, PSMA, bodies, from which the one or more, which bind to the mem FAPO, c-MET, endosialin and IGF-IR are cell surface anti brane proximal epitopes, are isolated. gens which are targets for therapy of cancer, including, but As used herein, a “bispecific single chain antibody' not limited to solid tumors. 35 denotes a single polypeptide chain comprising two binding In light of this, the target antigens PSMA, FAPO, c-MET, domains. Each binding domain comprises one variable region endosialin and IGF-IR can also be characterized as tumor from an antibody heavy chain (“VH region'), wherein the VH antigens. The term “tumor antigen” as used herein may be region of the first binding domain specifically binds to the understood as those antigens that are presented on tumor CD3 molecule, and the VH region of the second binding cells. These antigens can be presented on the cell Surface with 40 domain specifically binds to the extracellular domain of a an extracellular part, which is often combined with a trans membrane protein on a target cell, e.g. to PSMA, FAPC, membrane and cytoplasmic part of the molecule. These anti c-MET, endosialin/TEM1 or IGF-1R. The two binding gens can sometimes be presented only by tumor cells and domains are optionally linked to one another by a short never by the normal ones. Tumor antigens can be exclusively polypeptide spacer. A non-limiting example for a polypeptide expressed on tumor cells or might represent a tumor specific 45 spacer is Gly-Gly-Gly-Gly-Ser (G-G-G-G-S) and repeats mutation compared to normal cells. In this case, they are thereof. Each binding domain may additionally comprise one called tumor-specific antigens. More common are antigens variable region from an antibody light chain (“VL region'), that are presented by tumor cells and normal cells, and they the VH region and VL region within each of the first and are called tumor-associated antigens. These tumor-associated second binding domains being linked to one another via a antigens can be overexpressed compared to normal cells or 50 polypeptide linker, for example of the type disclosed and are accessible for antibody binding in tumor cells due to the claimed in EP 623679 B1, but in any case long enough to less compact structure of the tumor tissue compared to nor allow the VH region andVL region of the first binding domain mal tissue. and the VH region and VL region of the second binding In accordance with the present invention an independently domain to pair with one another such that, together, they are folded protein domain is defined as a discrete portion of a 55 able to specifically bind to the respective first and second protein formed by a single continuous stretch of amino acids binding domains. within the primary protein sequence, e.g. known from its The term “protein' is well known in the art and describes crystal structure, to take the “correct conformation' without biological compounds. Proteins comprise one or more amino requiring Support by other portions of the protein or predicted acid chains (polypeptides), whereby the amino acids are to do so by comparison with hidden Markow models in librar 60 bound among one another via a peptide bond. The term ies of described sequence domains, such as PFAM (Bateman "polypeptide' as used herein describes a group of molecules, (2000) Nucleic Acids Res. 28: 263-266) and SMART which consists of more than 30 amino acids. In accordance (Schultz (2000) Nucleic Acids Res. 28: 231-234), sequence with the invention, the group of polypeptides comprises “pro similarity searches in databases with the BLAST and PSI teins' as long as the proteins consist of a single polypeptide BLAST tools (Altschul (1997) Nucleic Acids Res. 25:3389 65 chain. Also in line with the definition the term “polypeptide' 3402) that rely on the concept of a common evolutionary describes fragments of proteins as long as these fragments ancestor among sequentially homologous sequences or any consist of more than 30 amino acids. Polypeptides may fur US 9,260,522 B2 13 14 ther form multimers such as dimers, trimers and higher oli Various procedures are known in the art and may be used gomers, i.e. consisting of more than one polypeptide mol for the production of such antibodies and/or fragments. Thus, ecule. Polypeptide molecules forming Such dimers, trimers the (antibody) derivatives can also be produced by peptido etc. may be identical or non-identical. The corresponding mimetics. Further, techniques described for the production of higher order structures of Such multimers are, consequently, 5 single chain antibodies (see, inter alia, U.S. Pat. No. 4,946, termed homo- or heterodimers, homo- or heterotrimers etc. 778) can be adapted to produce single chain antibodies spe An example for a hereteromultimer is an antibody molecule, cific for elected polypeptide(s). Also, transgenic animals may which, in its naturally occurring form, consists of two iden be used to express humanized or human antibodies specific tical light polypeptide chains and two identical heavy for polypeptides and fusion proteins of this invention. For the polypeptide chains. The terms “polypeptide' and “protein’ 10 preparation of monoclonal antibodies, any technique, provid also refer to naturally modified polypeptides/proteins ing antibodies produced by continuous cell line cultures can wherein the modification is effected e.g. by post-translational be used. Examples for such techniques include the hybridoma modifications like glycosylation, acetylation, phosphoryla technique (Kohler and Milstein Nature 256 (1975), 495-497), tion and the like. Such modifications are well known in theart. the trioma technique, the human B-cell hybridoma technique The term “binding domain characterizes in connection 15 (Kozbor, Immunology Today 4 (1983), 72) and the EBV with the present invention a domain of a polypeptide which hybridoma technique to produce human monoclonal antibod specifically binds to/interacts with a given target structure/ ies (Cole et al., Monoclonal Antibodies and Cancer Therapy, antigen/epitope. Thus, the binding domain is an “antigen Alan R. Liss, Inc. (1985), 77-96). Surface plasmon resonance interaction-site'. The term “antigen-interaction-site' defines, as employed in the BIAcore(R) system can be used to increase in accordance with the present invention, a motif of a the efficiency of phageantibodies which bind to an epitope of polypeptide, which is able to specifically interact with a spe a target polypeptide, such as CD3 (epsilon), PSMA or FAPO. cific antigen or a specific group of antigens, e.g. the identical c-MET, TEM1 or IGF-1R (Schier, Human Antibodies Hybri antigen in different species. Said binding/interaction is also domas 7 (1996), 97-105: Malmborg, J. Immunol. Methods understood to define a “specific recognition'. The term “spe 183 (1995), 7-13). It is also envisaged in the context of this cifically recognizing means in accordance with this inven 25 invention that the term “antibody’ comprises antibody con tion that the antibody molecule is capable of specifically structs, which may be expressed in a host as described herein interacting with and/or binding to at least two, preferably at below, e.g. antibody constructs which may be transfected least three, more preferably at least four amino acids of an and/or transduced via, interalia, viruses or plasmid vectors. antigen, e.g. the human CD3 antigen and the target antigens as The term “specific interaction” as used in accordance with defined herein. Such binding may be exemplified by the 30 the present invention means that the binding domain does not specificity of a “lock-and-key-principle”. Thus, specific or does not significantly cross-react with polypeptides which motifs in the amino acid sequence of the binding domain and have similar structure as those bound by the binding domain, the antigen bind to each other as a result of their primary, and which might be expressed by the same cells as the secondary or tertiary structure as well as the result of second polypeptide of interest. Cross-reactivity of a panel of binding ary modifications of said structure. The specific interaction of 35 domains under investigation may be tested, for example, by the antigen-interaction-site with its specific antigen may assessing binding of said panel of binding domains under result as well in a simple binding of said site to the antigen. conventional conditions (see, e.g., Harlow and Lane, Anti Moreover, the specific interaction of the binding domain/ bodies: A Laboratory Manual, Cold Spring Harbor Labora antigen-interaction-site with its specific antigen may alterna tory Press, 1988 and Using Antibodies: A Laboratory tively result in the initiation of a signal, e.g. due to the induc 40 Manual, Cold Spring Harbor Laboratory Press, 1999). tion of a change of the conformation of the antigen, an Examples for the specific interaction of a binding domain oligomerization of the antigen, etc. with a specific antigen comprise the specificity of a ligand for The term “antibody’ comprises derivatives or functional its receptor. Said definition particularly comprises the inter fragments thereof which still retain the binding specificity. action of ligands, which induce a signal upon binding to its Techniques for the production of antibodies are well known in 45 specific receptor. Examples for said interaction, which is also the art and described, e.g. in Harlow and Lane Antibodies. A particularly comprised by said definition, is the interaction of Laboratory Manual, Cold Spring Harbor Laboratory Press, an antigenic determinant (epitope) with the binding domain 1988 and Harlow and Lane “Using Antibodies: A Laboratory (antigenic binding site) of an antibody. Manual Cold Spring Harbor Laboratory Press, 1999. The According to a preferred embodiment of the method of the term “antibody also comprises immunoglobulins (Igs) of 50 invention the first binding domain binds to CD3 epsilon different classes (i.e. IgA, IgG, IgM, Ig|D and IgE) and Sub (CD3e) of human and non-chimpanzee primate. In this con classes (such as IgG1, IgG2 etc.). text it is particularly preferred that the first binding domain The definition of the term “antibody' also includes capable of binding to an epitope of human and non-chimpan embodiments such as chimeric, single chain and humanized Zee primate CD3e chain binds to an epitope, which is part of antibodies, as well as antibody fragments, like, interalia, Fab 55 an amino acid sequence comprised in the group consisting of fragments. Antibody fragments or derivatives further com SEQ ID NOS. 2, 4, 6, and 8. prise F(ab'). Fv, ScPV fragments or single domain antibodies, As used herein, “human” and “man” refers to the species single variable domain antibodies or immunoglobulin single Homo sapiens. As far as the medical uses of the constructs variable domain comprising merely one variable domain, described herein are concerned, human patients are to be which might be VH or VL, that specifically bind to an antigen 60 treated with the same molecule. or epitope independently of other V regions or domains; see, The term “human' antibody as used herein is to be under for example, Harlow and Lane (1988) and (1999), loc. cit. stood as meaning that the bispecific single chain antibody as Such immunoglobulin single variable domain encompasses defined herein, comprises (an) amino acid sequence(s) con not only an isolated antibody single variable domain polypep tained in the human germline antibody repertoire. For the tide, but also larger polypeptides that comprise one or more 65 purposes of definition herein, said bispecific single chain monomers of an antibody single variable domain polypeptide antibody may therefore be considered human if it consists of Sequence. Such (a) human germline amino acid sequence(s), i.e. if the US 9,260,522 B2 15 16 amino acid sequence(s) of the bispecific single chain anti generation of an immune murine antibody Sclv library; (c) body in question is (are) identical to (an) expressed human identification of CD3 epsilon specific binders by testing the germline amino acid sequence(s). A bispecific single chain capability to bind to at least SEQID NOS. 2, 4, 6, and 8. antibody as defined herein may also be regarded as human if The context-independence of the CD3 epitope provided it consists of (a) sequence(s) that deviate(s) from its (their) 5 herein corresponds to the first 27 N-terminal amino acids of closest human germline sequence(s) by no more than would CD3 epsilon or functional fragments of this 27 amino acid be expected due to the imprint of Somatic hypermutation. stretch. The phrase “context-independent, as used herein in Additionally, the antibodies of many non-human mammals, relation to the CD3 epitope means that binding of the herein for example rodents such as mice and rats, comprise VH described inventive binding molecules/antibody molecules CDR3 amino acid sequences which one may expect to existin 10 does not lead to a change or modification of the conformation, the expressed human antibody repertoire as well. Any Such sequence, or structure Surrounding the antigenic determinant sequence(s) of human or non-human origin which may be or epitope. In contrast, the CD3 epitope recognized by a expected to exist in the expressed human repertoire would conventional CD3 binding molecule (e.g. as disclosed in WO also be considered “human for the purposes of the present 99/54440 or WO 04/106380) is localized on the CD3 epsilon invention. 15 chain C-terminally to the N-terminal 1-27 amino acids of the Though T cell-engaging bispecific single chain antibodies context-independent epitope, where it only takes the correct described in the art have great therapeutic potential for the conformation if it is embedded within the rest of the epsilon treatment of malignant diseases, most of these bispecific mol chain and held in the right sterical position by heterodimer ecules are limited in that they are species specific and recog ization of the epsilon chain with either the CD3 gamma or nize only human antigen, and—due to genetic similarity— delta chain. Anti-CD3 binding domains as part of bispecific likely the chimpanzee counterpart. The advantage of the single chain molecules as provided herein and generated (and preferred embodiment of the invention is the provision of a directed) against a context-independent CD3 epitope provide bispecific single chain antibody comprising a binding domain for a Surprising clinical improvement with regard to T cell exhibiting cross-species specificity to human and non-chim redistribution and, thus, a more favourable safety profile. panzee primate of the CD3 epsilon chain. 25 Without being bound by theory, since the CD3 epitope is Herein described examples for preferred first binding context-independent, forming an autonomous self sufficient domains bind to an N-terminal 1-27 amino acid residue subdomain without much influence on the rest of the CD3 polypeptide fragment of the extracellular domain of CD3 complex, the CD3 binding domain of the bispecific single epsilon. This 1-27 amino acid residue polypeptide fragment chain molecules provided herein induces less allosteric was Surprisingly identified which in contrast to all other 30 changes in CD3 conformation than the conventional CD3 known epitopes of CD3 epsilon described in the art—main binding molecules (like molecules provided in WO99/54440 tains its three-dimensional structural integrity when taken out or WO 04/106380), which recognize context-dependent CD3 of its native environment in the CD3 complex (and optionally epitopes. fused to a heterologous amino acid sequence Such as EpCAM The context-independence of the CD3 epitope which is or an immunoglobulin Fc part). 35 recognized by the CD3 binding domain of the bispecific The present invention, therefore, provides for a bispecific single chain antibodies of the invention, respectively isolated single chain antibody molecule comprising a first binding by the method of the invention, is associated with less or no T domain capable of binding to an epitope of an N-terminal cell redistribution (T cell redistribution equates with an initial 1-27 amino acid residue polypeptide fragment of the extra episode of drop and subsequent recovery of absolute T cell cellular domain of CD3 epsilon (which CD3 epsilon is, for 40 counts) during the starting phase of treatment with said bispe example, taken out of its native environment and/or com cific single chain antibody. This results in a better safety prised by (presented on the surface of) a T-cell) of human and profile of the bispecific single chain antibodies of the inven at least one non-chimpanzee primate CD3 epsilon chain, tion compared to conventional CD3 binding molecules wherein the epitope is part of an amino acid sequence com known in the art, which recognize context-dependent CD3 prised in the group consisting of SEQID NOS. 2, 4, 6, and 8: 45 epitopes. Particularly, because T cell redistribution during the and a second binding domain capable of binding to prostate starting phase of treatment with CD3 binding molecules is a specific membrane antigen (PSMA). Preferred non-chimpan major risk factor for adverse events, like CNS adverse events, Zee primates are mentioned herein elsewhere. At least one (or the bispecific single chain antibodies of the invention by a selection thereof or all) primate(s) selected from Callithrix recognizing a context-independent rather than a context-de jacchus, Saguinus Oedipus, Saimiri Sciureus, and Macaca 50 pendent CD3 epitope has a Substantial safety advantage over fascicularis (either SEQ ID 863 or 864 or both), is (are) the CD3 binding molecules known in the art. Patients with particularily preferred. Macaca mulata, also known as such CNS adverse events related to T cell redistribution dur Rhesus Monkey is also envisaged as another preferred pri ing the starting phase of treatment with conventional CD3 mate. It is thus envisaged that antibodies of the invention bind binding molecules usually suffer from confusion and disori to (are capable of binding to) the context independent epitope 55 entation, in some cases also from urinary incontinence. Con of an N-terminal 1-27 amino acid residue polypeptide frag fusion is a change in mental status in which the patient is not ment of the extracellular domain of CD3 epsilon of human able to think with his or her usual level of clarity. The patient and Callithrix jacchus, Saguinus Oedipus, Saimiri Sciureus, usually has difficulties to concentrate and thinking is not only and Macaca fascicularis (either SEQID 863 or 864 or both), blurred and unclear but often significantly slowed down. and optionally also to Macaca mulatta. A bispecific single 60 Patients with CNS adverse events related to T cell redistribu chain antibody molecule comprising a first binding domain as tion during the starting phase of treatment with conventional defined herein can be obtained (is obtainable by) or can be CD3 binding molecules may also suffer from loss of memory. manufactured in accordance with the protocol set out in the Frequently, the confusion leads to the loss of ability to rec appended Examples (in particular Example 2). To this end, it ognize people, places, time or the date. Feelings of disorien is envisaged to (a) immunize mice with an N-terminal 1-27 65 tation are common in confusion, and the decision-making amino acid residue polypeptide fragment of the extracellular ability is impaired. CNS adverse events related to T cell domain of CD3 epsilon of human and/or Saimirisciureus; (b) redistribution during the starting phase of treatment with US 9,260,522 B2 17 18 conventional CD3 binding molecules may further comprise PSMA, e.g. macaque PSMA, can be determined, for instance, blurred speech and/or word finding difficulties. This disorder by FACS analysis. The FACS analysis is carried out in a way may impair both, the expression and understanding of lan that the respective monoclonal antibody is tested for binding guage as well as reading and writing. Besides urinary incon to human and non-chimpanzee primate cells, e.g. macaque tinence, Vertigo and dizziness may also accompany CNS cells, expressing said human and non-chimpanzee primate adverse events related to T cell redistribution during the start PSMA antigens, respectively. ing phase of treatment with conventional CD3 binding mol As used herein, CD3 epsilon denotes a molecule expressed ecules in some patients. The maintenance of the three-dimen as part of the T cell receptor and has the meaning as typically sional structure within the mentioned 27 amino acid ascribed to it in the prior art. In human, it encompasses in N-terminal polypeptide fragment of CD3 epsilon can be used 10 individual or independently combined form all known CD3 for the generation of preferably human, binding domains subunits, for example CD3 epsilon, CD3 delta, CD3 gamma, which are capable of binding to the N-terminal CD3 epsilon CD3 Zeta, CD3 alpha and CD3 beta. The non-chimpanzee polypeptide fragment in vitro and to the native (CD3 epsilon primate, non-human CD3 antigens as referred to herein are, subunit of the) CD3 complex on T cells in vivo with the same for example, Macaca fascicularis CD3 and Macaca mulatta binding affinity. These data strongly indicate that the N-ter 15 CD3. In Macaca fascicularis, it encompasses CD3 epsilon minal fragment as described herein forms a tertiary confor FN-18 negative and CD3 epsilon FN-18 positive, CD3 mation, which is similar to its structure normally existing in gamma and CD3 delta. In Macaca mulatta, it encompasses vivo. A very sensitive test for the importance of the structural CD3 epsilon, CD3 gamma and CD3 delta. Preferably, said integrity of the amino acids 1-27 of the N-terminal polypep CD3 as used herein is CD3 epsilon. tide fragment of CD3 epsilon was performed. Individual The human CD3 epsilon is indicated in GenBank Acces amino acids of amino acids 1-27 of the N-terminal polypep sion No. NM 000733 and comprises SEQ ID NO. 1. The tide fragment of CD3 epsilon were changed to alanine (ala human CD3 gamma is indicated in GenBank Accession NO. nine Scanning) to test the sensitivity of the amino acids 1-27 NM 000073. The human CD3 delta is indicated in GenBank of the N-terminal polypeptide fragment of CD3 epsilon for Accession No. NM 000732. minor disruptions. 25 The CD3 epsilon"FN-18 negative of Macaca fascicularis Unexpectedly, it has been found that the thus isolated, (i.e. CD3 epsilon not recognized by monoclonal antibody preferably human, bispecific single chain antibody of the FN-18 due to a polymorphism as set forth above) is indicated invention not only recognizes the human N-terminal frag in GenBank Accession No. AB073994. ment of CD3 epsilon, but also the corresponding homologous The CD3 epsilon “FN-18 positive' of Macaca fascicularis fragments of CD3 epsilon of various primates, including 30 (i.e. CD3 epsilon recognized by monoclonal antibody FN-18) New-World Monkeys (Marmoset, Callithrix jacchus, Sagui is indicated in GenBank Accession No. AB073993. The CD3 nus Oedipus, Saimiri sciureus) and Old-World Monkeys gamma of Macaca fascicularis is indicated in GenBank (Macaca fascicularis, also known as Cynomolgus Monkey; Accession No. AB073992. The CD3 delta of Macaca fascicu or Macaca mulatta, also known as Rhesus Monkey). Thus, laris is indicated in GenBank Accession No. AB073991. multi-primate specificity of the bispecific single chain anti 35 The nucleic acid sequences and amino acid sequences of bodies of the invention can be detected. The multi-primate the respective CD3 epsilon, gamma and delta homologs of specificity of the biding domains of the invention is defined Macaca mulatta can be identified and isolated by recombi herein as cross-species specificity. nant techniques described in the art (Sambrook et al. Molecu The amino acid sequence of the aformentioned N-terminal lar Cloning: A Laboratory Manual; Cold Spring Harbor fragments of CD3 epsilon are depicted in SEQ ID No. 2 40 Laboratory Press, 3' edition 2001). This applies mutatis (human), SEQID No. 4 (Callithrix jacchus); SEQID No. 6 mutandis to the CD3 epsilon, gamma and delta homologs of (Saguinus Oedipus); SEQ ID No. 8 (Saimiri Sciureus); SEQ other non-chimpanzee primates as defined herein. The iden ID No. 863 QDGNEEMGSITQTPYQVSISGTTILTC or tification of the amino acid sequence of Callithrix jacchus, SEQ ID No. 864 QDGNEEMGSITQTPYQVSISGTTVILT Saimiri Sciureus and Saguinus Oedipus is described in the (Macaca fascicularis, also known as Cynomolgus Monkey), 45 appended examples. The amino acid sequence of the extra and SEQ ID No. 865 QDGNEEMGSITQTPYHVSISGT cellular domain of the CD3 epsilon of Callithrix jacchus is TVILT (Macaca mulatta, also known as Rhesus Monkey). depicted in SEQ ID NO: 3, the one of Saguinus Oedipus is The term “cross-species specificity' or “interspecies speci depicted in SEQID NO: 5 and the one of Saimirisciureus is ficity' as used herein means binding of a binding domain depicted in SEQID NO: 7. described herein to the same target molecule in humans and 50 In line with the above, the term “epitope' defines an anti non-chimpanzee primates. Thus, "cross-species specificity' genic determinant, which is specifically bound/identified by a or “interspecies specificity' is to be understood as an inter binding domain as defined herein. The binding domain may species reactivity to the same molecule “X” expressed in specifically bind to/interact with conformational or continu different species, but not to a molecule other than “X”. Cross ous epitopes, which are unique for the target structure, e.g. the species specificity of a monoclonal antibody recognizing e.g. 55 human and non-chimpanzee primate CD3 epsilon chain. A human CD3 epsilon, to a non-chimpanzee primate CD3 epsi conformational or discontinuous epitope is characterized for lon, e.g. macaque CD3 epsilon, can be determined, for polypeptide antigens by the presence of two or more discrete instance, by FACS analysis. The FACS analysis is carried out amino acid residues which are separated in the primary in a way that the respective monoclonal antibody is tested for sequence, but come together on the Surface of the molecule binding to human and non-chimpanzee primate cells, e.g. 60 when the polypeptide folds into the native protein/antigen macaque cells, expressing said human and non-chimpanzee (Sela, (1969) Science 166, 1365 and Layer, (1990) Cell 61, primate CD3 epsilon antigens, respectively. An appropriate 553-6). The two or more discrete amino acid residues con assay is shown in the following examples. The above-men tributing to the epitope are present on separate sections of one tioned subject matter applies mutatis mutandis for the target or more polypeptide chain(s). These residues come together antigens PSMA, FAPC, endosialin (TEM1), c-MET and IGF 65 on the surface of the molecule when the polypeptide chain(s) 1R: Cross-species specificity of a monoclonal antibody rec fold(s) into a three-dimensional structure to constitute the ognizing e.g. human PSMA, to a non-chimpanzee primate epitope. In contrast, a continuous or linear epitope consists of US 9,260,522 B2 19 20 two or more discrete amino acid residues, which are present context-independent N-terminal 1-27 amino acid residue in a single linear segment of a polypeptide chain. Within the epitope and which are cross-species specific, i.e. bind to present invention, a “context-dependent CD3 epitope refers human and non-chimpanzee primate CD3 epsilon. to the conformation of said epitope. Such a context-depen As used herein, the term "humanized”, “humanization', dent epitope, localized on the epsilon chain of CD3, can only “human-like' or grammatically related variants thereof are develop its correct conformation if it is embedded within the used interchangeably to refer to a bispecific single chain rest of the epsilon chain and held in the right position by antibody comprising in at least one of its binding domains at heterodimerization of the epsilon chain with either CD3 least one complementarity determining region (“CDR) from gamma or delta chain. In contrast, a context-independent a non-human antibody or fragment thereof. Humanization CD3 epitope as provided herein refers to an N-terminal 1-27 10 approaches are described for example in WO 91/09968 and amino acid residue polypeptide or a functional fragment U.S. Pat. No. 6,407.213. As non-limiting examples, the term thereof of CD3 epsilon. This N-terminal 1-27 amino acid encompasses the case in which a variable region of at least residue polypeptide or a functional fragment thereof main one binding domain comprises a single CDR region, for tains its three-dimensional structural integrity and correct example the third CDR region of the VH(CDRH3), from conformation when taken out of its native environment in the 15 another non-human animal, for example a rodent, as well as CD3 complex. The context-independency of the N-terminal the case in which a or both variable region/s comprise at each 1-27 amino acid residue polypeptide or a functional fragment of their respective first, second and third CDRs the CDRs thereof, which is part of the extracellular domain of CD3 from said non-human animal. In the event that all CDRs of a epsilon, represents, thus, an epitope which is completely dif binding domain of the bispecific single chain antibody have ferent to the epitopes of CD3 epsilon described in connection been replaced by their corresponding equivalents from, for with a method for the preparation of human binding mol example, a rodent, one typically speaks of “CDR-grafting. ecules in WO 2004/106380. Said method used solely and this term is to be understood as being encompassed by the expressed recombinant CD3 epsilon. The conformation of term “humanized' or grammatically related variants thereof this solely expressed recombinant CD3 epsilon differed from as used herein. The term “humanized' or grammatically that adopted in its natural form, that is, the form in which the 25 related variants thereof also encompasses cases in which, in CD3 epsilon subunit of the TCR/CD3 complex exists as part addition to replacement of one or more CDR regions within a of a noncovalent complex with either the CD3 delta or the VH and/or VL of the first and/or second binding domain CD3-gamma subunit of the TCR/CD3 complex. When such further mutation/s (e.g. Substitutions) of at least one single solely expressed recombinant CD3 epsilon protein is used as amino acid residue/s within the framework (“FR) regions an antigen for selection of antibodies from an antibody 30 between the CDRs has/have been effected such that the amino library, antibodies specific for this antigen are identified from acids at that/those positions correspond/s to the amino acid/s the library although such a library does not contain antibodies at that those position/s in the animal from which the CDR with specificity for self-antigens/autoantigens. This is due to regions used for replacement is/are derived. As is known in the fact that solely expressed recombinant CD3 epsilon pro the art, such individual mutations are often made in the frame tein does not exist in vivo; it is not an autoantigen. Conse 35 work regions following CDR-grafting in order to restore the quently, Subpopulations of B cells expressing antibodies spe original binding affinity of the non-human antibody used as a cific for this protein have not been depleted in vivo; an CDR-donor for its target molecule. The term “humanized' antibody library constructed from such B cells would contain may further encompass (an) amino acid Substitution(s) in the genetic material for antibodies specific for solely expressed CDR regions from a non-human animal to the amino acid(s) recombinant CD3 epsilon protein. 40 of a corresponding CDR region from a human antibody, in However, since the context-independent N-terminal 1-27 addition to the amino acid substitutions in the framework amino acid residue polypeptide or a functional fragment regions as described above. thereof is an epitope, which folds in its native form, binding As used herein, the term “homolog”or “homology” is to be domains in line with the present invention cannot be identi understood as follows: Homology among proteins and DNA fied by methods based on the approach described in WO 45 is often concluded on the basis of sequence similarity, espe 2004/106380. Therefore, it could be verified in tests that cially in bioinformatics. For example, in general, if two or binding molecules as disclosed in WO 2004/106380 are not more genes have highly similar DNA sequences, it is likely capable of binding to the N-terminal 1-27 amino acid residues that they are homologous. But sequence similarity may arise of the CD3 epsilon chain. Hence, conventional anti-CD3 from different ancestors: short sequences may be similar by binding molecules or anti-CD3 antibody molecules (e.g. as 50 chance, and sequences may be similar because both were disclosed in WO 99/54.440) bind CD3 epsilon chain at a selected to bind to a particular protein, such as a transcription position which is more C-terminally located than the context factor. Such sequences are similar but not homologous. independent N-terminal 1-27 amino acid residue polypeptide Sequence regions that are homologous are also called con or a functional fragment provided herein. Prior art antibody served. This is not to be confused with conservation in amino molecules OKT3 and UCHT-1 have also a specificity for the 55 acid sequences in which the amino acid at a specific position epsilon-subunit of the TCR/CD3 complex between amino has changed but the physio-chemical properties of the amino acid residues 35 to 85 and, accordingly, the epitope of these acid remain unchanged. Homologous sequences are of two antibodies is also more C-terminally located. In addition, types: orthologous and paralogous. Homologous sequences UCHT-1 binds to the CD3 epsilon chain in a region between are orthologous if they were separated by a speciation event: amino acid residues 43 to 77 (Tunnacliffe, Int. Immunol. 1 60 when a species diverges into two separate species, the diver (1989), 546-50; Kjer-Nielsen, PNAS 101, (2004), 7675 gent copies of a single gene in the resulting species are said to 7680: Salmeron, J. Immunol. 147 (1991), 3047-52). There be orthologous. Orthologs, or orthologous genes, are genes in fore, prior art anti-CD3 molecules do not bind to and are not different species that are similar to each other because they directed against the herein defined context-independent originated from a common ancestor. The strongest evidence N-terminal 1-27 amino acid residue epitope (or a functional 65 that two similar genes are orthologous is the result of a phy fragment thereof). In particular, the state of the art fails to logenetic analysis of the gene lineage. Genes that are found provide anti-CD3 molecules which specifically binds to the within one Glade are orthologs, descended from a common US 9,260,522 B2 21 22 ancestor. Orthologs often, but not always, have the same mainly African but include the diverse genus of macaques function. Orthologous sequences provide useful information which are Asian and North African; and the Colobinae, which in taxonomic classification studies of organisms. The pattern include most of the Asian genera but also the African colobus of genetic divergence can be used to trace the relatedness of monkeys. organisms. Two organisms that are very closely related are 5 Specifically, within the subfamily Cercopithecinae, an likely to display very similar DNA sequences between two advantageous non-chimpanzee primate may be from the orthologs. Conversely, an organism that is further removed Tribe Cercopithecini, within the genus Allenopithecus evolutionarily from another organism is likely to display a (Allen's Swamp Monkey, Allenopithecus nigroviridis); greater divergence in the sequence of the orthologs being within the genus Miopithecus (Angolan Talapoin, Miopith studied. Homologous sequences are paralogous if they were 10 ecus talapoin; Gabon Talapoin, Miopithecus Ogouensis); separated by a gene duplication event: if a gene in an organ within the genus Erythrocebus (Patas Monkey, Erythrocebus ism is duplicated to occupy two different positions in the patas); within the genus Chlorocebus (Green Monkey, Chlo same genome, then the two copies are paralogous. A set of rocebus sabaceus: Grivet, Chlorocebus aethiops; Bale Moun sequences that are paralogous are called paralogs of each tains Vervet, Chlorocebus diamdiamensis; Tantalus Monkey, other. Paralogs typically have the same or similar function, 15 Chlorocebus tantalus; Vervet Monkey, Chlorocebus but sometimes do not: due to lack of the original selective pygerythrus; Malbrouck, Chlorocebus cynosuros); or within pressure upon one copy of the duplicated gene, this copy is the genus Cercopithecus (Dryas Monkey or Salongo Monkey, free to mutate and acquire new functions. An example can be Cercopithecus dryas; Diana Monkey, Cercopithecus diana; found in rodents such as rats and mice. Rodents have a pair of Roloway Monkey, Cercopithecus rolloway; Greater Spot paralogous insulin genes, although it is unclear if any diver nosed Monkey, Cercopithecus nictitans; Blue Monkey, Cer gence in function has occurred. Paralogous genes often copithecus mitis; Silver Monkey, Cercopithecus doggetti; belong to the same species, but this is not necessary: for Golden Monkey, Cercopithecus kanditi: Sykes's Monkey, example, the hemoglobin gene of humans and the myoglobin Cercopithecus albogularis; Mona Monkey, Cercopithecus gene of chimpanzees are paralogs. This is a common problem mona; Campbell's Mona Monkey, Cercopithecus Campbelli: in bioinformatics: when genomes of different species have 25 Lowe's Mona Monkey, Cercopithecus lowei: Crested Mona been sequenced and homologous genes have been found, one Monkey, Cercopithecus pogonias; Wolfs Mona Monkey, can not immediately conclude that these genes have the same Cercopithecus wolfi: Dent's Mona Monkey, Cercopithecus or similar function, as they could be paralogs whose function denti; Lesser Spot-nosed Monkey, Cercopithecus petaurista; has diverged. White-throated Guenon, Cercopithecus erythrogaster; Sclat As used herein, a “non-chimpanzee primate' or “non 30 er's Guenon, Cercopithecus Sclateri; Red-eared Guenon, chimp primate' or grammatical variants thereof refers to any Cercopithecus erythrotis; Moustached Guenon, Cercopith primate animal (i.e. not human) other than chimpanzee, i.e. ecus cephus; Red-tailed Monkey, Cercopithecus ascanius: other than an animal of belonging to the genus Pan, and L'Hoest's Monkey, Cercopithecus Ihoesti: Preuss's Monkey, including the species Pan paniscus and Pan troglodytes, also Cercopithecus preussi: Sun-tailed Monkey, Cercopithecus known as Anthropopithecus troglodytes or Simia satyrus. It 35 solatus; Hamlyn's Monkey or Owl-faced Monkey, Cercop will be understood, however, that it is possible that the anti ithecus hamlyni; De Brazza's Monkey, Cercopithecus bodies of the invention can also bind with their first and/or neglectus). second binding domain to the respective epitopes/fragments Alternatively, an advantageous non-chimpanzee primate, etc. of said chimpanzees. The intention is merely to avoid also within the subfamily Cercopithecinae but within the animal tests which are carried out with chimpanzees, if 40 Tribe Papionini, may be from within the genus Macaca (Bar desired. It is thus also envisaged that in another embodiment bary Macaque, Macaca sylvanus; Lion-tailed Macaque, the antibodies of the present invention also bind with their Macaca silenus; Southern Pig-tailed Macaque or Beruk, first and/or second binding domain to the respective epitopes Macaca nemestrina; Northern Pig-tailed Macaque, Macaca of chimpanzees. A "primate”, “primate species”, “primates' leonina; Pagai Island Macaque or Bokkoi, Macaca pagensis; or grammatical variants thereofdenote?s an order of eutherian 45 Siberut Macaque, Macaca Siberu; Moor Macaque, Macaca mammals divided into the two Suborders of prosimians and maura; Booted Macaque, Macaca ochreata; Tonkean anthropoids and comprising apes, monkeys and lemurs. Spe Macaque, Macaca tonkeana; Heck's Macaque, Macaca cifically, “primates' as used herein comprises the suborder hecki; Gorontalo Macaque, Macaca nigriscens; Celebes Strepsirrhini (non-tarsier prosimians), including the Crested Macaque or Black “Ape', Macaca nigra; Cynomol infraorder Lemuriformes (itself including the superfamilies 50 gus monkey or Crab-eating Macaque or Long-tailed Chemogaleoidea and Lemuroidea), the infraorder Chiromyi Macaque or Kera, Macaca fascicularis; Stump-tailed formes (itself including the family Daubentoniidae) and the Macaque or Bear Macaque, Macaca arctoides; Rhesus infraorder Lorisiformes (itself including the families Macaque, Macaca mulatta. Formosan Rock Macaque, Lorisidae and Galagidae). “Primates' as used herein also Macaca cyclopis; Japanese Macaque, Macaca fiscata; Toque comprises the suborder Haplorrhini, including the infraorder 55 Macaque, Macaca sinica; Bonnet Macaque, Macaca radiata; Tarsiiformes (itself including the family Tarsiidae), the Barbary Macaque, Macaca sylvanmus; Assam Macaque, infraorder Simiiformes (itself including the Platyrrhini, or Macaca assamensis; Tibetan Macaque or Milne-Edwards New-World monkeys, and the Catarrhini, including the Cer Macaque, Macaca thibetana; Arunachal Macaque or Mun copithecidea, or Old-World Monkeys). Zala, Macaca munzala); within the genus Lophocebus (Gray The non-chimpanzee primate species may be understood 60 cheeked Mangabey, Lophocebus albigena, Lophocebus albi within the meaning of the invention to be a lemur, a tarsier, a gena albigena, Lophocebus albigena Osmani Lophocebus gibbon, a marmoset (belonging to New-World Monkeys of albigena johnstoni; Black Crested Mangabey, Lophocebus the family Cebidae) or an Old-World Monkey (belonging to aterrimus; Opdenbosch's Mangabey, Lophocebus opdenbos the superfamily Cercopithecoidea). chi; Highland Mangabey, Lophocebus kipunji); within the As used herein, an “Old-World Monkey' comprises any 65 genus Papio (Hamadryas Baboon, Papio hamadryas; Guinea monkey falling in the Superfamily Cercopithecoidea, itself Baboon, Papio papio; Olive Baboon, Papio anubis; Yellow subdivided into the families: the Cercopithecinae, which are Baboon, Papio cynocephalus; Chacma Baboon, Papio ursi US 9,260,522 B2 23 24 nus); within the genus Theropithecus (Gelada, Theropithecus cunda; Mentawai Langur or Joja, Presbytis potenziani; gelada); within the genus Cercocebus (Sooty Mangabey, Natuna Island Surili, Presbytis natunae). Cercocebus atys, Cercocebus atys atys, Cercocebus atys Within the subfamily Colobinae, an advantageous non lunulatus; Collared Mangabey, Cercocebus torquatus; Agile chimpanzee primate may alternatively be from the Odd Mangabey, Cercocebus agilis; Golden-bellied Mangabey, Nosed group, within the genus Pygathrix (Red-shanked Cercocebus chrysogaster, Tana River Mangabey, Cercocebus Douc, Pygathrix nemaeus; Black-shanked Douc, Pygathrix galeritus; Sanje Mangabey, Cercocebus sanjei); or within the nigripes; Gray-shanked Douc, Pygathrix cinerea); within the genus Mandrillus (Mandrill, Mandrillus sphinx: Drill, Man genus Rhinopithecus (Golden Snub-nosed Monkey, Rhinop drillus leucophaeus). ithecus roxellana; Black Snub-nosed Monkey, Rhinopithecus Most preferred is Macaca fascicularis (also known as 10 bieti; Gray Snub-nosed Monkey, Rhinopithecus brelichi: Cynomolgus monkey and, therefore, in the Examples named Tonkin Snub-nosed Langur, Rhinopithecus avunculus); “Cynomolgus”) and Macaca mulatta (rhesus monkey, named within the genus Nasalis (Proboscis Monkey, Nasalis larva “rhesus'). tus); or within the genus Simias (Pig-tailed Langur, Simias Within the subfamily Colobinae, an advantageous non concolor). chimpanzee primate may be from the African group, within 15 As used herein, the term “marmoset denotes any New the genus Colobus (Black Colobus, Colobus satanas; Angola World Monkeys of the genus Callithrix, for example belong Colobus, Colobus angolensis; King Colobus, Colobus ing to the Atlantic marmosets of Subgenus Callithrix (sic!) polykomos; Ursine Colobus, Colobus vellerosus; Mantled (Common Marmoset, Callithrix (Callithrix) jacchus; Black Guereza, Colobus guereza); within the genus Piliocolobus tufted Marmoset, Callithrix (Callithrix) penicillata; Wied’s (Western Red Colobus, Piliocolobus badius, Piliocolobus Marmoset, Callithrix (Callithrix) kuhlii; White-headed Mar badius badius, Piliocolobus badius temminckii. Piliocolobus moset, Callithrix (Callithrix) geoffroyi; Buffy-headed Mar badius waldronae; Pennant's Colobus, Piliocolobus pennan moset, Callithrix (Callithrix) flaviceps; Buffy-tufted Marmo tii, Piliocolobus pennantipennantii, Piliocolobus pennantii set, Callithrix (Callithrix) aurita); belonging to the epieni, Piliocolobus pennantibouvieri; Preuss's Red Colo Amazonian marmosets of subgenus Mico (Rio Acari Marmo bus, Piliocolobus preussi: Thollon's Red Colobus, Piliocolo 25 set, Callithrix (Mico) acariensis; Manicore Marmoset, Cal bus tholloni; Central African Red Colobus, Piliocolobus foai; lithrix (Mico) manicorensis; Silvery Marmoset, Callithrix Piliocolobus foai foai; Piliocolobus foai ellioti, Piliocolobus (Mico) argentata: White Marmoset, Callithrix (Mico) foai oustaleti; Piliocolobus foai Semlikiensis, Piliocolobus leucippe: Emilia’s Marmoset, Callithrix (Mico) emiliae: foai partmentierorum; Ugandan Red Colobus, Piliocolobus Black-headed Marmoset, Callithrix (Mico) nigriceps; Mar tephrosceles; Uzyngwa Red Colobus, Piliocolobus gordo 30 ca's Marmoset, Callithrix (Mico)marcai: Black-tailed Mar norum; Zanzibar Red Colobus, Piliocolobus kirkii; Tana moset, Callithrix (Mico) melanura: Santarem Marmoset, River Red Colobus, Piliocolobus rufomitratus); or within the Callithrix (Mico) humerallifera; Maues Marmoset, Callithrix genus Procolobus (Olive Colobus, Procolobus verus). Within (Mico) mauesi; Gold-and-white Marmoset, Callithrix (Mico) the Subfamily Colobinae, an advantageous non-chimpanzee chrysoleuca; Hershkovitz's Marmoset, Callithrix (Mico) primate may alternatively be from the Langur (leaf monkey) 35 intermedia; Satere Marmoset, Callithrix (Mico) saterei); group, within the genus Semnopithecus (Nepal Gray Langur, Roosmallens' Dwarf Marmoset belonging to the Subgenus Semnopithecus schistaceus; Kashmir Gray Langur, Semnop Callibella (Callithrix (Callibella) humilis); or the Pygmy ithecus ajax, Tarai Gray Langur, Semnopithecus hector; Marmoset belonging to the subgenus Cebuella (Callithrix Northern Plains Gray Langur, Semnopithecus entellus; (Cebuella) pygmaea). Black-footed Gray Langur, Semnopithecus hypoleucos; 40 Other genera of the New-World Monkeys comprise tama Southern Plains Gray Langur, Semnopithecus dussumieri; rins of the genus Saguinus (comprising the S. Oedipus-group, Tufted Gray Langur, Semnopithecus priam); within the T. the S. midas group, the S. nigricollis group, the S. myStax vetulus group or the genus Trachypithecus (Purple-faced Lan group, the S. bicolor group and the S. inustus group) and gur, Trachypithecus vetulus; Nilgiri Langur, Trachypithecus squirrel monkeys of the genus Samiri (e.g. Saimirisciureus, johnii); within the T. Cristatus group of the genus Trachypith 45 Saimiri Oerstedii, Saimiri ustus, Saimiri boliviensis, Saimiri ecus (Javan Lutung, Trachypithecus auratus; Silvery Leaf vanzolini). Monkey or Silvery Lutung, Trachypithecus cristatus; Advantageously, the present invention provides also target Indochinese Lutung, Trachypithecus germaini; Tenasserim antigenxCD3 bispecific single chain antibodies comprising a Lutung, Trachypithecus barbel); within the T obscurus group second binding domain which binds both to the human target of the genus Trachypithecus (Dusky Leaf Monkey or Spec 50 antigen and to the macaque target antigen homolog, i.e. the tacled Leaf Monkey, Trachypithecus obscurus; Phayre's Leaf homolog of a non-chimpanzee primate. In a preferred Monkey, Trachypithecus phayrei); within the T. pileatus embodiment, the bispecific single chain antibody thus com group of the genus Trachypithecus (Capped Langur, Trachyp prises a second binding domain exhibiting cross-species ithecus pileatus; Shortridge's Langur, Trachypithecus short specificity to the human and a non-chimpanzee primate target ridgei; Gee's Golden Langur, Trachypithecus geei); within 55 antigen. In this case, the identical bispecific single chain the T francoisi group of the genus Trachypithecus (Francois antibody molecule can be used both for preclinical evaluation Langur, Trachypithecus francoisi; Hatinh Langur, Trachyp of safety, activity and/or pharmacokinetic profile of these ithecus hatinhensis; White-headed Langur, Trachypithecus binding domains in primates and as drug in humans. Put in poliocephalus; Laotian Langur, Trachypithecus laotum; other words, the same molecule can be used in preclinical Delacours Langur, Trachypithecus delacouri; Indochinese 60 animal studies as well as in clinical studies in humans. This Black Langur, Trachypithecus ebenus); or within the genus leads to highly comparable results and a much-increased Presbytis (Sumatran Surili, Presbytis melalophos; Banded predictive power of the animal studies compared to species Surili, Presbytis femoralis; Sarawak Surili, Presbytis chry specific surrogate molecules. Since both the CD3 and the somelas; White-thighed Surili, Presbytis Siamensis; White target antigen binding domain of the target antigenxCD3 fronted Surili, Presbytis frontata; Javan Surili, Presbytis 65 bispecific single chain antibody of the invention are cross comata; Thomas’s Langur, Presbytis thomasi. Hose's Lan species specific, i.e. reactive with the human and non-chim gur, Presbytis hosei: Maroon Leaf Monkey, Presbytis rubi panzee primates antigens, it can be used both for preclinical US 9,260,522 B2 25 26 evaluation of safety, activity and/or pharmacokinetic profile Another major advantage of the, preferably human, target of these binding domains in primates and in the identical antigenxCD3 bispecific single chain antibody of the inven form—as drug in humans. It will be understood that in a tion is its applicability for preclinical testing in various pri preferred embodiment, the cross-species specificity of the mates. The behavior of a drug candidate in animals should first and second binding domain of the antibodies of the ideally be indicative of the expected behavior of this drug invention is identical. candidate upon administration to humans. As a result, the data It has been found in the present invention that it is possible obtained from Such preclinical testing should therefore gen to generate a, preferably human, target antigenxCD3 bispe erally have a highly predictive power for the human case. cific single chain antibody wherein the identical molecule can However, as learned from the tragic outcome of the recent be used in preclinical animal testing, as well as clinical stud 10 Phase I clinical trial on TGN1412 (a CD28 monoclonal anti ies and even in therapy in human. This is due to the unex body), a drug candidate may act differently in a primate pected identification of the, preferably human, target anti species than in humans: Whereas in preclinical testing of said genxCD3 bispecific single chain antibody, which, in addition antibody no or only limited adverse effects have been to binding to human CD3 epsilon and target antigen, respec observed in animal studies performed with cynomolgus mon tively, (and due to genetic similarity likely to the chimpanzee 15 keys, six human patients developed multiple organ failure counterpart), also binds to the homologs of said antigens of upon administration of said antibody (Lancet 368 (2006), non-chimpanzee primates, including New-World Monkeys 2206-7). The results of these dramatic, non-desired negative and Old-World Monkeys. The preferably human, target anti events suggest that it may not be sufficient to limit preclinical genxCD3 bispecific single chain antibody of the invention testing to only one (non-chimpanzee primate) species. The can be used as therapeutic agent against various diseases, fact that the target antigenxCD3 bispecific single chain anti including, but not limited, to cancer. In view of the above, the body of the invention binds to a series of New-World and need to construct a surrogate target antigenxCD3 bispecific Old-World Monkeys may help to overcome the problems single chain antibody for testing in a phylogenetic distant faced in the case mentioned above. Accordingly, the present (from humans) species disappears. As a result, the identical invention provides means and methods for minimizing spe molecule can be used in animal preclinical testing as is 25 cies differences in effects when drugs for human therapy are intended to be administered to humans in clinical testing as being developed and tested. well as following market approval and therapeutic drug With the, preferably human, cross-species specific target administration. The ability to use the same molecule for pre antigenxCD3 bispecific single chain antibody of the inven clinical animal testing as in later administration to humans tion it is also no longer necessary to adapt the test animal to virtually eliminates, or at least greatly reduces, the danger 30 the drug candidate intended for administration to humans, that the data obtained in preclinical animal testing have lim Such as e.g. the creation of transgenic animals. The, prefer ited applicability to the human case. In short, obtaining pre ably human, target antigenxCD3 bispecific single chain anti clinical safety data in animals using the same molecule as will body of the invention exhibiting cross-species specificity actually be administered to humans does much to ensure the according to the uses and the methods of invention can be applicability of the data to a human-relevant scenario. In 35 directly used for preclinical testing in non-chimpanzee pri contrast, in conventional approaches using Surrogate mol mates, without any genetic manipulation of the animals. As ecules, said Surrogate molecules have to be molecularly well known to those skilled in the art, approaches in which the adapted to the animal test system used for preclinical safety test animal is adapted to the drug candidate always bear the assessment. Thus, the molecule to be used in human therapy risk that the results obtained in the preclinical safety testing in fact differs in sequence and also likely in structure from the 40 are less representative and predictive for humans due to the Surrogate molecule used in preclinical testing in pharmaco modification of the animal. For example, in transgenic ani kinetic parameters and/or biological activity, with the conse mals, the proteins encoded by the transgenes are often highly quence that data obtained in preclinical animal testing have over-expressed. Thus, data obtained for the biological activity limited applicability/transferability to the human case. The of an antibody against this protein antigen may be limited in use of surrogate molecules requires the construction, produc 45 their predictive value for humans in which the protein is tion, purification and characterization of a completely new expressed at much lower, more physiological levels. construct. This leads to additional development costs and A further advantage of the uses of the preferably human time necessary to obtain that molecule. In Sum, Surrogates target antigenxCD3 bispecific single chain antibody of the have to be developed separately in addition to the actual drug invention exhibiting cross-species specificity is the fact that to be used in human therapy, so that two lines of development 50 chimpanzees as an endangered species are avoided for animal for two molecules have to be carried out. Therefore, a major testing. Chimpanzees are the closest relatives to humans and advantage of the, preferably human, target antigenxCD3 were recently grouped into the family of hominids based on bispecific single chain antibody of the invention exhibiting the genome sequencing data (Wildman et al., PNAS 100 cross-species specificity described herein is that the identical (2003), 7181). Therefore, data obtained with chimpanzee is molecule can be used for therapeutic agents in humans and in 55 generally considered to be highly predictive for humans. preclinical animal testing. However, due to their status as endangered species, the num It is preferred that at least one of said first or second binding ber of chimpanzees, which can be used for medical experi domains of the bispecific single chain antibody of the inven ments, is highly restricted. As stated above, maintenance of tion is CDR-grafted, humanized or human, as set forth in chimpanzees for animal testing is therefore both costly and more detail below. Preferably, both the first and second bind 60 ethically problematic. The uses of the, preferably human, ing domains of the bispecific single chain antibody of the target antigenxCD3 bispecific single chain antibody of the invention are CDR-grafted, humanized or human. For the invention avoid both ethical objections and financial burden preferably human, target antigenxCD3 bispecific single during preclinical testing without prejudicing the quality, i.e. chain antibody of the invention, the generation of an immune applicability, of the animal testing data obtained. In light of reaction against said binding molecule is excluded to the 65 this, the uses of the, preferably human, target antigenxCD3 maximum possible extent upon administration of the mol bispecific single chain antibody of the invention provide for a ecule to human patients. reasonable alternative for studies in chimpanzees. US 9,260,522 B2 27 28 A still further advantage of the, preferably human, target Moreover, the cytotoxic activity of the target antigenxCD3 antigenxCD3 bispecific single chain antibody of the inven bispecific single chainantibody of the invention is higher than tion is the ability of extracting multiple blood samples when the activity of antibodies described in the art for the exempli using it as part of animal preclinical testing, for example in the fied targets. course of pharmacokinetic animal studies. Multiple blood 5 It is further preferred for the method of the invention that extractions can be much more readily obtained with a non the first binding domain capable of binding to an epitope of chimpanzee primate than with lower animals, e.g. a mouse. human and non-chimpanzee primate CD3e chain comprises a The extraction of multiple blood samples allows continuous VL region comprising CDR-L1, CDR-L2 and CDR-L3 testing of blood parameters for the determination of the bio selected from: logical effects induced by the, preferably human, target anti 10 genxCD3 bispecific single chain antibody of the invention. (a) CDR-L1 as depicted in SEQ ID NO. 27, CDR-L2 as Furthermore, the extraction of multiple blood samples depicted in SEQ ID NO. 28 and CDR-L3 as depicted in enables the researcher to evaluate the pharmacokinetic profile SEQID NO. 29; of the, preferably human, target antigenxCD3 bispecific (b) CDR-L1 as depicted in SEQ ID NO. 117, CDR-L2 as single chain antibody of the invention as defined herein. In 15 depicted in SEQID NO. 118 and CDR-L3 as depicted in addition, potential side effects, which may be induced by SEQID NO. 119; and said, preferably human, target antigenxCD3 bispecific single (c) CDR-L1 as depicted in SEQ ID NO. 153, CDR-L2 as chain antibody of the invention reflected in blood parameters depicted in SEQID NO. 154 and CDR-L3 as depicted in can be measured in different blood samples extracted during SEQID NO. 155. the course of the administration of said antibody. This allows More preferably, the first binding domain capable of bind the determination of the potential toxicity profile of the, pref ing to an epitope of human and non-chimpanzee primate erably human, target antigenxCD3 bispecific single chain CD3e chain comprises a VL region selected from the group antibody of the invention as defined herein. consisting of a VL region as depicted in SEQID NO. 35, 39, The advantages of the, preferably human, target antigenx 125, 129, 161 or 165. CD3 bispecific single chain antibody of the invention as 25 It is alternatively preferred for the method of the invention defined herein exhibiting cross-species specificity may be that the first binding domain capable of binding to an epitope briefly summarized as follows: of human and non-chimpanzee primate CD3e chain com First, the preferably human, target antigenxCD3 bispe prises aVH region comprising CDR-H1, CDR-H2 and CDR cific single chain antibody of the invention as defined herein H3 selected from: used in preclinical testing is the same as the one used in 30 (a) CDR-H1 as depicted in SEQ ID NO. 12, CDR-H2 as human therapy. Thus, it is no longer necessary to develop two depicted in SEQ ID NO. 13 and CDR-H3 as depicted in independent molecules, which may differ in their pharmaco SEQID NO. 14; kinetic properties and biological activity. This is highly (b) CDR-H1 as depicted in SEQ ID NO. 30, CDR-H2 as advantageous in that e.g. the pharmacokinetic results are depicted in SEQ ID NO. 31 and CDR-H3 as depicted in more directly transferable and applicable to the human setting 35 SEQID NO. 32: than e.g. in conventional Surrogate approaches. (c) CDR-H1 as depicted in SEQ ID NO. 48, CDR-H2 as Second, the uses of the, preferably human, target antigenx depicted in SEQ ID NO. 49 and CDR-H3 as depicted in CD3 bispecific single chain antibody of the invention as SEQID NO. 50: defined herein for the preparation of therapeutics in human is (d) CDR-H1 as depicted in SEQ ID NO. 66, CDR-H2 as less cost- and labor-intensive than Surrogate approaches. 40 depicted in SEQ ID NO. 67 and CDR-H3 as depicted in Third, the, preferably human, target antigenxCD3 bispe SEQID NO. 68: cific single chain antibody of the invention as defined herein (e) CDR-H1 as depicted in SEQ ID NO. 84, CDR-H2 as can be used for preclinical testing not only in one primate depicted in SEQ ID NO. 85 and CDR-H3 as depicted in species, but in a series of different primate species, thereby SEQID NO.86; limiting the risk of potential species differences between 45 (f) CDR-H1 as depicted in SEQ ID NO. 102, CDR-H2 as primates and human. depicted in SEQID NO. 103 and CDR-H3 as depicted in Fourth, chimpanzee as an endangered species for animal SEQID NO. 104; testing can be avoided if desired. (g) CDR-H1 as depicted in SEQ ID NO. 120, CDR-H2 as Fifth, multiple blood samples can be extracted for exten depicted in SEQID NO. 121 and CDR-H3 as depicted in sive pharmacokinetic studies. 50 SEQID NO. 122; Sixth, due to the human origin of the preferably human, (h) CDR-H1 as depicted in SEQ ID NO. 138, CDR-H2 as binding molecules according to a preferred embodiment of depicted in SEQID NO. 139 and CDR-H3 as depicted in the invention, the generation of an immune reaction against SEQID NO. 140; said binding molecules is minimalized when administered to (i) CDR-H1 as depicted in SEQ ID NO. 156, CDR-H2 as human patients. Induction of an immune response with anti 55 depicted in SEQID NO. 157 and CDR-H3 as depicted in bodies specific for a drug candidate derived from a non SEQID NO. 158; and human species as e.g. a mouse leading to the development of (i) CDR-H1 as depicted in SEQ ID NO. 174, CDR-H2 as human-anti-mouse antibodies (HAMAS) against therapeutic depicted in SEQID NO. 175 and CDR-H3 as depicted in molecules of murine origin is excluded. SEQID NO. 176. Last but not least, the therapeutic use of the target antigenx 60 More preferably, the binding domain capable of binding to CD3 bispecific single chain antibody of the invention pro an epitope of human and non-chimpanzee primate CD3e vides a novel and inventive therapeutic approach for cancer, chain comprises a VH region selected from the group con preferably solid tumors, more preferably carcinomas and sisting of a VH region as depicted in SEQID NO. 15, 19, 33, prostate cancer. As shown in the following examples, the 37, 51,55, 69,73, 87,91, 105,109, 123, 127, 141, 145, 159, target antigenxCD3 bispecific single chain antibody of the 65 163, 177 or 181. invention provides an advantageous tool in order to kill target It is preferred for the method of the invention that the first antigen-expressing human target cells, e.g. cancer cells. binding domain capable of binding to an epitope of human US 9,260,522 B2 29 30 and non-chimpanzee primate CD3e chain comprises a VL three EGF-like domains (SEQ ID NO: 441), or the Sushi/ region and a VH region selected from the group consisting of SCR/CCP domain (SEQ ID NO: 442) of the extracellular (a) a VL region as depicted in SEQID NO. 17 or 21 and a VH domain of TEM1. region as depicted in SEQID NO. 15 or 19: According to a further alternatively preferred embodiment (b) a VL region as depicted in SEQID NO. 35 or 39 and a VH of the method of the invention the second binding domain region as depicted in SEQID NO. 33 or 37; binds to epitopes/binding sites in the extracellular domain of (c) a VL region as depicted in SEQID NO. 53 or 57 and a VH IGF-1R. This target antigen and its expression characteristics region as depicted in SEQID NO. 51 or 55; have been described herein above. For endosialina extracel (d) a VL region as depicted in SEQID NO. 71 or 75 and a VH lular domain consisting of 905 aa and the following sequen 10 tial arrangement of independently folded extracellular region as depicted in SEQID NO. 69 or 73: domains from membrane-proximal to membrane-distal is (e) a VL region as depicted in SEQID NO. 89 or 93 and a VH described in the art: three fibronectin type III domains of region as depicted in SEQID NO. 87 or 91; together 447 aa (residues 460-906), an L2 domain of 160 aa (f) a VL region as depicted in SEQID NO. 107 or 111 and a (residues 300-459), a cystein-rich domain of 149aa (residues VH region as depicted in SEQID NO. 105 or 109: 15 151-299) and an L1 domain of 150 aa (residues 1-150). (g) a VL region as depicted in SEQID NO. 125 or 129 and a Accordingly, it is preferred for the method of the invention VH region as depicted in SEQID NO. 123 or 127: that the second binging domain binds to epitopes/binding (h) a VL region as depicted in SEQID NO. 143 or 147 and a sites in the three fibronectin type III domains (SEQ ID NO: VH region as depicted in SEQID NO. 141 or 145; 444), and the L2 domain (SEQID NO: 445) of the extracel (i) a VL region as depicted in SEQID NO. 161 or 165 and a lular domain of IGF-1R. VH region as depicted in SEQID NO. 159 or 163; and An alternative embodiment of the invention relates to a (i) a VL region as depicted in SEQID NO. 179 or 183 and a bispecific single chain antibody comprising a first domain VH region as depicted in SEQID NO. 177 or 181. binding domain capable of binding to CD3 epsilon (CD3e) of More preferably, the first binding domain capable of bind human and non-chimpanzee primate and a second domain ing to an epitope of human and non-chimpanzee primate 25 binding domain capable of binding to the extracellular CD3e chain comprises an amino acid sequence selected from domain of the mutated human PSMA having an amino acid the group consisting of SEQID NOs: 23, 25, 41, 43, 59, 61. sequence as depicted in SEQ ID NO: 447 but not to the 77,79, 95, 97,113, 115, 131, 133, 149, 151,167, 169, 185 or extracellular domain of the rodent PSMA. In other words, the 187. bispecific antibody of the invention specifically binds to As already discussed herein above, it is also preferred for 30 membrane proximal epitopes, i.e. epitopes formed only by the method of the invention that also the second binding amino acid resides of the extracellular domain of PSMA, the domain binds to epitopes/binding sites in the extracellular alpha C-atom of which has a distance of less than 60 A from domain of a high molecular weight antigen of human and the reference C-atom (the alpha C-atom of the 13' aa as non-chimpanzee primate. counted from the junction of transmembrane and extracellu In a preferred embodiment of the method of the invention 35 lar region). The specific Superior characteristics of these the second binding domain binds to epitopes/binding sites in PSMAxCD3 bispecific single chain antibodies have been the extracellular domain of c-MET. This target antigen and its described herein above. Moreover, corresponding antibodies expression characteristics have been described herein above. are exemplified and characterized in the appended examples. The MET tyrosine kinase receptor with an extracellular It is preferred for the bispecific ingle chain antibody com region of 908 aa with the following sequential arrangement of 40 prising a first domain binding domain capable of binding to independently folded extracellular domains from membrane CD3 epsilon (CD3e) and a second domain binding domain proximal to membrane-distal: Four Ig domains of together capable of binding to the extracellular domain of the mutated 362 aa (residues 563-924), a cystein-rich domain of 42 aa human PSMA having an amino acid sequence as depicted in (residues 520-561), the beta-chain of a sema domain of 212 aa SEQ ID NO: 447 but not to the extracellular domain of the (residues 308-519) and the alpha-chain of the sema domain of 45 rodent PSMA that the first domain capable of binding to an 282 aa (residues 25-307). Accordingly, it is preferred for the epitope of human and non-chimpanzee primate CD3e chain method of the invention, that the second binging domain comprises an amino acid sequence selected from the group binds to epitopes/binding sites in the four Ig domains (SEQ consisting of SEQID NOs: 23, 25, 41, 43, 59, 61, 77,79, 95, IDNO:436), a cystein-rich domain (SEQID NO:437), or the 97, 113, 115, 131, 133, 149, 151, 167, 169, 185 or 187. It is beta-chain of a sema domain (SEQID NO: 438) of the extra 50 preferred for the bispecific single chain antibodies that the cellular domain of c-MET, which are all below the 640 aa second domain comprises a VL region comprising CDR-L1, threshold. CDR-L2 and CDR-L3 selected from: In an alternatively preferred embodiment of the method of (a) CDR-L1 as depicted in SEQ ID NO. 269, CDR-L2 as the invention the second binding domain binds to epitopes/ depicted in SEQ ID NO: 270 and CDR-L3 as depicted in binding sites in the extracellular domain of endosialin 55 SEQID NO. 271; (TEM1). This target antigen and its expression characteristics (b) CDR-L1 as depicted in SEQ ID NO. 283, CDR-L2 as have been described herein above. For endosialina extracel depicted in SEQ ID NO: 284 and CDR-L3 as depicted in lular domain consisting of 665 aa and the following sequen SEQID NO. 285: tial arrangement of independently folded extracellular (c) CDR-L1 as depicted in SEQ ID NO. 297, CDR-L2 as domains from membrane-proximal to membrane-distal is 60 depicted in SEQ ID NO: 298 and CDR-L3 as depicted in described in the art: a mucin domain of 326 aa (residues SEQID NO. 299; 360-685), three EGF-like domains of together 116 aa (resi (d) CDR-L1 as depicted in SEQ ID NO. 311, CDR-L2 as dues 235-350), a Sushi/SCR/CCP domain of 55 aa (residues depicted in SEQ ID NO:312 and CDR-L3 as depicted in 176-230) and a C-type lectin domain of 129 aa (residues SEQID NO. 313; 29-157). Accordingly, it is preferred for the method of the 65 (e) CDR-L1 as depicted in SEQ ID NO. 325, CDR-L2 as invention that the second binging domain binds to epitopes/ depicted in SEQID NO. 326 and CDR-L3 as depicted in binding sites in the mucin domain (SEQ ID NO: 440), the SEQID NO. 327; US 9,260,522 B2 31 32 (f) CDR-L1 as depicted in SEQ ID NO. 255, CDR-L2 as (a) anamino acid sequence as depicted in any of SEQID NOS. depicted in SEQID NO. 256 and CDR-L3 as depicted in 280, 294,308,322, 336, 266 or 487; SEQID NO. 257; and (b) an amino acid sequence encoded by a nucleic acid (g) CDR-L1 as depicted in SEQ ID NO. 481, CDR-L2 as sequence as depicted in any of SEQID NOs: 281,295,309, depicted in SEQID NO. 482 and CDR-L3 as depicted in 5 267, 323,337 or 488; and SEQ ID NO. 483. (c) an amino acid sequence at least 90% identical, more It is also preferred for the bispecific single chain antibodies preferred at least 95 identical, most preferred at least 96% of the invention that the second domain comprises a VH identical to the amino acid sequence of (a) or (b). region comprising CDR-H1, CDR-H2 and CDR-H3 selected The invention relates to a bispecific single chain antibody from: 10 molecule comprising an amino acid sequence as depicted in any of SEQID NOs: 280, 294, 266,308,322,336 or 487, as (a) CDR-H1 as depicted in SEQ ID NO. 274, CDR-H2 as well as to an amino acid sequences at least 85% identical, depicted in SEQID NO: 275 and CDR-H3 as depicted in preferably 90%, more preferred at least 95% identical, most SEQ ID NO. 276: preferred at least 96.97, 98, or 99% identical to the amino (b) CDR-H1 as depicted in SEQ ID NO. 288, CDR-H2 as 15 acid sequence of SEQID NOs: 280,294, 266,308,322,336 depicted in SEQID NO: 289 and CDR-H3 as depicted in or 487. The invention relates also to the corresponding SEQ ID NO. 290; nucleic acid sequences as depicted in any of SEQID NOs: (c) CDR-H1 as depicted in SEQ ID NO. 302, CDR-H2 as 281,295, 267, 309,323,337 or 488, as well as to nucleic acid depicted in SEQID NO:303 and CDR-H3 as depicted in sequences at least 85% identical, preferably 90%, more pre SEQ ID NO. 304; ferred at least 95% identical, most preferred at least 96.97, (d) CDR-H1 as depicted in SEQ ID NO. 316, CDR-H2 as 98, or 99% identical to the nucleic acid sequences shown in depicted in SEQID NO:317 and CDR-H3 as depicted in SEQID NOs: 281,295, 267, 309, 323,337 or 488. Preferred SEQ ID NO. 318; domain arrangements in the PSMAxCD3 bispecific single (e) CDR-H1 as depicted in SEQ ID NO. 330, CDR-H2 as chain antibody constructs of the invention are shown in the depicted in SEQID NO:331 and CDR-H3 as depicted in 25 following examples. SEQ ID NO. 332; In a preferred embodiment of the invention, the bispecific (f) CDR-H1 as depicted in SEQ ID NO. 260, CDR-H2 as single chain antibodies are cross-species specific for CD3 depicted in SEQID NO: 261 and CDR-H3 as depicted in epsilon and for the human and non-chimpanzee primate cell SEQID NO. 262; and surface antigen PSMA, recognized by their second binding (g) CDR-H1 as depicted in SEQ ID NO. 476, CDR-H2 as 30 domain. depicted in SEQID NO: 477 and CDR-H3 as depicted in In an alternative embodiment the invention provides a SEQ ID NO.478. bispecific single chain antibody comprising a first domain In a further preferred embodiment of a bispecific single binding domain capable of binding to CD3 epsilon (CD3e) of chain antibody of the invention the second domain comprises human and non-chimpanzee primate and a second domain a VL region and a VH region selected from the group con 35 binding domain capable of binding to the extracellular sisting of: domain of the mutated human FAPC. chimera having an (a) a VL region as depicted in SEQ ID NO. 268 and a VH amino acid sequence as depicted in SEQID NO: 448 but not region as depicted in SEQID NO. 273: to the extracellular domain of the rodent FAPC. In other (b) a VL region as depicted in SEQ ID NO. 282 and a VH words, the bispecific antibody of the invention specifically region as depicted in SEQID NO. 287: 40 binds to membrane proximal epitopes, i.e. epitopes formed (c) a VL region as depicted in SEQ ID NO. 296 and a VH only by amino acid resides of the extracellular domain of region as depicted in SEQID NO. 301; FAPO, the alpha C-atom of which has a distance of less than (d) a VL region as depicted in SEQ ID NO. 310 and a VH 60 A from the reference C-atom (the alpha C-atom of the 13' region as depicted in SEQID NO. 315; aa as counted from the junction of transmembrane and extra (e) a VL region as depicted in SEQ ID NO. 324 and a VH 45 cellular region). region as depicted in SEQID NO. 329; According to a preferred embodiment of the invention an (f) a VL region as depicted in SEQ ID NO. 254 and a VH above characterized bispecific single chain antibody mol region as depicted in SEQID NO. 259; and ecule comprises a group of the following sequences as CDR (g) a VL region as depicted in SEQ ID NO. 480 and a VH H1, CDR H2, CDR H3, CDRL1, CDRL2 and CDRL3 in the region as depicted in SEQID NO. 475. 50 second binding domain selected from: More preferably, the second domain comprises an amino CDR H1-3 of SEQID NO: 1137-1139 and CDRL 1-3 of acid sequence selected from the group consisting of SEQID SEQID NO: 1132-1134. NOs: 278, 292, 306, 320,334, 485 or 264. The sequences of the corresponding VL- and VH-regions It is preferred for the bispecific single chain antibody com of the second binding domain of the bispecific single chain prising a first domain binding domain capable of binding to 55 antibody molecule of the invention as well as of the respective CD3 epsilon (CD3e) of human and non-chimpanzee primate ScFVs are shown in the sequence listing. and a second domain binding domain capable of binding to According to a preferred embodiment of the invention an the extracellular domain of the mutated human PSMA chi above characterized bispecific single chain antibody mol mera that the first domain capable of binding to an epitope of ecule comprises a group of the following sequences as CDR human and non-chimpanzee primate CD3e chain comprises 60 H1, CDR H2, CDR H3, CDRL1, CDRL2 and CDRL3 in the an amino acid sequence selected from the group consisting of second binding domain selected from the group consisting of SEQID NOs: 23, 25, 41,43, 59,61, 77,79, 95, 97,113, 115, a) CDR H1-3 of SEQID NO: 808–810 and CDRL 1-3 of 131, 133, 149, 151,167, 169, 185 or 187. SEQ ID NO: 813-815; A particularly preferred embodiment of the invention con b) CDR H1-3 of SEQID NO: 794-796 and CDRL1-3 of cerns an above characterized polypeptide, wherein the bispe 65 SEQ ID NO: 799-801; cific single chain antibody molecule comprises a sequence c) CDR H1-3 of SEQID NO: 738-740 and CDRL 1-3 of selected from: SEQ ID NO: 743-745; US 9,260,522 B2 33 34 d) CDR H1-3 of SEQID NO: 752-754 and CDRL 1-3 of H1, CDR H2, CDR H3, CDRL1, CDRL2 and CDRL3 in the SEQID NO: 757-759; second binding domain selected from the group consisting of e) CDR H1-3 of SEQ ID NO: 822-824 and CDRL 1-3 of a) CDR H1-3 of SEQID NO: 500-502 and CDRL 1-3 of SEQID NO: 827-829: SEQ ID NO: 505-507: f) CDR H1-3 of SEQID NO: 766-768 and CDRL 1-3 of 5 b) CDR H1-3 of SEQIDN O: 514-516 and CDRL 1-3 of SEQID NO: 771-773; and SEQ ID NO: 519-521; g) CDR H1-3 of SEQID NO: 780-782 and CDRL 1-3 of c) CDR H1-3 of SEQIDN O : 528-530 and CDRL 1-3 of SEQID NO: 785-787. SEQ ID NO: 533-535: In the bispecific single chain antibody molecule of the d) CDR H1-3 of SEQIDN O : 542-544 and CDRL 1-3 of invention the binding domains are arranged in the order VL- 10 SEQ ID NO:547-549; VH-VH-VL, VL-VH-VL-VH, VH-VL-VH-VL or VH-VL e) CDR H1-3 of SEQID NO. 556-558 and CDRL 1-3 of VL-VH, as exemplified in the appended examples. Prefer SEQ ID NO:561-563: ably, the binding domains are arranged in the order VHFAP f) CDR H1-3 of SEQID N O : 570-572 and CDRL1-3 of alpha-VL FAP alpha-VH CD3-VL CD3 or VL FAP alpha SEQ ID NO:575-577; -VH FAP alpha -VH CD3-VL CD3. More preferred, the 15 g) CDR H1-3 of SEQIDN O : 584-586 and CDRL 1-3 of binding domains are arranged in the order VL FAP alpha-VH SEQ ID NO:589-591; FAP alpha-VH CD3-VLCD3. h) CDR H1-3 of SEQID N O : 598-600 and CDRL 1-3 of It is preferred for the bispecific single chain antibody com SEQ ID NO: 603-605; prising a first domain binding domain capable of binding to i) CDR H1-3 of SEQID NO: 612-614 and CDRL 1-3 of CD3 epsilon (CD3e) of human and non-chimpanzee primate 20 SEQ ID NO: 617-619; and a second domain binding domain capable of binding to j) CDR H1-3 of SEQID NO: 626-628 and CDRL 1-3 of the extracellular domain of the mutated human FAPO. chi SEQ ID NO:631-633; mera that the first domain capable of binding to an epitope of k) CDR H1-3 of SEQID NO: 640-642 and CDRL1-3 of human and non-chimpanzee primate CD3e chain comprises SEQ ID NO: 645-647; an amino acid sequence selected from the group consisting of 25 1) CDR H1-3 of SEQID NO: 654-656 and CDRL 1-3 of SEQID NOs: 23, 25, 41,43, 59,61, 77,79, 95, 97,113, 115, SEQ ID NO: 659-661; 131, 133, 149, 151,167, 169, 185 or 187. m) CDR H1-3 of SEQID NO: 668-670 and CDRL1-3 of A particularly preferred embodiment of the invention con SEQ ID NO: 673-675; cerns an above characterized polypeptide, wherein the bispe n) CDR H1-3 of SEQID NO: 682-684 and CDRL1-3 of cific single chain antibody molecule comprises a sequence 30 SEQ ID NO: 687-689; selected from: o) CDR H1-3 of SEQID NO: 696-698 and CDRL 1-3 of (a) anamino acid sequence as depicted in any of SEQID NOS. SEQ ID NO: 701-703; 819, 805, 749, 763, 833, 777 or 791; p) CDR H1-3 of SEQID NO: 710-712 and CDRL1-3 of (b) an amino acid sequence encoded by a nucleic acid SEQ ID NO: 715–717; and sequence as depicted in any of SEQIDNOs: 820,806, 750, 35 q) CDR H1-3 of SEQID NO: 724-726 and CDRL 1-3 of 764, 834, 778 or 792; and SEQ ID NO: 729-731. (c) an amino acid sequence at least 90% identical, more The sequences of the corresponding VL- and VH-regions preferred at least 95 identical, most preferred at least 96% of the second binding domain of the bispecific single chain identical to the amino acid sequence of (a) or (b). antibody molecule of the invention as well as of the respective The invention relates to a bispecific single chain antibody 40 ScFVs are shown in the sequence listing. molecule comprising an amino acid sequence as depicted in In the bispecific single chain antibody molecule of the any of SEQID NOs: 819, 805, 749, 763, 833, 777 or 791, as invention the binding domains are arranged in the order VL well as to an amino acid sequences at least 85% identical, VH-VH-VL, VL-VH-VL-VH, VH-VL-VH-VL or VH-VL preferably 90%, more preferred at least 95% identical, most VL-VH, as exemplified in the appended examples. Prefer preferred at least 96.97, 98, or 99% identical to the amino 45 ably, the binding domains are arranged in the order VH acid sequence of SEQID NOs: 819, 805, 749, 763, 833, 777 C-MET-VL C-MET-VH CD3-VL, CD3 or VL C-MET-VH or 791. The invention relates also to the corresponding C-MET-VH CD3-VL, CD3. nucleic acid sequences as depicted in any of SEQID NOs: More preferably, the first domain capable of binding to an 820, 806, 750, 764, 834, 778 or 792, as well as to nucleic acid epitope of human and non-chimpanzee primate CD3e chain sequences at least 85% identical, preferably 90%, more pre- 50 comprises an amino acid sequence selected from the group ferred at least 95% identical, most preferred at least 96.97, consisting of SEQID NOs: 23, 25, 41, 43, 59, 61, 77,79, 95, 98, or 99% identical to the nucleic acid sequences shown in 97, 113, 115, 131, 133, 149, 151,167, 169, 185 or 187. SEQID NOs: 820, 806, 750, 764,834, 778 or 792. Preferred A particularly preferred embodiment of the invention con domain arrangements in the FAPOxCD3 bispecific single cerns an above characterized polypeptide, wherein the bispe chain antibody constructs of the invention are shown in the 55 cific single chain antibody molecule comprises a sequence following examples. selected from: In a further alternative embodiment the invention provides (a) an amino acid sequence as depicted in any of SEQID a bispecific single chain antibody comprising a first domain NOs. 511,525,539,553,567 or 581; binding domain capable of binding to CD3 epsilon (CD3e) of (b) an amino acid sequence encoded by a nucleic acid human and non-chimpanzee primate and a second domain 60 sequence as depicted in any of SEQID NOs: 512, 526, binding domain capable of binding to the four Ig domains 540, 554, 568 or 582; and (SEQIDNO:436), a cystein-rich domain (SEQID NO:437), (c) an amino acid sequence at least 90% identical, more or the beta-chain of a sema domain (SEQID NO: 438) of the preferred at least 95% identical, most preferred at least extracellular domain of c-MET. 96% identical to the amino acid sequence of (a) or (b). According to a preferred embodiment of the invention an 65 The invention relates to a bispecific single chain antibody above characterized bispecific single chain antibody mol molecule comprising an amino acid sequence as depicted in ecule comprises a group of the following sequences as CDR any of SEQID NOs: 511, 525, 539,553, 567 or 581, as well US 9,260,522 B2 35 36 as to an amino acid sequences at least 85% identical, prefer (b) an amino acid sequence encoded by a nucleic acid ably 90%, more preferred at least 95% identical, most pre sequence as depicted in any of SEQID NOs: 848, or 862; ferred at least 96, 97,98, or 99% identical to the amino acid and sequence of SEQIDNOs: 511,525,539,553,567 or 581. The (c) an amino acid sequence at least 90% identical, more invention relates also to the corresponding nucleic acid 5 preferred at least 95% identical, most preferred at least sequences as depicted in any of SEQID NOs: 512, 526, 540, 96% identical to the amino acid sequence of (a) or (b). 554, 568 or 582 as well as to nucleic acid sequences at least The invention relates to a bispecific single chain antibody 85% identical, preferably 90%, more preferred at least 95% molecule comprising an amino acid sequence as depicted in identical, most preferred at least 96, 97.98, or 99% identical any of SEQID NOs: 847 or 861, as well as to an amino acid 10 sequences at least 85% identical, preferably 90%, more pre to the nucleic acid sequences shown in SEQ ID NOs: 512, ferred at least 95% identical, most preferred at least 96.97, 526,540, 554, 568 or 582. 98, or 99% identical to the amino acid sequence of SEQ ID Preferred domain arrangements in the c-METxCD3 bispe NOs: 847 or 861. The invention relates also to the correspond cific single chain antibody constructs of the invention are ing nucleic acid sequences as depicted in any of SEQID NOs: shown in the following examples. 15 848, or 862 as well as to nucleic acid sequences at least 85% According to an alternative embodiment the invention pro identical, preferably 90%, more preferred at least 95% iden vides a bispecific single chain antibody comprising a first tical, most preferred at least 96.97, 98, or 99 identical to the domain binding domain capable of binding to CD3 epsilon nucleic acid sequences shown in SEQID NOs: 848, or 862. (CD3e) of human and non-chimpanzee primate and a second Preferred domain arrangements in the IGF-1 RxCD3 bispe domain binding domain capable of binding to the mucin cific single chain antibody constructs of the invention are domain (SEQID NO:440), the three EGF-like domains (SEQ shown in the following examples. ID NO:441), or the Sushi/SCR/CCP domain (SEQID NO: The invention relates to bispecific single chain antibody 442) of the extracellular domain of endosialin (TEM1). molecule comprising the above identified amino acid Preferably the first domain capable of binding to an epitope sequences, as well as to amino acid sequences at least 85% of human and non-chimpanzee primate CD3e chain com 25 identical, preferably 90%, more preferred at least 95% iden prises an amino acid sequence selected from the group con tical, most preferred at least 96.97,98, or 99% identical to the sisting of SEQID NOs: 23, 25, 41,43, 59, 61,77, 79, 95.97, amino acid sequence of said sequences. As described herein 113, 115, 131, 133, 149, 151,167, 169, 185 or 187. above, the specificity of an antibody is generally understood Moreover, in an alternative embodiment the invention pro to be determined by the CDR sequences. Accordingly, in vides a bispecific single chain antibody comprising a first 30 order to maintain the specificity of a given group of CDRs, e.g domain binding domain capable of binding to CD3 epsilon three CDRs of a heavy chain and three CDRs of a light chain, (CD3e) of human and non-chimpanzee primate and a second the sequence of these CDRs has to be conserved. Accordingly domain binding domain capable of binding to the three variants of bispecific single antibodies as identified herein, fibronectin type III domains (SEQ ID NO: 444), the L2 which also fall under the present invention are preferably domain (SEQ ID NO: 445) of the extracellular domain of 35 variants having more than one amino acid Substitutions in FR IGF-1R. According to a preferred embodiment of the inven regions instead of the CDRs. It is to be understood that the tion an above characterized bispecific single chain antibody sequence identity is determined over the entire nucleotide or molecule comprises a group of the following sequences as amino acid sequence. For sequence alignments, for example, CDRH1, CDR H2, CDR H3, CDRL1, CDRL2 and CDRL3 the programs Gap or BestFit can be used (Needleman and in the second binding domain selected from the group con 40 Wunsch J. Mol. Biol. 48 (1970), 443-453; Smith and Water sisting of: man, Adv. Appl. Math 2 (1981), 482-489), which is contained a) CDR H1-3 of SEQ ID NO: 836-838 and CDRL 1-3 of in the GCG software package (Genetics Computer Group, SEQID NO: 841-843; and 575 Science Drive, Madison, Wis., USA53711 (1991). It is a b) CDR H1-3 of SEQID NO: 850-852 and CDRL 1-3 of routine method for those skilled in the art to determine and SEQID NO: 855-857. 45 identify a nucleotide oramino acid sequence having e.g. 85% The sequences of the corresponding VL- and VH-regions (90%. 95%,96%.97%, 98% or 99%) sequence identity to the of the second binding domain of the bispecific single chain nucleotide or amino acid sequences of the bispecific single antibody molecule of the invention as well as of the respective chain antibody of the invention by using e.g. one of the above ScFVs are shown in the sequence listing. mentioned programs. For example, according to Crick's In the bispecific single chain antibody molecule of the 50 Wobble hypothesis, the 5' base on the anti-codon is not as invention the binding domains are arranged in the order VL spatially confined as the other two bases, and could thus have VH-VH-VL, VL-VH-VL-VH, VH-VL-VH-VL or VH-VL non-standard base pairing. Put in other words: the third posi VL-VH, as exemplified in the appended examples. Prefer tion in a codon triplet may vary so that two triplets which ably, the binding domains are arranged in the order VH IGF differ in this third position may encode the same amino acid 1R-VLIGF-1R-VH CD3-VL, CD3 or VLIGF-1R-VH IGF 55 residue. Said hypothesis is well knownto the person skilled in 1R-VH CD3-VL, CD3. the art (see e.g. http://en.wikipedia.org/wiki/Wobble Hy Preferably, the first domain capable of binding to an pothesis: Crick, J Mol Biol 19 (1966): 548-55). epitope of human and non-chimpanzee primate CD3e chain In an alternative embodiment the present invention pro comprises an amino acid sequence selected from the group vides a nucleic acid sequence encoding an above described consisting of SEQID NOS: 23, 25, 41, 43, 59, 61, 77, 79,95, 60 bispecific single chain antibody molecule of the invention. 97, 113, 115, 131, 133, 149, 151,167, 169, 185 or 187. The present invention also relates to a vector comprising A particularly preferred embodiment of the invention con the nucleic acid molecule of the present invention. cerns an above characterized polypeptide, wherein the bispe Many suitable vectors are known to those skilled in cific single chain antibody molecule comprises a sequence molecular biology, the choice of which would depend on the selected from: 65 function desired and include plasmids, cosmids, viruses, bac (a) an amino acid sequence as depicted in any of SEQID teriophages and other vectors used conventionally in genetic NOs: 847 or 861, engineering. Methods which are well known to those skilled US 9,260,522 B2 37 38 in the art can be used to construct various plasmids and The leader sequence(s) is (are) assembled in appropriate vectors; see, for example, the techniques described in Sam phase with translation, initiation and termination sequences, brook et al. (loc cit.) and Ausubel, Current Protocols in and preferably, a leader sequence capable of directing secre Molecular Biology, Green Publishing Associates and Wiley tion of translated protein, or a portion thereof, into the peri Interscience, N.Y. (1989), (1994). Alternatively, the poly 5 plasmic space or extracellular medium. Optionally, the het nucleotides and vectors of the invention can be reconstituted erologous sequence can encode a fusion protein including an into liposomes for delivery to target cells. As discussed in N-terminal identification peptide imparting desired charac further details below, a cloning vector was used to isolate teristics, e.g., Stabilization or simplified purification of individual sequences of DNA. Relevant sequences can be expressed recombinant product; see Supra. In this context, transferred into expression vectors where expression of a 10 Suitable expression vectors are known in the art such as particular polypeptide is required. Typical cloning vectors Okayama-Berg cDNA expression vector pcDV1 (Pharma include pBluescript SK, pGEM, puC9, pBR322 and pGBT9. cia), pCDM8, pRc/CMV, pcDNA1, pcDNA3 (Invitrogene), Typical expression vectors include pTRE, pCAL-n-EK, pEF-DHFR, pEF-ADA or pEF-neo (Mack et al. PNAS pESP-1, pCP13CAT. (1995) 92, 7021-7025 and Raum et al. Cancer Immunol Preferably said vector comprises a nucleic acid sequence 15 Immunother (2001) 50(3), 141-150) or pSPORT1 (GIBCO which is a regulatory sequence operably linked to said nucleic BRL). acid sequence defined herein. Preferably, the expression control sequences will be The term “regulatory sequence” refers to DNA sequences, eukaryotic promoter systems in vectors capable of transform which are necessary to effect the expression of coding ing of transfecting eukaryotic host cells, but control sequences to which they are ligated. The nature of such con sequences for prokaryotic hosts may also be used. Once the trol sequences differs depending upon the host organism. In vector has been incorporated into the appropriate host, the prokaryotes, control sequences generally include promoter, host is maintained under conditions suitable for high level ribosomal binding site, and terminators. In eukaryotes gen expression of the nucleotide sequences, and as desired, the erally control sequences include promoters, terminators and, collection and purification of the bispecific single chain anti in some instances, enhancers, transactivators or transcription 25 body molecule of the invention may follow; see, e.g., the factors. The term “control sequence' is intended to include, at appended examples. a minimum, all components the presence of which are nec An alternative expression system, which can be used to essary for expression, and may also include additional advan express a cell cycle interacting protein is an insect system. In tageous components. one Such system, Autographa Californica nuclear polyhedro The term “operably linked’ refers to a juxtaposition 30 sis virus (AcNPV) is used as a vector to express foreign genes wherein the components so described are in a relationship in Spodoptera frugiperda cells or in Trichoplusia larvae. The permitting them to function in their intended manner. A con coding sequence of a recited nucleic acid molecule may be trol sequence “operably linked to a coding sequence is cloned into a nonessential region of the virus, such as the ligated in Such a way that expression of the coding sequence polyhedrin gene, and placed under control of the polyhedrin is achieved under conditions compatible with the control 35 promoter. Successful insertion of said coding sequence will sequences. In case the control sequence is a promoter, it is render the polyhedrin gene inactive and produce recombinant obvious for a skilled person that double-stranded nucleic acid virus lacking coat protein coat. The recombinant viruses are is preferably used. then used to infect S. frugiperda cells or Trichoplusia larvae in Thus, the recited vector is preferably an expression vector. which the protein of the invention is expressed (Smith, J. An “expression vector is a construct that can be used to 40 Virol. 46 (1983), 584: Engelhard, Proc. Nat. Acad. Sci. USA transform a selected host and provides for expression of a 91 (1994), 3224-3227). coding sequence in the selected host. Expression vectors can Additional regulatory elements may include transcrip for instance be cloning vectors, binary vectors or integrating tional as well as translational enhancers. Advantageously, the vectors. Expression comprises transcription of the nucleic above-described vectors of the invention comprise a select acid molecule preferably into a translatable mRNA. Regula 45 able and/or scorable marker. tory elements ensuring expression in prokaryotes and/or Selectable marker genes useful for the selection of trans eukaryotic cells are well known to those skilled in the art. In formed cells and, e.g., plant tissue and plants are well known the case of eukaryotic cells they comprise normally promot to those skilled in the art and comprise, for example, antime ers ensuring initiation of transcription and optionally poly-A tabolite resistance as the basis of selection for dhfr, which signals ensuring termination of transcription and stabilization 50 confers resistance to methotrexate (Reiss, Plant Physiol. (Life of the transcript. Possible regulatory elements permitting Sci. Adv.) 13 (1994), 143-149): npt, which confers resistance expression in prokaryotic host cells comprise, e.g., the Plac, to the aminoglycosides neomycin, kanamycin and paromycin trp or tac promoter in E. coli, and examples of regulatory (Herrera-Estrella, EMBO J. 2 (1983), 987-995) and hygro, elements permitting expression in eukaryotic host cells are which confers resistance to hygromycin (Marsh, Gene 32 the AOX1 or GAL1 promoter in yeast or the CMV-, SV40-, 55 (1984), 481-485). Additional selectable genes have been RSV-promoter (Rous sarcoma virus), CMV-enhancer, SV40 described, namely trpB, which allows cells to utilize indole in enhancer or a globin intron in mammalian and other animal place of tryptophan; hisD, which allows cells to utilize histi cells. nol in place of histidine (Hartman, Proc. Natl. Acad. Sci. USA Beside elements, which are responsible for the initiation of 85 (1988), 8047); mannose-6-phosphate isomerase which transcription Such regulatory elements may also comprise 60 allows cells to utilize mannose (WO 94/20627) and ODC transcription termination signals, such as the SV40-poly-A (ornithine decarboxylase) which confers resistance to the site or the tk-poly-A site, downstream of the polynucleotide. ornithine decarboxylase inhibitor, 2-(difluoromethyl)-DL Furthermore, depending on the expression system used leader ornithine, DFMO (McConlogue, 1987. In: Current Commu sequences capable of directing the polypeptide to a cellular nications in Molecular Biology, Cold Spring Harbor Labora compartment or secreting it into the medium may be added to 65 tory ed.) or deaminase from Aspergillus terreus which the coding sequence of the recited nucleic acid sequence and confers resistance to Blasticidin S (Tamura, Biosci. Biotech are well known in the art; see also the appended Examples. nol. Biochem. 59 (1995), 2336-2338). US 9,260,522 B2 39 40 Useful scorable markers are also known to those skilled in sequence of the bispecific single chain antibody molecule of the art and are commercially available. Advantageously, said the invention and genetically fused thereto an N-terminal marker is a gene encoding luciferase (Giacomin, Pl. Sci. 116 FLAG-tag and/or C-terminal His-tag. Preferably, the length (1996), 59-72: Scikantha, J. Bact. 178 (1996), 121), green of said FLAG-tag is about 4 to 8 amino acids, most preferably fluorescent protein (Gerdes, FEBS Lett. 389 (1996), 44-47) or 8 amino acids. An above described polynucleotide can be |B-glucuronidase (Jefferson, EMBO.J. 6 (1987), 3901-3907). used to transform or transfect the host using any of the tech This embodiment is particularly useful for simple and rapid niques commonly known to those of ordinary skill in the art. screening of cells, tissues and organisms containing a recited Furthermore, methods for preparing fused, operably linked Vector. genes and expressing them in, e.g., mammalian cells and As described above, the recited nucleic acid molecule can 10 be used alone or as part of a vector to express the bispecific bacteria are well-known in the art (Sambrook, loc cit.). Pref single chain antibody molecule of the invention in cells, for, erably, said the host is a bacterium or an insect, fungal, plant e.g., purification but also for gene therapy purposes. The or animal cell. nucleic acid molecules or vectors containing the DNA It is particularly envisaged that the recited host may be a sequence(s) encoding any one of the above described bispe 15 mammalian cell. Particularly preferred host cells comprise cific single chain antibody molecule of the invention is intro CHO cells, COS cells, myeloma cell lines like SP2/0 or NS/0. duced into the cells which in turn produce the polypeptide of As illustrated in the appended examples, particularly pre interest. Gene therapy, which is based on introducing thera ferred are CHO-cells as hosts. peutic genes into cells by ex-Vivo or in-vivo techniques is one More preferably said host cell is a human cellor humancell of the most important applications of gene transfer. Suitable line, e.g. percó (Kroos, Biotechnol. Prog., 2003, 19:163– vectors, methods or gene-delivery systems for in-vitro or 168). in-vivo gene therapy are described in the literature and are In a further embodiment, the present invention thus relates known to the person skilled in the art; see, e.g., Giordano, to a process for the production of a bispecific single chain Nature Medicine 2 (1996), 534-539; Schaper, Circ. Res. 79 antibody molecule of the invention, said process comprising (1996), 911-919; Anderson, Science 256 (1992), 808–813; 25 culturing a host of the invention under conditions allowing the Verma, Nature 389 (1994), 239; Isner, Lancet 348 (1996), expression of the bispecific single chain antibody molecule of 370-374; Muhlhauser, Circ. Res. 77 (1995), 1077-1086: the invention and recovering the produced polypeptide from Onodera, Blood 91 (1998), 30-36; Verma, Gene Ther. 5 the culture. (1998), 692-699; Nabel, Ann. N.Y. Acad. Sci. 811 (1997), The transformed hosts can be grown in fermentors and 289-292; Verzeletti, Hum. Gene Ther. 9 (1998), 2243-51: 30 cultured according to techniques known in the art to achieve Wang, Nature Medicine 2 (1996), 714-716; WO94/29469; optimal cell growth. The bispecific single chain antibody WO 97/00957, U.S. Pat. No. 5,580,859; U.S. Pat. No. 5,589, molecule of the invention can then be isolated from the 466; or Schaper, Current Opinion in Biotechnology 7 (1996), growth medium, cellular lysates, or cellular membrane frac 635-640; dos Santos Coura and Nardi Virol J. (2007), 4:99. tions. The isolation and purification of the, e.g., microbially The recited nucleic acid molecules and vectors may be 35 expressed bispecific single chain antibody molecules may be designed for direct introduction or for introduction via lipo by any conventional means such as, for example, preparative Somes, or viral vectors (e.g., adenoviral, retroviral) into the chromatographic separations and immunological separations cell. Preferably, said cell is a germline cell, embryonic cell, or Such as those involving the use of monoclonal or polyclonal egg cell or derived there from, most preferably said cell is a antibodies directed, e.g., against a tag of the bispecific single stem cell. An example for an embryonic stem cell can be, inter 40 chain antibody molecule of the invention or as described in alia, a stem cell as described in Nagy, Proc. Natl. Acad. Sci. the appended examples. USA 90 (1993), 8424-8428. The conditions for the culturing of a host, which allow the The invention also provides for a host transformed or trans expression are known in the art to depend on the host system fected with a vector of the invention. Said host may be pro and the expression system/vector used in Such process. The duced by introducing the above described vector of the inven 45 parameters to be modified in order to achieve conditions tion or the above described nucleic acid molecule of the allowing the expression of a recombinant polypeptide are invention into the host. The presence of at least one vector or known in the art. Thus, suitable conditions can be determined at least one nucleic acid molecule in the host may mediate the by the person skilled in the art in the absence of further expression of a gene encoding the above described single inventive input. chain antibody constructs. 50 Once expressed, the bispecific single chain antibody mol The described nucleic acid molecule or vector of the inven ecule of the invention can be purified according to standard tion, which is introduced in the host may either integrate into procedures of the art, including ammonium Sulfate precipita the genome of the host or it may be maintained extrachromo tion, affinity columns, column chromatography, gel electro Somally. phoresis and the like; see, Scopes, “Protein Purification', The host can be any prokaryote or eukaryotic cell. 55 Springer-Verlag, N.Y. (1982). Substantially pure polypep The term “prokaryote' is meant to include all bacteria, tides of at least about 90 to 95% homogeneity are preferred, which can be transformed or transfected with DNA or RNA and 98 to 99% or more homogeneity are most preferred, for molecules for the expression of a protein of the invention. pharmaceutical uses. Once purified, partially or to homoge Prokaryotic hosts may include gram negative as well as gram neity as desired, the bispecific single chainantibody molecule positive bacteria Such as, for example, E. coli, S. typhimu 60 of the invention may then be used therapeutically (including rium, Serratia marcescens and Bacillus subtilis. The term extracorporeally) or in developing and performing assay pro “eukaryotic' is meant to include yeast, higher plant, insect cedures. Furthermore, examples for methods for the recovery and preferably mammalian cells. Depending upon the host of the bispecific single chain antibody molecule of the inven employed in a recombinant production procedure, the protein tion from a culture are described in detail in the appended encoded by the polynucleotide of the present invention may 65 examples. be glycosylated or may be non-glycosylated. Especially pre Furthermore, the invention provides for a composition ferred is the use of a plasmid or a virus containing the coding comprising a bispecific single chain antibody molecule of the US 9,260,522 B2 41 42 invention or a bispecific single chain antibody as produced by vessel, thereby allowing a direct contact between the pump the process disclosed above. Preferably, said composition is a system and the skin of the patient. The pump system can be pharmaceutical composition. attached to the skin of the patient for 24 hours up to several The invention provides also for a bispecific single chain days. The pump system may be of small size with a reservoir antibody molecule as defined herein, or produced according for Small Volumes. As a non-limiting example, the Volume of to the process as defined herein, wherein said bispecific single the reservoir for the suitable pharmaceutical composition to chain antibody molecule is for use in the prevention, treat be administered can be between 0.1 and 50 ml. ment or amelioration of cancer. Preferably, said cancer is a The continuous administration may be transdermal by way Solid tumor, more preferably a carcinoma or prostate cancer. of a patch worn on the skin and replaced at intervals. One of It is preferred that the bispecific single chain is further com 10 skill in the art is aware of patch systems for drug delivery prising Suitable formulations of carriers, stabilizers and/or suitable for this purpose. It is of note that transdermal admin excipients. Moreover, it is preferred that said bispecific single istration is especially amenable to uninterrupted administra chain antibody molecule is suitable to be administered in tion, as exchange of a first exhausted patch can advanta combination with an additional drug. Said drug may be a geously be accomplished simultaneously with the placement non-proteinaceous compound or a proteinaceous compound 15 of a new, second patch, for example on the Surface of the skin and may be administered simultaneously or non-simulta immediately adjacent to the first exhausted patch and imme neously with the bispecific single chain antibody molecule as diately prior to removal of the first exhausted patch. Issues of defined herein. flow interruption or power cell failure do not arise. In accordance with the invention, the term “pharmaceutical The composition of the present invention, comprising in composition” relates to a composition for administration to a particular bispecific single chain antibodies preferably patient, preferably a human patient. The particular preferred directed against and generated against context-independent pharmaceutical composition of this invention comprises CD3 epitopes may further comprise a pharmaceutically bispecific single chain antibodies directed against and gener acceptable carrier. Examples of Suitable pharmaceutical car ated against context-independent CD3 epitopes. Preferably, riers are well known in the art and include solutions, e.g. the pharmaceutical composition comprises suitable formula 25 phosphate buffered saline Solutions, water, emulsions. Such tions of carriers, stabilizers and/or excipients. In a preferred as oil/water emulsions, various types of wetting agents, sterile embodiment, the pharmaceutical composition comprises a Solutions, liposomes, etc. Compositions comprising such car composition for parenteral, transdermal, intraluminal, riers can beformulated by well known conventional methods. intraarterial, intrathecal and/or intranasal administration or Formulations can comprise carbohydrates, buffer Solutions, by direct injection into tissue. It is in particular envisaged that 30 amino acids and/or surfactants. Carbohydrates may be non said composition is administered to a patient via infusion or reducing Sugars, preferably trehalose. Sucrose, octasulfate, injection. Administration of the suitable compositions may be sorbitol or xylitol. Such formulations may be used for con effected by different ways, e.g., by intravenous, intraperito tinuous administrations which may be intravenuous or Sub neal, Subcutaneous, intramuscular, topical or intradermal cutaneous with and/or without pump systems. Amino acids administration. In particular, the present invention provides 35 may be charged amino acids, preferably lysine, lysine acetate, for an uninterrupted administration of the Suitable composi arginine, glutamate and/or histidine. Surfactants may be tion. As a non-limiting example, uninterrupted, i.e. continu detergents, preferably with a molecular weight of >1.2 KD ous administration may be realized by a small pump system and/or a polyether, preferably with a molecular weight of >3 worn by the patient for metering the influx of therapeutic KD. Non-limiting examples for preferred detergents are agent into the body of the patient. The pharmaceutical com 40 Tween 20, Tween 40, Tween 60, Tween 80 or Tween 85. position comprising the bispecific single chain antibodies Non-limiting examples for preferred polyethers are PEG directed against and generated against context-independent 3000, PEG 3350, PEG 4000 or PEG 5000. Buffer systems CD3 epitopes of the invention can be administered by using used in the present invention can have a preferred pH of 5-9 said pump systems. Such pump systems are generally known and may comprise citrate. Succinate, phosphate, histidine and in the art, and commonly rely on periodic exchange of car 45 acetate. The compositions of the present invention can be tridges containing the therapeutic agent to be infused. When administered to the subject at a suitable dose which can be exchanging the cartridge in Such a pump system, a temporary determined e.g. by dose escalating studies by administration interruption of the otherwise uninterrupted flow of therapeu of increasing doses of the bispecific single chain antibody tic agent into the body of the patient may ensue. In Sucha case, molecule of the invention exhibiting cross-species specificity the phase of administration prior to cartridge replacement and 50 described herein to non-chimpanzee primates, for instance the phase of administration following cartridge replacement macaques. As set forth above, the bispecific single chain would still be considered within the meaning of the pharma antibody molecule of the invention exhibiting cross-species ceutical means and methods of the invention together make specificity described herein can be advantageously used in up one “uninterrupted administration of such therapeutic identical form in preclinical testing in non-chimpanzee pri agent. 55 mates and as drug in humans. These compositions can also be The continuous or uninterrupted administration of these administered in combination with other proteinaceous and bispecific single chain antibodies directed against and gener non-proteinaceous drugs. These drugs may be administered ated against context-independent CD3 epitopes of this inven simultaneously with the composition comprising the bispe tion may be intravenuous or Subcutaneous by way of a fluid cific single chain antibody molecule of the invention as delivery device or Small pump system including a fluid driv 60 defined herein or separately before or after administration of ing mechanism for driving fluid out of a reservoir and an said polypeptide in timely defined intervals and doses. The actuating mechanism for actuating the driving mechanism. dosage regimen will be determined by the attending physi Pump systems for Subcutaneous administration may include a cian and clinical factors. AS is well known in the medical arts, needle or a cannula for penetrating the skin of a patient and dosages for any one patient depend upon many factors, delivering the suitable composition into the patient’s body. 65 including the patient's size, body Surface area, age, the par Said pump systems may be directly fixed or attached to the ticular compound to be administered, sex, time and route of skin of the patient independently of a vein, artery or blood administration, general health, and other drugs being admin US 9,260,522 B2 43 44 istered concurrently. Preparations for parenteral administra lished. Pharmacokinetic parameters of the drug influencing tion include sterile aqueous or non-aqueous solutions, Sus the ability of a drug for treating a certain disease entity pensions, and emulsions. Examples of non-aqueous solvents include, but are not limited to: half-life, Volume of distribu are propylene glycol, polyethylene glycol, vegetable oils such tion, hepatic first-pass metabolism and the degree of blood as olive oil, and injectable organic esters such as ethyl oleate. serum binding. The efficacy of a given drug agent can be Aqueous carriers include water, alcoholic/aqueous Solutions, influenced by each of the parameters mentioned above. emulsions or Suspensions, including saline and buffered “Half-life” means the time where 50% of an administered media. Parenteral vehicles include sodium chloride solution, drug are eliminated through biological processes, e.g. Ringer's dextrose, dextrose and sodium chloride, lactated metabolism, excretion, etc. Ringer's, or fixed oils. Intravenous vehicles include fluid and 10 By “hepatic first-pass metabolism' is meant the propensity nutrient replenishers, electrolyte replenishers (such as those of a drug to be metabolized upon first contact with the liver, based on Ringer's dextrose), and the like. Preservatives and i.e. during its first pass through the liver. “Volume of distri other additives may also be present such as, for example, bution” means the degree of retention of a drug throughout antimicrobials, anti-oxidants, chelating agents, inert gases the various compartments of the body, like e.g. intracellular and the like. In addition, the composition of the present inven 15 and extracellular spaces, tissues and organs, etc. and the dis tion might comprise proteinaceous carriers, like, e.g., serum tribution of the drug within these compartments. “Degree of albuminor immunoglobulin, preferably of human origin. It is blood serum binding means the propensity of a drug to envisaged that the composition of the invention might com interact with and bind to blood serum proteins, such as albu prise, in addition to the bispecific single chain antibody mol min, leading to a reduction or loss of biological activity of the ecule of the invention defined herein, further biologically drug. active agents, depending on the intended use of the compo Pharmacokinetic parameters also include bioavailability, sition. Such agents might be drugs acting on the gastro lag time (Tlag), Tmax, absorption rates, more onset and/or intestinal system, drugs acting as cytostatica, drugs prevent Cmaxfor a given amount of drug administered. ing hyperurikemia, drugs inhibiting immunoreactions (e.g. “Bioavailability” means the amount of a drug in the blood corticosteroids), drugs modulating the inflammatory 25 compartment. response, drugs acting on the circulatory system and/or “Lag time” means the time delay between the administra agents such as cytokines known in the art. tion of the drug and its detection and measurability in blood or The biological activity of the pharmaceutical composition plasma. defined herein can be determined for instance by cytotoxicity “Tmax’ is the time after which maximal blood concentra assays, as described in the following examples, in WO 30 tion of the drug is reached, and “Cmax’ is the blood concen 99/54440 or by Schlereth et al. (Cancer Immunol. Immu tration maximally obtained with a given drug. The time to nother. 20 (2005), 1-12). “Efficacy” or “in vivo efficacy” as reach a blood or tissue concentration of the drug which is used herein refers to the response to therapy by the pharma required for its biological effect is influenced by all param ceutical composition of the invention, using e.g. standardized eters. Pharmacokinetik parameters of bispecific single chain NC1 response criteria. The success or in vivo efficacy of the 35 antibodies exhibiting cross-species specificity, which may be therapy using a pharmaceutical composition of the invention determined in preclinical animal testing in non-chimpanzee refers to the effectiveness of the composition for its intended primates as outlined above are also set forth e.g. in the pub purpose, i.e. the ability of the composition to cause its desired lication by Schlereth et al. (Cancer Immunol. Immunother. 20 effect, i.e. depletion of pathologic cells, e.g. tumor cells. The (2005), 1-12). in vivo efficacy may be monitored by established standard 40 The term “toxicity' as used herein refers to the toxic effects methods for the respective disease entities including, but not of a drug manifested in adverse events or severe adverse limited to white blood cell counts, differentials, Fluorescence events. These side events might refer to a lack of tolerability Activated Cell Sorting, bone marrow aspiration. In addition, of the drug in general and/or a lack of local tolerance after various disease specific clinical chemistry parameters and administration. Toxicity could also include teratogenic or other established standard methods may be used. Further 45 carcinogenic effects caused by the drug. more, computer-aided tomography, X-ray, nuclear magnetic The term “safety”, “in vivo safety” or “tolerability” as used resonance tomography (e.g. for National Cancer Institute herein defines the administration of a drug without inducing criteria based response assessment Cheson BD, Horning SJ, severe adverse events directly after administration (local tol Coiffier B, Shipp MA, Fisher R I, Connors J M, Lister TA, erance) and during a longer period of application of the drug. Vose J. Grillo-Lopez A. Hagenbeek A, Cabanillas F. Klippen 50 “Safety”, “in vivo safety” or “tolerability” can be evaluated sten D. Hiddemann W. Castellino R. Harris NL, Armitage J e.g. at regular intervals during the treatment and follow-up O. Carter W. Hoppe R, Canellos G. P. Report of an interna period. Measurements include clinical evaluation, e.g. organ tional workshop to standardize response criteria for non manifestations, and Screening of laboratory abnormalities. Hodgkin’s lymphomas. NC Sponsored International Work Clinical evaluation may be carried out and deviating to nor ing Group. J Clin Oncol. 1999 April; 17(4): 1244), positron 55 mal findings recorded/coded according to NC1-CTC and/or emission tomography scanning, white blood cell counts, MedDRA standards. Organ manifestations may include cri differentials, Fluorescence Activated Cell Sorting, bone mar teria Such as allergy/immunology, blood/bone marrow, car row aspiration, lymph node biopsies/histologies, and various diac arrhythmia, coagulation and the like, as set forth e.g. in cancer specific clinical chemistry parameters (e.g. lactate the Common Terminology Criteria for adverse events v3.0 dehydrogenase) and other established standard methods may 60 (CTCAE). Laboratory parameters which may be tested be used. include for instance haematology, clinical chemistry, coagu Another major challenge in the development of drugs such lation profile and urine analysis and examination of other as the pharmaceutical composition of the invention is the body fluids such as serum, plasma, lymphoid or spinal fluid, predictable modulation of pharmacokinetic properties. To liquor and the like. Safety can thus be assessed e.g. by physi this end, a pharmacokinetic profile of the drug candidate, i.e. 65 cal examination, imaging techniques (i.e. ultrasound, X-ray, a profile of the pharmacokinetic parameters that effect the CT scans, Magnetic Resonance Imaging (MRI), other mea ability of a particular drug to treat a given condition, is estab Sures with technical devices (i.e. electrocardiogram), vital US 9,260,522 B2 45 46 signs, by measuring laboratory parameters and recording Preferably, said proteinaceous compound or non-proteina adverse events. For example, adverse events in non-chimpan ceous compound may be administered simultaneously or Zee primates in the uses and methods according to the inven non-simultaneously with the bispecific single chain antibody tion may be examined by histopathological and/or his molecule of the invention, a nucleic acid molecule as defined tochemical methods. hereinabove, a vector as defined as defined hereinabove, or a The term “effective and non-toxic dose” as used herein host as defined as defined hereinabove. refers to a tolerable dose of the bispecific single chain anti Another aspect of the invention relates to a method for the body as defined herein which is high enough to cause deple prevention, treatment or amelioration of a disease in a subject tion of pathologic cells, tumor elimination, tumor shrinkage in the need thereof, said method comprising the step of or stabilization of disease without or essentially without 10 administration of an effective amount of a pharmaceutical major toxic effects. Such effective and non-toxic doses may composition of the invention. Preferably, said disease is can be determined e.g. by dose escalation studies described in the cer or an autoimmune disease. Preferably, said cancer is art and should be below the dose inducing severe adverse side (a) a solid tumor, more preferably a carcinoma or prostate events (dose limiting toxicity, DLT). Cancer, The above terms are also referred to e.g. in the Preclinical 15 (b) a carcinoma, sarcoma, glioblastoma/astrocytoma, mela safety evaluation of biotechnology-derived pharmaceuticals noma, mesothelioma, Wilms tumor or a hematopoietic S6; ICH Harmonised Tripartite Guideline; ICH Steering malignancy Such as leukemia, lymphoma or multiple Committee meeting on Jul. 16, 1997. myeloma; Moreover, the invention relates to a pharmaceutical com (c) carcinomas (breast, kidney, lung, colorectal, colon, pan position comprising a bispecific single chain antibody mol creas mesothelioma), sarcomas, and neuroectodermal ecule of this invention or produced according to the process tumors (melanoma, glioma, neuroblastoma); according to the invention for the prevention, treatment or (d) epithelial cancer, or amelioration of cancer oran autoimmune disease. Preferably, (e) bone or soft tissue cancer (e.g. Ewing sarcoma), breast, said cancer is a: liver, lung, head and neck, colorectal, prostate, leiomyosa (a) a solid tumor, more preferably a carcinoma or prostate 25 rcoma, cervical and endometrial cancer, ovarian, prostate, Cancer, and pancreatic cancer. (b) a carcinoma, sarcoma, glioblastoma/astrocytoma, mela In another preferred embodiment of the method of the noma, mesothelioma, Wilms tumor or a hematopoietic invention said pharmaceutical composition is Suitable to be malignancy Such as leukemia, lymphoma or multiple administered in combination with an additional drug, i.e. as myeloma; 30 part of a co-therapy. In said co-therapy, an active agent may be (c) carcinomas (breast, kidney, lung, colorectal, colon, pan optionally included in the same pharmaceutical composition creas mesothelioma), sarcomas, and neuroectodermal as the bispecific single chain antibody molecule of the inven tumors (melanoma, glioma, neuroblastoma); tion, or may be included in a separate pharmaceutical com (d) epithelial cancer, or position. In this latter case, said separate pharmaceutical (e) bone or soft tissue cancer (e.g. Ewing sarcoma), breast, 35 composition is suitable for administration prior to, simulta liver, lung, head and neck, colorectal, prostate, leiomyosa neously as or following administration of said pharmaceuti rcoma, cervical and endometrial cancer, ovarian, prostate, cal composition comprising the bispecific single chain anti and pancreatic cancer. body molecule of the invention. The additional drug or Preferably, said pharmaceutical composition further com pharmaceutical composition may be a non-proteinaceous prises suitable formulations of carriers, stabilizers and/or 40 compound or a proteinaceous compound. In the case that the excipients. additional drug is a proteinaceous compound, it is advanta A further aspect of the invention relates to a use of a geous that the proteinaceous compound be capable of provid bispecific single chain antibody molecule/polypeptide as ing an activation signal for immune effector cells. defined herein above or produced according to a process Preferably, said proteinaceous compound or non-proteina defined herein above, for the preparation of a pharmaceutical 45 ceous compound may be administered simultaneously or composition for the prevention, treatment or amelioration of non-simultaneously with the bispecific single chain antibody a disease. Preferably, said disease is cancer. More preferably, molecule of the invention, a nucleic acid molecule as defined said cancer is a solid tumor, preferably a carcinoma or pros hereinabove, a vector as defined as defined hereinabove, or a tate Cancer. host as defined as defined hereinabove. In another preferred embodiment of use of the bispecific 50 It is preferred for the above described method of the inven single chain antibody molecule of the invention said pharma tion that said Subject is a human. ceutical composition is Suitable to be administered in combi In a further aspect, the invention relates to a kit comprising nation with an additional drug, i.e. as part of a co-therapy. In a bispecific single chain antibody molecule of the invention, said co-therapy, an active agent may be optionally included in a nucleic acid molecule of the invention, a vector of the the same pharmaceutical composition as the bispecific single 55 invention, or a host of the invention. chain antibody molecule of the invention, or may be included These and other embodiments are disclosed and encom in a separate pharmaceutical composition. In this latter case, passed by the description and Examples of the present inven said separate pharmaceutical composition is Suitable for tion. Recombinant techniques and methods in immunology administration prior to, simultaneously as or following are described e.g. in Sambrook et al. Molecular Cloning: A administration of said pharmaceutical composition compris 60 Laboratory Manual; Cold Spring Harbor Laboratory Press, ing the bispecific single chainantibody molecule of the inven 3" edition 2001: Lefkovits; Immunology Methods Manual: tion. The additional drug orpharmaceutical composition may The Comprehensive Sourcebook of Techniques; Academic be a non-proteinaceous compound or a proteinaceous com Press, 1997: Golemis; Protein-Protein Interactions: A pound. In the case that the additional drug is a proteinaceous Molecular Cloning Manual: Cold Spring Laboratory Press, compound, it is advantageous that the proteinaceous com 65 2002. Further literature concerning any one of the antibodies, pound be capable of providing an activation signal for methods, uses and compounds to be employed in accordance immune effector cells. with the present invention may be retrieved from public US 9,260,522 B2 47 48 libraries and databases, using for example electronic devices. PSMA, rat PSMA mutated to the homologous human amino For example, the public database “Medline', available on the acid at every mismatched amino acid position with a mem Internet, may be utilized. Further databases and addresses are brane-distance of >60 A, and unmutated rat PSMA. The bold known to the person skilled in the art. lines show staining by cell culture supernatant of CHO cells The figures show: transfected with PSMA-directed bispecific antibody con FIG. 1 structs. Cell culture supernatant of untransfected CHO cells Flow cytometry of CHO cells transfected with native served as negative control (thin lines). The PSMA-directed human EpCAM (positive control) and different EpCAM bscAb P6xI2C binds to (unmutated) human PSMA but nei hNG2 fusion proteins, respectively. Staining with the murine ther to unmutated nor to mutated rat PSMA and thus confirms parental IgG1 antibody MAb 5-10 (bold lines) directed 10 against human EpCAM was performed as described (Brisch a membrane-distance of <60 A for the PSMA-epitope of this wein (2007) J Immunother 30: 798-807). PBS/2% FCS bispecific construct. By contrast, PSMA-directed bscAb instead of MAb 5-10 was used as negative control (thin lines). D3xI2C does not bind to unmutated rat PSMA but to (unmu FIG 2 tated) human and mutated rat PSMA consistent with a T cell cytotoxicity redirected by bscAb 5-10xI2C against 15 PSMA-epitope of a membrane-distance >60 A. CHO cells transfected with human EpCAM (positive control) FIG. 7 and different EpCAM-hNG2 fusion proteins as measured in a T cell cytotoxicity redirected by I2C-based PSMA-di chromium 51 (Cr) release assay (y-axis). Stimulated human rected bscAbs to CHO cells transfected with human PSMA as CD4/CD56 depleted PBMC served as effector T cells. Effec measured in a chromium 51 (Cr) release assay. As source of tor- to target cell ratio was 10:1. BscAb 5-10xI2C was used as effector T cells stimulated human CD4/CD56 depleted culture supernatant at different dilutions as indicated on the PBMC were used. The effector-to-target cell ratio was 10:1. X-axis. Assay duration was 18 hours. PSMA-directed bscAbs were used as cell culture superna FIG.3 tants from transfected CHO cells at different dilutions as Amino acid sequence alignment of full-length mature indicated. The assay duration was 18 hours. PSMA-directed human (SEQID NO: 202) and rat (SEQID NO: 206, residues 25 bscAb P6xI2C, whose PSMA-epitope has a membrane-dis 1-732) PSMA. Mismatching homologous amino acid posi tance of <60 A is substantially more potent in redirecting T tions are underlined and highlighted by bold character style. cell cytotoxicity than PSMA-directed bscAb D3xI2C, whose Numbering of amino acid positions refers to human PSMA PSMA-epitope has a membrane-distance of 60 A. (SEQID NO: 202) and starts with the amino acid methionine FIG.8 encoded by the start codon of human PSMA. Intracellular 30 FACS binding analysis of I2C-based bscAbs directed to domain aa 1-aa 19; transmembrane domain aa 20-aa 43; extracellular domain aa 44-aa 750. membrane-proximal PSMA-epitopes on CHO cells trans FIG. 4 fected with macaque PSMA, the CD3 positive human T cell FACS binding analysis of I2C-based anti-human PSMA leukemia cell line HPB-ALL and the CD3 positive macaque bscAbs on CHO cells transfected with (unmutated) human 35 T cell line 4119LnPx. The bold lines show staining by cell PSMA, rat PSMA mutated to the homologous human amino culture supernatant of CHO cells transfected with designated acid at every mismatched amino acid position with a mem PSMA-directed bispecific antibody constructs. Cell culture brane-distance of >60 A, and unmutated rat PSMA. The bold supernatant of untransfected CHO cells served as negative lines show staining by cell culture supernatant of CHO cells control (thin lines). The designated PSMA-directed bscAbs transfected with PSMA-directed bispecific antibody con 40 in addition to human PSMA also bind to macaque PSMA, structs. Cell culture supernatant of untransfected CHO cells human CD3 and macaque CD3. served as negative control (thin lines). PSMA-directed FIG.9 bscAbs P1XI2C, P2XI2C, P3XI2C, P4XI2C and P5XI2C bind Amino acid sequence alignment of full-length mature to (unmutated) human PSMA but neither to unmutated nor to human (SEQ ID NO: 367) and murine (SEQ ID NO:371, mutated rat PSMA and thus confirms a membrane-distance of 45 residues 1-761) FAPO. Mismatching homologous amino acid <60 A for the PSMA-epitope of each of these bispecific positions are underlined and highlighted by bold character constructs. By contrast, PSMA-directed bscAbs DlxI2C and style. Numbering of amino acid positions refers to human D2xI2C do not bind to unmutated rat PSMA but to (unmu FAPalpha (SEQID NO: 367) and starts with the amino acid tated) human and mutated rat PSMA consistent with PSMA methionine encoded by the start codon of human FAPalpha. epitopes of a membrane-distance >60 A. 50 Intracellular domain aa 1-aa 9; transmembrane domain aa FIG.S 10-aa 26; extracellular domain aa 27-aa 760. T cell cytotoxicity redirected by I2C-based PSMA-di FIG 10 rected bscAbs to CHO cells transfected with human PSMA as FACS binding analysis of cell surface expression on CHO measured in a chromium 51 (Cr) release assay. As source of cells expressing the murine FAPalpha antigen as described in effector T cells stimulated human CD4/CD56 depleted 55 Example 6.3 and CHO cells expressing the mutated human PBMC were used. The effector-to-target cell ratio was 10:1. FAPalpha antigen with murine membrane-distal epitopes as PSMA-directed bscAbs were used as cell culture superna described in Example 6.4, respectively. The FACS staining tants from transfected CHO cells at different dilutions as was performed as described in Examples 6.3 and 6.4. The indicated. The assay duration was 18 hours. PSMA-directed bold lines represent cells incubated with the detection anti bscAbs P1XI2C, P2XI2C, P3XI2C, P4XI2C and P5XI2C, 60 bodies—the Penta His antibody in case of murine FAPalpha whose PSMA-epitopes have a membrane-distance of <60 A and the anti-FLAG M2 antibody in case of the mutated human are Substantially more potent in redirecting T cell cytotoxicity FAPalpha antigen with murine membrane-distal epitopes. than PSMA-directed bscAbs D1XI2C and D2xI2C, whose The thin lines represent the negative controls. For both cell PSMA-epitopes have a membrane-distance of 60 A. lines the overlay of the histograms for the anti-FLAG M2 FIG. 6 65 antibody and the Penta His antibody, respectively, shows a FACS binding analysis of I2C-based anti-human PSMA significant expression level of the respective antigen. Expres bscAbs on CHO cells transfected with (unmutated) human sion levels were comparable for the two cell lines. US 9,260,522 B2 49 50 FIG 11 c-MET antigen with human membrane-distal epitopes and FACS binding analysis of designated bispecific single not to the murine c-MET antigen. chain constructs to CHO cells expressing human FAPalpha as FIG. 15 described in Example 6.1, the human CD3+ T cell line HPB The diagrams show results of chromium release assays ALL, CHO cells expressing macaque FAPalpha as described 5 measuring cytotoxic activity induced by designated c-MET in Example 6.15 and the macaque T cell line 4119LnPx, specific single chain constructs redirected to the indicated respectively. The FACS staining was performed as described target cell lines generated as described in Example 7.17. in Examples 6.13 and 6.16. The bold lines represent cells Effector cells were also used as indicated. The assays were incubated with cell culture supernatant of transfected cells performed as described in Example 7.14. The diagrams 10 clearly demonstrate for each construct the potent recruitment expressing the bispecific antibody constructs. The thin lines of cytotoxic activity of human effector T cells against target represent the negative controls. Supernatant of untransfected cells positive for human c-MET. No significant recruitment of CHO cells was used as negative control. For each bispecific cytotoxic activity of human effector T cells against target single chain construct the overlays of the histograms show cells positive for murine c-MET and target cells positive for specific binding of the construct to human and macaque 15 the mutated murine c-MET antigen with human membrane FAPalpha and human and macaque CD3. distal epitopes, respectively, was detectable. FIG. 12 FIG 16 The diagrams show results of chromium release assays FACS binding analysis of designated cross-species specific measuring cytotoxic activity induced by designated FAPal schv antibodies to CHO cells expressing human c-MET as pha specific single chain constructs redirected to the indicated 20 described in Example 7.17. CHO cells expressing the murine target cell lines generated as described in Examples 6.1, 6.3 c-MET antigen as described in Example 7.17, CHO cells and 6.4. Effector cells were also used as indicated. The assays expressing the mutated human c-MET antigen with murine were performed as described in Example 6.14. The diagrams membrane-distal epitopes as described in Example 7.17 and clearly demonstrate for each construct the potent recruitment CHO cells expressing the mutated murine c-MET antigen of cytotoxic activity of human effector T cells against target 25 with human membrane-distal epitopes as described in cells positive for human FAPalpha and target cells positive for Example 7.17, respectively. The FACS staining was per the mutated human FAPalpha antigen with murine mem formed as described in Example 7.9. The bold lines represent brane-distal epitopes. No significant recruitment of cytotoxic cells incubated with periplasmic preparations containing the activity of human effector T cells against target cells positive c-MET specific sclv antibodies. The filled histograms show for murine FAPalpha was detectable. 30 the negative controls. The Buffer used for periplasmic prepa FIG. 13 rations was used as negative control. For each c-MET specific FACS binding analysis of designated bispecific single scFv antibody the overlays of the histograms show specific chain constructs to CHO cells expressing human c-MET as binding of the construct to human c-MET and human c-MET described in Example 7.17, the human CD3+ T cell line with murine membrane-distal epitopes. No significant bind HPB-ALL, CHO cells expressing macaque c-MET as 35 ing to cells positive for murine c-MET and to cells positive for described in Example 7.17 and the macaque T cell line the mutated murine c-MET with human membrane-distal 41 19LnPx, respectively. The FACS staining was performed epitopes, respectively, was detectable. as described in Examples 7.13 and 7.16. The bold lines rep FIG. 17 resent cells incubated with cell culture supernatant of trans FACS binding analysis of designated bispecific single fected cells expressing the bispecificantibody constructs. The 40 chain constructs to CHO cells expressing human IGF-1R as filled histograms show the negative controls. Supernatant of described in Example 9.1, the human CD3+ T cell line HPB untransfected CHO cells was used as negative control. For ALL, CHO cells expressing macaque IGF-1R as described in each bispecific single chain construct the overlays of the Example 9.15 and the macaque T cell line 4119LnPx, respec histograms show specific binding of the construct to human tively. The FACS staining was performed as described in and macaque c-MET and human and macaque CD3. 45 Examples 9.13 and 9.16. The bold lines represent cells incu FIG. 14 bated with cell culture supernatant of transfected cells FACS binding analysis of designated bispecific single expressing the bispecific antibody constructs. The filled his chain constructs to CHO cells expressing the murine c-MET tograms show the negative controls. Supernatant of untrans antigen as described in Example 7.17. CHO cells expressing fected CHO cells was used as negative control. For each the mutated human c-MET antigen with murine membrane- 50 bispecific single chain construct the overlays of the histo distal epitopes as described in Example 7.17 and CHO cells grams show specific binding of the construct to human and expressing the mutated murine c-MET antigen with human macaque IGF-1R and human and macaque CD3. membrane-distal epitopes as described in Example 7.17. FIG. 18 respectively. The FACS staining was performed as described FACS binding analysis of designated bispecific single in Example 7.13. The bold lines represent cells incubated 55 chain constructs to CHO cells expressing the murine IGF-1R with cell culture Supernatant of transfected cells expressing antigen as described in Example 9.3, CHO cells expressing the bispecific antibody constructs. The filled histograms show the mutated human IGF-1R antigen with murine membrane the negative controls. Supernatant of untransfected CHO distal epitopes as described in Example 9.4 and CHO cells cells was used as negative control. An anti-FLAG M2 anti expressing the mutated murine IGF-1R antigen with human body was used to detect expression levels of the respective 60 membrane-distal epitopes as described in Example 9.5, antigens. For each cell line the overlay of the histograms for respectively. The FACS staining was performed as described the anti-FLAG M2 antibody shows high expression levels of in Example 9.13. The bold lines represent cells incubated the respective antigen. Expression levels were comparable for with cell culture Supernatant of transfected cells expressing the three cell lines. For each bispecific single chain construct the bispecific antibody constructs. The filled histograms show the overlays of the histograms show specific binding of the 65 the negative controls. Supernatant of untransfected CHO construct to the mutated human c-MET antigen with murine cells was used as negative control. For each bispecific single membrane-distal epitopes but not to the mutated murine chain construct the overlays of the histograms show specific US 9,260,522 B2 51 52 binding of the construct to the mutated human IGF-1R anti acid (i.e. the reference aa) as counted from the junction of gen with murine membrane-distal epitopes but not to the transmembrane and extracellular region are marked in bold. mutated murine IGF-1R antigen with human membrane-dis The distances between alpha C-atoms of two amino acids tal epitopes and not to the murine IGF-1R antigen. within human PSMA were determined using the crystal struc FIG. 19 ture of human PSMA (accession No 1 Z8L; obtained from the The diagrams show results of chromium release assays RCSB pab, protein data bank of the Research Collaboratory measuring cytotoxic activity induced by designated IGF-1R for Structural Bioinformatics) and the “measure distance specific single chain constructs redirected to the CHO cells mode' of the software "3D molecule viewer' (a component expressing human IGF-1R as described in Example 9.1. of Vector NTIR Suite 8.0, Informax Inc.). Numbering of Effector cells were used as indicated. The assays were per 10 amino acid positions refers to human PSMA and starts with formed as described in Example 9.14. The diagrams clearly the amino acid methionine encoded by the start codon of demonstrate for each construct the potent recruitment of cyto human PSMA. The reference aa is histidine at position 56. toxic activity of human effector T cells against target cells Table 2 positive for human IGF-1R. All extracellular amino acids of human FAPalpha mis FIG. 20 15 matching with the homologous murine FAPalpha amino acid FACS binding analysis of designated bispecific single sequence. Those mismatched extracellular human FAPalpha chain constructs to CHO cells expressing designated human/ amino acids, whose alpha C-atoms have a distance of>60 A rat PSMA chimeras as described in Example 10.2.1. The from the alpha C-atom of the thirteenth extracellular human FACS staining was performed as described in Example FAPalpha amino acid (i.e. the reference aa) as counted from 10.2.2. The bold lines represent cells incubated with cell the junction of transmembrane and extracellular region are culture Supernatant of transfected cells expressing the bispe marked in bold. The distances between alpha C-atoms of two cific antibody constructs. The filled histograms show the amino acids within human FAPalpha were determined using negative controls. Supernatant of untransfected CHO cells the crystal structure of human FAPalpha (Accession No was used as negative control. For each bispecific single chain 1Z68; obtained from the RCSB pab, protein data bank of the construct the overlays of the histograms show specific bind 25 Research Collaboratory for Structural Bioinformatics: http:// ing of the construct to the chimeric constructs huPSMA www.rcsb.org/pdb) and the “measure distance mode of the rat 140-169, huPSMArat281-284, huPSMArat300-344, software "3D molecule viewer' (a component of Vector NTI huPSMArat683-690 and huPSMArat716-750. Compared Suite 8.0, Informax Inc.). Numbering of amino acid positions with the signals obtained for the other bispecific single chain refers to human FAPalpha and starts with the amino acid construct there is a clear lack of binding for the bispecific 30 methionine encoded by the start codon of human FAPalpha. single chain antibody construct PSMA-P7 HLXI2CHL to the The reference aa is methionine at position 39. chimeric PSMA construct huPSMArats 98-617. The present invention is additionally described by way of FIG 21 the following illustrative non-limiting examples that provide The Figure shows binding signals obtained with periplas a better understanding of the present invention and of its many mic preparations of the schv antibody of the PSMA specific 35 advantages. binder of PSMA-D4 HLXI2C HL to 15-mer peptides span ning over the extracellular domain of human PSMA and EXAMPLES overlapping with their neighboring peptides by 14 amino acids. Signals obtained for the peptides are plotted on the 1. Cytotxicity with Respect to the Distance of the X-axis in order of the N-terminal peptides on the left to the 40 Target Cell Epitopes Distance from the Target Cell C-terminal peptides on the right. Strength of ELISA signals Membrane using His detection is plotted on the Y-axis. The ELISA was performed as described in Example 10.3. A distinct maxi 1.1. Generation of CHO Cells Expressing the Human mum signal is detectable for the peptide spanning over the EpCAM Antigen amino acids threonine 334 to threonine 339. 45 The sequence of the human EpCAM antigen (NM FIG.22 002354, Homo sapiens tumor-associated calcium signal The diagram shows results of a CytoTox-GloTM cytotoxic transducer 1 (TACSTD1), mRNA, National Center for Bio ity assay measuring cytotoxic activity of unstimulated human technology Information, http://www.ncbi.nlm.nih.gov/en T cells induced by designated PSMA specific bispecific trez) was used to obtain a synthetic molecule by gene synthe single chain constructs against CHO cells expressing human 50 sis according to standard protocols. The gene synthesis PSMA as described in Example 2.1. The assay was performed fragment was also designed as to contain a Kozak site for as described in Example X.4. The diagram clearly demon eukaryotic expression of the construct and restriction sites at strates the superior cytotoxic activity of PSMA bispecific the beginning and the end of the DNA. The introduced restric single chain antibody PSMA-P7 HLXI2C HL directed at a tion sites Xbal at the 5' end and Salat the 3' end were utilised membrane-proximal target epitope of human PSMA over 55 in the following cloning procedures. The gene synthesis frag PSMA bispecific single chain antibody PSMA-D4 HLXI2C ment was cloned via Xbaland Sal into a plasmid designated HL directed at a membrane-distal target epitope of human pEF-DHFR (pEF-DHFR is described in Raum et al. Cancer PSMA. Immunol Immunother 50 (2001) 141-150). The aforemen tioned procedures were carried out according to standard TABLE LEGENDS 60 protocols (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd edition, Cold Spring Harbour Laboratory Press, Table 1 Cold Spring Harbour, N.Y. (2001)). A clone with sequence All extracellular amino acids of human PSMA mismatch verified nucleotide sequence was transfected into DHFR defi ing with the homologous rat PSMA amino acid sequence. cient CHO cells for eukaryotic expression of the construct. Those mismatched extracellular human PSMA amino acids, 65 Eukaryotic protein expression in DHFR deficient CHO cells whose alpha C-atoms have a distance of>60 A from the alpha was performed as described by Kaufmann R. J. (1990) Meth C-atom of the thirteenth extracellular human PSMA amino ods Enzymol. 185, 537-566. Gene amplification of the con US 9,260,522 B2 53 54 struct was induced by increasing concentrations of methotr stationary culture cell culture Supernatant was collected and exate (MTX) to a final concentration of up to 20 nM MTX. used in the Subsequent experiments. 1.2. Generation of CHO Cells Expressing EpCAM-hNG2 1.4. Flowcytometry of CHO Cells Transfected with Different Fusion Proteins EpCAM-hNG2 Fusion Proteins The coding sequences of EpCAM-hNG2 fusion proteins The presence of the EpCAM-epitope of bispecific single EpCAM-D1-hNG2 (SEQID Nos 189 and 190), EpCAM-D3 chain antibody 5-10xI2C on the CHO cells transfected with hNG2 (SEQ ID Nos 191 and 192), EpCAM-D1D3-hNG2 native EpCAM and the EpCAM-hNG2 fusion proteins (SEQID Nos 193 and 194), EpCAM-D1D2-hNG2 (SEQID EpCAM-D1-hNG2, EpCAM-D3-hNG2, EpCAM-D1D3 Nos 195 and 196) and EpCAM-hNG2 (SEQID Nos 197 and hNG2, EpCAM-D1D2-hNG2 and EpCAM-hNG2, respec 198) were obtained by gene synthesis according to standard 10 tively, was confirmed by flow cytometry with the murine protocols. The gene synthesis fragment was designed as to parental IgG1 antibody Mab 5-10 as described (Brischwein contain first the coding sequence of an immunoglobulin (2007) J Immunother 30: 798-807). The result is shown in leader peptide followed subsequently by human EpCAM, the FIG 1. respective extracellular part of human NG2 and the trans 15 1.5. T Cell Cytotoxicity Redirected by bscAb 5-10x12C membrane and cytoplasmic domain of the human NG2 (Plus Against CHO Cells Transfected with Different EpCAM chke (1996) PNAS 93: 9710-9715). The different compo hNG2 Fusion Proteins nents were connected by short peptide linkers. The gene T cell cytotoxicity redirected by bscAb 5-10xI2C against synthesis fragments were also designed as to introduce CHO cells transfected with different EpCAM-hNG2 fusion restriction sites at the 5' end (EcoRI) and at the 3' end (Sal I) proteins was measured in a chromium 51 (Cr) release in for cloning into the mammalian cell expression vector pEF vitro cytotoxicity assay. As source of effector T cells stimu DHFR(pEF-DHFR is described in Raumetal. Cancer Immu lated human CD4/CD56 depleted PBMC were used. Stimu nol Immunother 50 (2001) 141-150). The aforementioned lated human PBMC were obtained as follows: A Petri dish procedures were carried out according to standard protocols (145 mm diameter, Greiner bio-one GmbH, Kremsmünster) (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd 25 was coated with a commercially available anti-CD3 specific edition, Cold Spring Harbour Laboratory Press, Cold Spring antibody (e.g. OKT3. Orthoclone) in a final concentration of Harbour, N.Y. (2001)). For each EpCAM-hNG2 fusion pro 1 ug/ml for 1 hour at 37°C. Unbound protein was removed by tein a clone with sequence-verified nucleotide sequence was one washing step with PBS. The fresh PBMC were isolated transfected into DHFR deficient CHO cells for eukaryotic from peripheral blood (30-50 ml human blood) by Ficoll expression. Eukaryotic protein expression in DHFR deficient 30 gradient centrifugation according to standard protocols. 3-5x CHO cells was performed as described by Kaufmann R. J. 107 PBMC were added to the precoated petridish in 120 ml of (1990) Methods Enzymol. 185, 537-566. Gene amplification RPMI 1640 with stabilized glutamine/10% FCS/IL-2 20 of the construct was induced by increasing concentrations of U/ml (Proleukin, Chiron) and stimulated for 2 days. On the methotrexate (MTX) to a final concentration of up to 20 nM third day the cells were collected and washed once with RPMI MTX. 35 1640. IL-2 was added to a final concentration of 20U/ml and 1.3. Generation of the EpCAM and CD3 Bispecific Single the cells were cultured again for one day in the same cell Antibody 5-10xI2C culture medium as above. By depletion of CD4+ T cells and The bispecific single chain antibody 5-10xI2C comprising CD56+ NK cells according to standard protocols CD8+ cyto the sclv binding domain 5-10 (in VL-VH arrangement) toxic T lymphocytes (CTLs) were enriched. Target cells were directed at human EpCAM (Brischwein (2007).JImmunother 40 washed twice with PBS and labelled with 11.1 MBq'Crina 30: 798-807) and the scEv binding domain I2C (in VL-VH final volume of 100 ul RPMI with 50% FCS for 60 minutes at arrangement) directed at CD3epsilon on human T cells was 37° C. Subsequently the labelled target cells were washed 3 obtained by gene synthesis. The gene synthesis fragment was times with 5 ml RPMI and then used in the cytotoxicity assay. designed as to contain first a Kozak site for eukaryotic expres The assay was performed in a 96 well plate in a total volume sion of the construct, followed by a 19 amino acid immuno 45 of 250 ul supplemented RPMI (as above) with an E:T ratio of globulin leader peptide, followed in frame by the coding 10:1. BscAb 5-10xI2C was added as culture supernatant from sequence of the bispecific single chain antibody 5-10xI2C, transfected CHO cells at different dilutions. The assay time followed in frame by the coding sequence of a 6 histidine tag was 18 hours. Cytotoxicity was measured as relative values of and a stop codon (the cDNA and amino acid sequence of the released chromium related to the difference of maximum construct is listed under SEQID Nos 199 and 200). The gene 50 lysis (addition of Triton-X) and spontaneous lysis (without synthesis fragment was also designed as to introduce Suitable effector cells). All measurements were carried out in quadru restriction sites at the beginning (EcoRI) and at the end of the plicates. Measurement of released chromium activity was fragment (Sal I) for cloning of the gene synthesis fragment performed with a Wizard R. 3" gammacounter (Perkin Elmer into the mammalian cell expression vectorpEF-DHFR (pEF Life Sciences GmbH, Koln, Germany). Analysis of the DHFR is described in Raum et al. Cancer Immunol Immu 55 experimental data was performed with Prism 4 for Win nother 50 (2001) 141-150). The aforementioned procedures dows(R (version 4.02, GraphPad Software Inc., San Diego, were carried out according to standard protocols (Sambrook, Calif., USA). Sigmoidal dose response curves typically had Molecular Cloning: A Laboratory Manual, 3rd edition, Cold R values >0.90 as determined by the software. As shown in Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. FIG. 2 the T cell cytotoxicity redirected by bscAb 5-10xI2C (2001)). A clone with sequence-verified nucleotide sequence 60 against the indicated target cells critically depends on the was transfected into DHFR deficient CHO cells for eukary varying distance of the target EpCAM epitope of bsc 5-10x otic expression of the construct. Eukaryotic protein expres I2C in the different EpCAM-hNG2 model antigens from the sion in DHFR deficient CHO cells was performed as target cell membrane. The descending order of T cell cyto described by Kaufmann R.J. (1990) Methods Enzymol. 185, toxicity with increasing membrane distance of the target 537-566. Gene amplification of the construct was induced by 65 epitope as given in parentheses is EpCAM-CHO (positive increasing concentrations of methotrexate (MTX) to a final control)>EpCAM-D1-hNG2 (640 aa)>EpCAM-D3-hNG2 concentration of up to 20 nM MTX. After two passages of (679 aa)>EpCAM-D1D3-hNG2 (871 aa). There was no cyto US 9,260,522 B2 55 56 toxic activity detectable against CHO cells expressing (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd EpCAM-D1 D2-hNG2 (1511 aa) or EpCAM-hNG2 (21.90 edition, Cold Spring Harbour Laboratory Press, Cold Spring aa). Harbour, N.Y. (2001)). A clone with sequence-verified nucle otide sequence is transfected into DHFR deficient CHO cells 2. Generation of Bispecific Single Chain Antibodies for eukaryotic expression of the construct. Eukaryotic protein Directed at Membrane-proximal Target Epitopes of expression in DHFR deficient CHO cells is performed as Human PSMA described by Kaufmann R.J. (1990) Methods Enzymol. 185, 537-566. Gene amplification of the construct is induced by 2.1 Generation of CHO Cells Expressing Human PSMA increasing concentrations of methotrexate (MTX) to a final The coding sequence of human PSMA as published in 10 concentration of up to 20 nM MTX. After two passages of GenBank (Accession number NM 004476) is obtained by stationary culture the cells are grown in roller bottles with gene synthesis according to standard protocols. The gene nucleoside-free HyO PF CHO liquid soy medium (with 4.0 synthesis fragment is designed as to contain first a Kozak site mML-Glutamine with 0.1% Pluronic F-68; HyClone) for 7 for eukaryotic expression of the construct followed by the days before harvest. The cells are removed by centrifugation coding sequence of the human PSMA protein and a stop 15 and the Supernatant containing the expressed protein is stored codon (the cDNA and amino acid sequence of the construct is at -20° C. Alternatively a clone of the expression plasmid listed under SEQID Nos 201 and 202). The gene synthesis with sequence-verified nucleotide sequence is used for trans fragment is also designed as to introduce restriction sites at fection and protein expression in the FreeStyle 293 Expres the beginning and at the end of the fragment. The introduced sion System (Invitrogen GmbH, Karlsruhe, Germany) restriction sites, Xbal at the 5' end and Sal at the 3' end, are according to the manufacturer's protocol. Supernatant con utilized in the following cloning procedures. The gene Syn taining the expressed protein is obtained, cells are removed by thesis fragment is cloned via Xbal and SalI into a plasmid centrifugation and the Supernatant is stored at -20°C. designated pEF-DHFR (pEF-DHFR is described in Raum et Purification of the soluble human PSMA fusion protein is al. Cancer Immunol Immunother 50 (2001) 141-150) follow performed as follows: Akta R. Explorer System (GE Health ing standard protocols. The aforementioned procedures are 25 Systems) and Unicorn R Software are used for chromatogra carried out according to standard protocols (Sambrook, phy. Immobilized metal affinity chromatography (“IMAC) Molecular Cloning: A Laboratory Manual, 3rd edition, Cold is performed using a Fractogel EMD Chelate R (Merck) Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. which is loaded with ZnCl2 according to the protocol pro (2001)). A clone with sequence-verified nucleotide sequence vided by the manufacturer. The column is equilibrated with is transfected into DHFR deficient CHO cells for eukaryotic 30 buffer A (20 mM sodium phosphate buffer pH 7.2, 0.1 M expression of the construct. Eukaryotic protein expression in NaCl) and the cell culture supernatant (500 ml) is applied to DHFR deficient CHO cells is performed as described by the column (10 ml) at a flow rate of 3 ml/min. The column is Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. washed with buffer A to remove unbound sample. Bound Gene amplification of the construct is induced by increasing protein is eluted using a two step gradient of buffer B (20 mM concentrations of methotrexate (MTX) to a final concentra 35 sodium phosphate buffer pH 7.2, 0.1 M NaCl, 0.5 MImida tion of up to 20 nM MTX. Zole) according to the following procedure: 2.2 Generation of a Soluble Human PSMA Fusion Protein Step 1: 20% Buffer B in 6 Column Volumes The coding sequence of human PSMA as described in Step 2: 100% Buffer B in 6 Column Volumes Example 2.1 and the coding sequence of murine Lag3 as Eluted protein fractions from step 2 are pooled for further published in GenBank (Accession number NM 008479) are 40 purification. All chemicals are of research grade and pur used for the construction of an artificial clNA sequence chased from Sigma (Deisenhofen) or Merck (Darmstadt). encoding a soluble fusion protein of human PSMA and Gel filtration chromatography is performed on a HiLoad murine Lag3. To generate a construct for expression of the 16/60 Superdex 200 prep grade column (GE/Amersham) soluble human PSMA fusion protein a cDNA fragment is equilibrated with Equi-buffer (25 mM Citrate, 200 mM obtained by gene synthesis according to standard protocols 45 Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow (the cDNA and amino acid sequence of the construct is listed rate 1 ml/min) are subjected to standard SDS-PAGE and under SEQ ID Nos 203 and 204). The gene synthesis frag Western Blot for detection. Prior to purification, the columnis ment is designed as to contain first a Kozak site for eukaryotic calibrated for molecular weight determination (molecular expression of the construct followed by the coding sequence weight marker kit, Sigma MW GF-200). Protein concentra of the murine Lag3 protein from amino acid 1 to 441 corre 50 tions are determined using OD280 nm. sponding to the signal peptide and extracellular domains of 2.3 Generation of CHO Cells Expressing Rat PSMA murine Lag3, followed in frame by the coding sequence of an The sequence of rat PSMA (NM 057185, Rattus norvegi artificial Ser-Gly-Ser-linker, followed in frame by the cod cus folate hydrolase (Folh1), mRNA, National Center for ing sequence of the human PSMA protein from amino acid 44 Biotechnology Information, http://www.ncbi.nlm.nih.gov/ to 750 corresponding to the extracellular domains of human 55 entrez) is used to obtain a synthetic cDNA molecule by gene PSMA, followed in frame by the coding sequence of an synthesis according to standard protocols. The gene synthesis artificial Ser-Gly-linker, followed in frame by the coding fragment is designed as to contain first a Kozak site for sequence of a 6 histidine tag and a stop codon. The gene eukaryotic expression of the construct followed by the com synthesis fragment is also designed as to introduce restriction plete coding sequence of the rat PSMA antigen, followed in sites at the beginning and at the end of the fragment. The 60 frame by the coding sequence of a FLAG-tag and a stop codon introduced restriction sites, Xbal at the 5' end and Sal at the (the cDNA and amino acid sequence of the construct is listed 3' end, are utilized in the following cloning procedures. The under SEQ ID Nos 205 and 206). The gene synthesis frag gene synthesis fragment is cloned via Xbal and SalI into a ment is also designed as to introduce restriction sites at the 5' plasmid designated pEF-DHFR (pEF-DHFR is described in end (EcoRI) and at the 3' end (SalI) of the cDNA fragment for Raum et al. Cancer Immunol Immunother 50 (2001) 141 65 cloning into the mammalian cell expression vector pEF 150) following standard protocols. The aforementioned pro DHFR(pEF-DHFR is described in Raumetal. Cancer Immu cedures are all carried out according to standard protocols nol Immunother 50 (2001) 141-150). The aforementioned US 9,260,522 B2 57 58 procedures are carried out according to standard protocols described in Raum et al. Cancer Immunol Immunother 50 (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd (2001) 141-150). The aforementioned procedures are carried edition, Cold Spring Harbour Laboratory Press, Cold Spring out according to standard protocols (Sambrook, Molecular Harbour, N.Y. (2001)). A clone with sequence-verified nucle Cloning: A Laboratory Manual, 3rd edition, Cold Spring otide sequence is transfected into DHFR deficient CHO cells Harbour Laboratory Press, Cold Spring Harbour, N.Y. for eukaryotic expression of the construct. Eukaryotic protein (2001)). A clone with sequence-verified nucleotide sequence expression in DHFR deficient CHO cells is performed as is transfected into DHFR deficient CHO cells for eukaryotic described by Kaufmann R.J. (1990) Methods Enzymol. 185, expression of the construct. Eukaryotic protein expression in 537-566. Gene amplification for increased antigen expres DHFR deficient CHO cells is performed as described by sion is induced by increasing concentrations of methotrexate 10 Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. (MTX) to a final concentration of up to 20 nM MTX. Gene amplification for increased antigen expression is 2.4 Generation of CHO Cells Expressing a Mutated Human induced by increasing concentrations of methotrexate (MTX) PSMA Antigen with Rat Membrane-distal Epitopes to a final concentration of up to 20 nM MTX. The coding sequence of a mutated human PSMA antigen 2.6 Immunization of Mice Using a Soluble Human PSMA with rat membrane-distal epitopes is obtained by gene Syn 15 Fusion Protein thesis according to standard protocols The gene synthesis Twelve weeks old F1 mice from BALB? cxC57BL/6 cross fragment is designed as to contain first a Kozak site for ings are immunized with the soluble human PSMA fusion eukaryotic expression of the construct followed by the com protein as described in Example 2.2 To this end for each plete coding sequence of the human PSMA antigen mutated animal 40 ug of the soluble human PSMA fusion protein are at 12 specific amino acid positions as explained below, fol mixed with 10 nmol of a thioate-modified CpG-Oligonucle lowed in frame by the coding sequence of a FLAG tag and a otide (5'-tccatgacgttcctgatgct-3") in 300 ul PBS and are stop codon (the cDNA and amino acid sequence of the con injected intraperitoneally. Mice receive booster immuniza struct is listed under SEQID Nos 207 and 208). All extracel tions after 21, 42 and optionally 63 days in the same way. Ten lular amino acids of human PSMA mismatching with the days after the first booster immunization, blood samples are homologous rat sequence, whose alpha C-atoms have a dis 25 taken and antibody serum titers against human PSMA are tance of >60 A from the alpha C-atom of the thirteenth extra tested by flow cytometry according to standard protocols. To cellular amino acid (i.e. the reference aa) as counted from the this end 200,000 cells of the human PSMA transfected CHO junction of transmembrane and extracellular region, are cells as described in Example 2.1 are incubated for 30 min on mutated to the homologous mismatched rat amino acid. This ice with 50 ul of serum of the immunized animals diluted applies to the 12 amino acids that are listed in table Table 1 30 1:1000 in PBS with 2% FCS. The cells are washed twice in and marked in bold. The homologous mismatched amino PBS with 2% FCS and binding of serum antibodies is acids between human and rat PSMA are identified by detected with an mouse Fc gamma-specific antibody (Di sequence alignment as shown in FIG. 3. The gene synthesis anova) conjugated to phycoerythrin, diluted 1:100 in PBS fragment is also designed as to introduce restriction sites at with 2% FCS. Serum of the animals obtained prior to immu the 5' end (Xba I) and at the 3' end (Sal I) of the cDNA 35 nization is used as a negative control. Flow cytometry is fragment for cloning into the mammalian cell expression performed on a FACS-Calibur apparatus, the CellOuest soft vector pEF-DHFR (pEF-DHFR is described in Raum et al. ware is used to acquire and analyze the data (Becton Dickin Cancer Immunol Immunother 50 (2001) 141-150). The afore son biosciences, Heidelberg). FACS staining and measuring mentioned procedures are carried out according to standard of the fluorescence intensity are performed as described in protocols (Sambrook, Molecular Cloning: A Laboratory 40 Current Protocols in Immunology (Coligan, Kruisbeek, Mar Manual, 3rd edition, Cold Spring Harbour Laboratory Press, gullies. Shevach and Strober, Wiley-Interscience, 2002). Cold Spring Harbour, N.Y. (2001)). A clone with sequence Animals demonstrating significant serum reactivity verified nucleotide sequence is transfected into DHFR defi against human PSMA as determined by the FACS analysis cient CHO cells for eukaryotic expression of the construct. are used in the Subsequent experiment. Eukaryotic protein expression in DHFR deficient CHO cells 45 2.7 Generation of an Immune Murine Antibody schv Library: is performed as described by Kaufmann R.J. (1990) Methods Construction of a Combinatorial Antibody Library and Phage Enzymol. 185, 537-566. Gene amplification for increased Display antigen expression is induced by increasing concentrations of Three days after the last booster immunization spleen cells methotrexate (MTX) to a final concentration of up to 20 nM of reactive animals are harvested for the preparation of total MTX. 50 RNA according to standard protocols. 2.5 Generation of CHO Cells Expressing a Mutated Rat A library of murine immunoglobulin (Ig) light chain PSMA Antigen with Human Membrane-distal Epitopes (kappa) variable region (VK) and Ig heavy chain variable The coding sequence of a mutated rat PSMA antigen with region (VH) DNA-fragments is constructed by RT-PCR on human membrane-distal epitopes is obtained by gene synthe murine spleen RNA using VK- and VH specific primers. sis according to standard protocols The gene synthesis frag 55 cDNA is synthesized according to standard protocols. ment is designed as to contain first a Kozak site for eukaryotic The primers are designed in a way to give rise to a 5'-XhoI expression of the construct followed by the complete coding and a 3'-BstEII recognition site for the amplified heavy chain sequence of the rat PSMA antigen mutated at the same 12 V-fragments and to a 5'-SacI and a 3'-Spel recognition site for specific extracellular amino acid positions as identified in the amplified VK DNA fragments. foregoing Example 2.4 to the respective homologous human 60 For the PCR-amplification of the VH DNA-fragments amino acid, followed in frame by the coding sequence of a eight different 5'-VH-family specific primers (MVH1(GC) FLAG tag and a stop codon (the cDNA and amino acid AG GTG CAG CTC GAG GAG TCAGGA CCT SEQ ID sequence of the construct is listed under SEQID Nos 209 and NO: 211; MVH2 GAG GTC CAG CTC GAG CAG TCT 210). The gene synthesis fragment is also designed as to GGA CCTSEQID NO: 212; MVH3 CAG GTCCAA CTC introduce restriction sites at the 5' end (EcoRII) and at the 3' 65 GAG CAGCCT GGG GCT SEQID NO: 213: MVH4 GAG end (Sal I) of the cDNA fragment for cloning into the mam GTT CAG CTC GAG CAG TCT GGG GCASEQ ID NO: malian cell expression vector pEF-DHFR (pEF-DHFR is 214; MVH5 GA(AG) GTGAAGCTC GAG GAG TCTGGA US 9,260,522 B2 59 60 GGASEQ ID NO: 215; MVH6 GAG GTGAAG CTTCTC 0.4 ml of PBS/0.1% BSA and incubated with 10 to 107 CHO GAG TCTGGAGGT SEQID NO: 216; MVH7 GAA GTG cells expressing the mutated human PSMA antigen with rat AAG CTC GAG GAG TCT GGGGGA SEQID NO: 217; membrane-distal epitopes as described in example 2.4 for 1 MVH8 GAG GTT CAG CTC GAG CAG TCT GGA GCT hour on ice under slow agitation. These CHO cells are grown SEQID NO: 218) are each combined with one 3'-VH primer 5 beforehand, harvested by centrifugation, washed in PBS and (3'MuVHBstEIItgagga gacggt gaccgtggt ccc ttggcc cca g resuspended in PBS/1% FCS (containing Na Azide). schv SEQIDNO: 219); for the PCR amplification of the VK-chain phage which do not specifically bind to the CHO cells are fragments seven different 5'-VK-family specific primers eliminated by up to five washing steps with PBS/1% FCS (MUVK1 CCA GTTCCGAGCTCG TTGTGA CTCAGG (containing Na Azide). After washing, binding entities are AAT CT SEQID NO: 220; MUVK2 CCA GTT CCGAGC 10 eluted from the cells by resuspending the cells in HCl-glycine TCG TGT TGA CGC AGC CGC CC SEQ ID NO: 221; pH 2.2 (10 min incubation with Subsequent Vortexing) and MUVK3 CCA GTT CCG AGC TCG TGC TCA CCC AGT after neutralization with 2M Tris pH 12, the eluate is used for CTC CASEQ ID NO: 222; MUVK4 CCA GTTCCGAGC infection of a fresh uninfected E. coli XL1 Blue culture TCC AGA TGA CCC AGT CTC CASEQ ID NO: 223: (OD600>0.5). The E. coli culture containing E. coli cells MUVK5 CCA GAT GTG. AGC TCG TGA TGA CCC AGA 15 Successfully transduced with a phagemid copy, encoding a CTC CASEQ ID NO: 224; MUVK6 CCA GAT GTGAGC murine ScFV-fragment, are again selected for carbenicillin TCG TCA TGA CCC AGT CTC CASEQ ID NO: 225; resistance and subsequently infected with VCMS13 helper MUVK7 CCA GTT CCG AGC TCG TGA TGA CAC AGT phage to start the second round of antibody display and in CTC CASEQ ID NO: 226) are each combined with one vitro selection. Typically a total of 4 to 5 rounds of selections 3'-VK primer (3"MuVkHindIII/BsiW1 tygtgc act agt cgtacg are carried out. tttgat cte aag cttggt ccc SEQID NO: 227). 2.9 Screening for Membrane-proximal Target Binders on The following PCR program is used for amplification: CHO Cells Expressing the Human Psma Antigen, the Rat denaturation at 94° C. for 20 sec; primer annealing at 52°C. PSMA Antigen and the Mutated Rat PSMA Antigen with for 50 sec and primer extension at 72° C. for 60 sec and 40 Human Membrane-distal Epitopes cycles, followed by a 10 min final extension at 72°C. 25 Plasmid DNA corresponding to 4 and 5 rounds of panning 450 ng of the kappa light chain fragments (SacI-Spel is isolated from E. coli cultures after selection. For the pro digested) are ligated with 1400 ng of the phagemid duction of soluble schv-protein, VH-VL-DNA fragments are pComb3H5Bhis (SacI-Spel digested; large fragment). The excised from the plasmids (XhoI-SpeI). These fragments are resulting combinatorial antibody library is then transformed cloned via the same restriction sites in the plasmid into 300 ul of electrocompetent Escherichia coli XL 1 Blue 30 pComb3H5BFlag? His differing from the original cells by electroporation (2.5 kV, 0.2 cm gap cuvette, 25uFD, pComb3H5BHis in that the expression construct (e.g. scFv) 200 Ohm, Biorad gene-pulser) resulting in a library size of includes a Flag-tag (TGDYKDDDDK) between the scFv and more than 107 independent clones. After one hour of pheno the His6-tag and the additional phage proteins are deleted. type expression, positive transformants are selected for car After ligation, each pool (different rounds of panning) of benicillin resistance encoded by the pComb3H5BHis vector 35 plasmid DNA is transformed into 100 ul heat shock compe in 100 ml of liquid super broth (SB)-culture over night. Cells tent E. coli TG1 or XLI blue and plated onto carbenicillin are then harvested by centrifugation and plasmid preparation LB-agar. Single colonies are picked into 100 ul of LB carb is carried out using a commercially available plasmid prepa (LB with 50 lug/ml carbenicillin). ration kit (Qiagen). After induction with 1 mMIPTG. E. coli transformed with 2800 ng of this plasmid-DNA containing the VK-library 40 pComb3H5BFlag? His containing a VL- and VH-segment (XhoI-BstEII digested; large fragment) are ligated with 900 produce soluble scFv in sufficient amounts. Due to a suitable ng of the heavy chain V-fragments (XhoI-BstEII digested) signal sequence, the Sclv is exported into the periplasma and again transformed into two 300 ulaliquots of electrocom where it folds into a functional conformation. petent E. coli XL 1 Blue cells by electroporation (2.5 kV, 0.2 Single E. coli bacterial colonies from the transformation cm gap cuvette, 25uFD, 200Ohm) resulting in a totalVH-VK 45 plates are picked for periplasmic Small scale preparations and schv (single chain variable fragment) library size of more grown in SB-medium (e.g. 10 ml) supplemented with 20 mM than 107 independent clones. MgCl, and carbenicillin 50 ug/ml (and re-dissolved in PBS After phenotype expression and slow adaptation to carbe (e.g. 1 ml) after harvesting. A temperature shock is applied by nicillin, the E. coli cells containing the antibody library are four rounds of freezing at -70° C. and thawing at 37° C. transferred into SB-Carbenicillin (SB with 50 lug/mL carbe 50 whereby the outer membrane of the bacteria is destroyed and nicillin) selection medium. The E. coli cells containing the the soluble periplasmic proteins including the Sclvs are antibody library are then infected with an infectious dose of released into the supernatant. After elimination of intact cells 10' particles of helper phage VCSM13 resulting in the pro and cell-debris by centrifugation, the Supernatant containing duction and secretion of filamentous M13 phage, wherein the murine anti-human PSMA-scFVs is collected and used for each phage particle contains single stranded 55 further examination. pComb3H5BHis-DNA encoding a murine schv-fragment Screening of the isolated schvs for membrane-proximal and displays the corresponding scFV-proteinas a translational target binders is performed by flow cytometry on CHO cells fusion to phage coat protein III. This pool of phages display expressing the human PSMA antigen as described in ing the antibody library is later used for the selection of Example 2.1, the rat PSMA antigen as described in Example antigen binding entities. 60 2.3 and the mutated rat PSMA antigen with human mem 2.8 Phage Display Based Selection of Membrane-proximal brane-distal epitopes as described in Example 2.5. For flow Target Binders on CHO Cells Expressing the Mutated Human cytometry 2.5x10 cells of the respective cell lines are incu Psma Antigen with Rat Membrane-distal Epitopes bated with 50 ul supernatant. The binding of the constructs is The phage library carrying the cloned schv-repertoire is detected with an anti-His antibody (Penta-His-Antibody, BSA harvested from the respective culture supernatant by 65 free, Qiagen GmbH, Hilden, FRG) at 2 ug/ml in 50 lul PBS PEG8000/NaCl precipitation and centrifugation. Approxi with 2% FCS. As a second step reagent a R-Phycoerythrin mately 10' to 10" scFv phage particles are resuspended in conjugated affinity purified F(ab')2 fragment, goat anti US 9,260,522 B2 61 62 mouse IgG (Fc-gamma fragment specific), diluted 1:100 in screened, identified and confirmed as described above for the 50 ul PBS with 2% FCS (Dianova, Hamburg, FRG) is used. parental non-human (murine) anti-PSMA schv. Single clones The samples are measured on a FACSscan (BD biosciences, are then analyzed for favorable properties and amino acid Heidelberg, FRG). sequence. Those ScFVs, which are closest in amino acid Only constructs which show binding to CHO cells express sequence homology to human germline V-segments, are pre ing the human PSMA antigen and do not show binding to ferred. CHO cells expressing the rat PSMA antigen and also do not Human/humanized anti-PSMA schvs to membrane-proxi show binding to CHO cells expressing the mutated rat PSMA mal target epitopes of human PSMA are converted into antigen with human membrane-distal epitopes are selected recombinant bispecific single chain antibodies and further for further use. 10 characterized as follows. 2.10 Generation of Human/Humanized Equivalents of Non 2.11 Generation of I2C-based Bispecific Single Chain Anti human schvs to Membrane-proximal Target Epitopes of bodies Directed at Membrane-proximal Target Epitopes of Human PSMA Human PSMA The VH region of a murine anti-PSMA schv to a mem Anti-PSMA schvs to membrane-proximal target epitopes brane-proximal target epitope of human PSMA is aligned 15 of human PSMA with favorable properties and amino acid against human antibody germline amino acid sequences. The sequence are converted into recombinant bispecific single human antibody germline VH sequence is chosen which has chain antibodies by joining them via a Gly-Ser-linker with the closest homology to the non-human VH and a direct the CD3 specific scFv 120 (SEQ ID NO: 185) to result in alignment of the two amino acid sequences is performed. constructs with the domain arrangement VHs-(Gly There are a number of framework residues of the non-human Ser)-VLS-Ser Gly-Ser-VH-(Gly Ser)--Vls. VH that differ from the human VH framework regions (“dif Alternatively further constructs with different domain ferent framework positions'). Some of these residues may arrangements can be generated according to standard proto contribute to the binding and activity of the antibody to its colls. For expression in CHO cells the coding sequences of (i) target. an N-terminal immunoglobulin heavy chain leader compris To construct a library that contains the murine CDRs and at 25 ing a start codon embedded within a Kozak consensus every framework position that differs from the chosen human sequence and (ii) a C-terminal His6-tag followed by a stop VH sequence both possible residues (the human and the codon are both attached in frame to the nucleotide sequence maternal murine amino acid residue), degenerated oligo encoding the bispecific single chain antibodies prior to inser nucleotides are synthesized. These oligonucleotides incorpo tion of the resulting DNA-fragment as obtained by gene Syn rate at the differing positions the human residue with a prob 30 thesis into the multiple cloning site of the expression vector ability of 75% and the murine residue with a probability of pEF-DHFR (Raum et al. Cancer Immunol Immunother 50 25%. For one human VH e.g. six of these oligonucleotides (2001) 141-150). A clone with sequence-verified nucleotide have to be synthesized that overlap in a terminal stretch of sequence is transfected into DHFR deficient CHO cells for approximately 20 nucleotides. To this end every second eukaryotic expression of the construct. Eukaryotic protein primer is an antisense primer. Restriction sites within the 35 expression in DHFR deficient CHO cells is performed as oligonucleotides needed for later cloning are deleted. described by Kaufmann R.J. (1990) Methods Enzymol. 185, These primers may have a length of 60 to 90 nucleotides, 537-566. Gene amplification of the construct is induced by depending on the number of primers that are needed to span increasing concentrations of methotrexate (MTX) to a final over the whole V sequence. concentration of up to 20 nM MTX. These e.g. six primers are mixed in equal amounts (e.g. 1 ul 40 2.12 Expression and Purification of Bispecific Single Chain of each primer (primer stocks 20 to 100 uM) to a 20 lul PCR Antibody Molecules Directed at Membrane-proximal Target reaction) and added to a PCR mix consisting of PCR buffer, Epitopes of Human PSMA nucleotides and Taq polymerase. This mix is incubated at 94° Bispecific single chain antibody molecules are expressed C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° in Chinese hamster ovary cells (CHO). Eukaryotic protein C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° 45 expression in DHFR deficient CHO cells is performed as C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. described by Kaufmann R.J. (1990) Methods Enzymol. 185, Subsequently the product is run in an agarose gel electro 537-566. Gene amplification of the constructs is induced by phoresis and the product of a size from 200 to 400 base pairs addition of increasing concentrations of MTX up to final isolated from the gel according to standard methods. concentrations of 20 nMMTX. After two passages of station This PCR product is then used as a template for a standard 50 ary culture cell culture Supernatantis collected and used in the PCR reaction using primers that incorporate suitable N-ter Subsequent experiments. To generate Supernatant for purifi minal and C-terminal cloning restriction sites. The DNA frag cation after two passages of Stationary culture the cells are ment of the correct size (for a VH approximately 350 nucle grown in roller bottles with nucleoside-free HyO PF CHO otides) is isolated by agarose gel electrophoresis according to liquid soy medium (with 4.0 mM L-Glutamine with 0.1% standard methods. In this way sufficientVHDNA fragment is 55 Pluronic F-68; HyClone) for 7 days before harvest. The cells amplified. This VH fragment is now a pool of VH fragments are removed by centrifugation and the Supernatant containing that have each one a different amount of human and murine the expressed protein is stored at -20°C. Alternatively, con residues at the respective differing framework positions (pool structs are transiently expressed in HEK 293 cells. Transfec of humanized VH). The same procedure is performed for the tion is performed with 293fectin reagent (Invitrogen, #12347 VL region of the murine anti-PSMA scFv to a membrane 60 019) according to the manufacturer's protocol. Furthermore proximal target epitope of human PSMA (pool of humanized the constructs are alternatively expressed in transiently trans VL). fected DHFR deficient CHO cells using for example The pool of humanized VH is then combined with the pool FuGENER HD Transfection Reagent (Roche Diagnostics of humanized VL in the phage display vectorpComb3H5Bhis GmbH, Cat. No. 04709691001) according to the manufactur to form a library of functional scFVs from which—after dis 65 er's protocol. play on filamentous phage—anti-PSMA binders to mem Akta R. Explorer System (GE Health Systems) and Uni brane-proximal target epitopes of human PSMA are selected, corn(R) Software are used for chromatography. Immobilized US 9,260,522 B2 63 64 metal affinity chromatography (“IMAC) is performed using with an Fc gamma-specific antibody (Dianova) conjugated to a Fractogel EMD ChelateR (Merck) which is loaded with phycoerythrin, diluted 1:100 in PBS with 2% FCS. Superna ZnCl2 according to the protocol provided by the manufac tant of untransfected cells is used as a negative control. turer. The column is equilibrated with buffer A (20 mM Flow cytometry is performed on a FACS-Calibur appara sodium phosphate buffer pH 7.2, 0.1 M NaCl) and the cell tus, the CellOuest Software is used to acquire and analyze the culture supernatant (500 ml) is applied to the column (10 ml) data (Becton Dickinson biosciences, Heidelberg). FACS at a flow rate of 3 ml/min. The column is washed with buffer staining and measuring of the fluorescence intensity are per A to remove unbound sample. Bound protein is eluted using formed as described in Current Protocols in Immunology a two step gradient of buffer B (20 mM sodium phosphate (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley buffer pH 7.2, 0.1 MNaCl, 0.5 MImidazole) according to the 10 Interscience, 2002). following procedure: Only those constructs that show bispecific binding to Step 1: 20% buffer B in 6 column volumes human CD3 as well as to human PSMA and neither bind to Step 2: 100% buffer B in 6 column volumes the rat PSMA antigen nor to the mutated rat PSMA antigen Eluted protein fractions from step 2 are pooled for further with human membrane-distal epitopes are selected for further purification. All chemicals are of research grade and pur 15 SC. chased from Sigma (Deisenhofen) or Merck (Darmstadt). 2.14 Bioactivity of Bispecific Antibodies Directed at Mem Gel filtration chromatography is performed on a HiLoad brane-proximal Target Epitopes of Human PSMA 16/60 Superdex 200 prep grade column (GE/Amersham) Bioactivity of generated bispecific single chain antibodies equilibrated with Equi-buffer (25 mM Citrate, 200 mM is analyzed by chromium 51 (Cr) release in vitro cytotox Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow icity assays using the CHO cells transfected with human rate 1 ml/min) are subjected to standard SDS-PAGE and PSMA described in Example 2.1. As effector cells stimulated Western Blot for detection. Prior to purification, the columnis human CD4/CD56 depleted PBMC are used. calibrated for molecular weight determination (molecular Stimulated Human PBMC are obtained as follows: weight marker kit, Sigma MW GF-200). Protein concentra A Petri dish (145 mm diameter, Greiner bio-one GmbH, tions are determined using OD280 nm. 25 Kremsmünster) is coated with a commercially available anti Purified bispecific single chain antibody protein is ana CD3 specific antibody (e.g. OKT3. Orthoclone) in a final lyzed in SDS PAGE under reducing conditions performed concentration of 1 lug/ml for 1 hour at 37°C. Unbound protein with pre-cast 4-12% Bis Tris gels (Invitrogen). Sample prepa is removed by one washing step with PBS. The fresh PBMC ration and application are performed according to the proto are isolated from peripheral blood (30-50 ml human blood) col provided by the manufacturer. The molecular weight is 30 by Ficoll gradient centrifugation according to standard pro determined with MultiMark protein standard (Invitrogen). tocols. 3-5x107 PBMC are added to the precoated petridishin The gel is stained with colloidal Coomassie (Invitrogen pro 120 ml of RPMI 1640 with stabilized glutamine/10% FCS/ tocol). The purity of the isolated protein is typically >95% as IL-2 20 U/ml (Proleukin, Chiron) and stimulated for 2 days. determined by SDS-PAGE. On the third day the cells are collected and washed once with The bispecific single chain antibody has a molecular 35 RPMI 1640. IL2 is added to a final concentration of 20U/ml weight of about 52 kDa under native conditions as determined and the cells are cultivated again for one day in the same cell by gel filtration in PBS. culture medium as above. Western Blot is performed using an Optitran?R) BA-S83 By depletion of CD4+ T cells and CD56+ NK cells accord membrane and the Invitrogen Blot Module according to the ing to standard protocols CD8+ cytotoxic T lymphocytes protocol provided by the manufacturer. The antibody used is 40 (CTLs) are enriched. directed against the His Tag (Penta His, Qiagen) and a Goat Target cells are washed twice with PBS and labeled with anti-mouse Ig labeled with alkaline phosphatase (AP) 11.1 MBq 'Cr in a final volume of 100 ul RPMI with 50% (Sigma) is used as second step reagent, and BCIP/NBT FCS for 60 minutes at 37° C. Subsequently the labeled target (Sigma) as substrate. Aband detected at 52 kD corresponds to cells are washed 3 times with 5 ml RPMI and then used in the purified bispecific single chain antibodies. 45 cytotoxicity assay. The assay is performed in a 96 well plate 2.13 Flow Cytometric Binding Analysis of Bispecific Anti in a total volume of 250 ul supplemented RPMI (as above) bodies Directed at Membrane-proximal Target Epitopes of with an E:T ratio of 10:1. 1 ug/ml of purified bispecific single Human PSMA chain antibody molecule and 20 threefold dilutions thereof In order to test the functionality of bispecific antibody are applied. The assay time is 18 hours. Cytotoxicity is mea constructs regarding the capability to bind to CD3 and to 50 Sured as relative values of released chromium in the Superna membrane-proximal target epitopes of human PSMA, tant related to the difference of maximum lysis (addition of respectively, a FACS analysis is performed. For this purpose Triton-X) and spontaneous lysis (without effector cells). All CHO cells transfected with human PSMA as described in measurements are done in quadruplicates. Measurement of Example 2.1 and the human CD3 positive T cell leukemia cell chromium activity in the Supernatants is performed with a line HPB-ALL (DSMZ. Braunschweig, ACC483) are used. 55 Wizard R. 3" gammacounter (Perkin Elmer Life Sciences For confirmation of binding to membrane-proximal target GmbH, Koln, Germany). Analysis of the experimental data is epitopes of human PSMA in addition CHO cells express performed with Prism 4 for Windows(R (version 4.02, Graph ing the rat PSMA antigen as described in Example 2.3 and Pad Software Inc., San Diego, Calif., USA). Sigmoidal dose CHO cells expressing the mutated rat PSMA antigen with response curves typically have R values >0.90 as determined human membrane-distal epitopes as described in Example 60 by the software. EC50 values calculated by the analysis pro 2.5 are used. 200,000 cells of the respective cell lines are gram are used for comparison of bioactivity. incubated for 30 min on ice with 50 ul of cell culture super Only those constructs showing potent recruitment of cyto natant of transfected cells expressing the bispecific antibody toxic activity of effector T cells against target cells positive constructs. The cells are washed twice in PBS with 2% FCS for PSMA are selected for further use. and binding of the construct is detected with a murine Penta 65 2.15 Generation of CHO Cells Expressing Macaque PSMA His antibody (Qiagen; diluted 1:20 in 50 ul PBS with 2% The cDNA sequence of macaque PSMA is obtained by a FCS). After washing, bound anti His antibodies are detected set of five PCRs on cDNA from macaque monkey prostate US 9,260,522 B2 65 66 prepared according to standard protocols. The following gated by FACS analysis. For this purpose the macaque PSMA reaction conditions: 1 cycle at 94° C. for 2 minutes followed transfected CHO cells as described in example 2.15 and the by 40 cycles with 94°C. for 1 minute, 52°C. for 1 minute and macaque T cell line 4119LnPx (kindly provided by Prof 72° C. for 1.5 minutes followed by a terminal cycle of 72°C. Fickenscher, Hygiene Institute, Virology, Erlangen-Nuern for 3 minutes and the following primers are used: 5 berg; published in Knappe A, et al., and Fickenscher H., Blood 2000, 95, 3256-61) are used. 200,000 cells of the respective cell lines are incubated for 30 minonice with 50 ul 1. forward primer: 5' - cactgtggcc.caggttcgagg-3' SEO ID NO: 228 of cell culture Supernatant of transfected cells expressing the reverse primer: cross-species specific bispecific antibody constructs. The 5'-gacataccacacaaattcaatacgg-3' SEO ID NO: 229 10 cells are washed twice in PBS with 2% FCS and binding of 2. forward primer: the construct is detected with a murine Penta His antibody 5'-gctctgct cqc.gc.cgagatgtgg-3 SEO ID NO: 230 (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After reverse primer: washing, bound anti His antibodies are detected with an Fc 5'-acgctggacaccacctic cagg-3' SEO ID NO: 231 15 gamma-specific antibody (Dianova) conjugated to phyco 3. forward primer: erythrin, diluted 1:100 in PBS with 2% FCS. Supernatant of 5'-ggttctactgagtgggcagagg-3 SEO ID NO: 232 untransfected cells is used as a negative control. Flow cytom reverse primer: etry is performed on a FACS-Calibur apparatus, the 5'-acttgttgttggctgcttggagc-3' SEO ID NO: 233 CellOuest software is used to acquire and analyze the data 4. forward primer: (Becton Dickinson biosciences, Heidelberg). FACS staining 5'-gggtgaagt cctato cagatgg-3' SEO ID NO: 234 and measuring of the fluorescence intensity are performed as reverse primer: described in Current Protocols in Immunology (Coligan, 5'-gtgctctgcctgaagcaatticc-3' SEO ID NO: 235 Kruisbeek, Margulies. Shevach and Strober, Wiley-Inter 5. forward primer: science, 2002). 5' - ct cqgct tcct citt.cgggtgg-3' SEO ID NO: 236 25 reverse primer: Example 3 5'-gcatatt catttgctggg talacct gg-3 SEQ ID NO: 237 Those PCRs generate five overlapping fragments, which 3.1. Generation of PSMA- and CD3-directed Bispecific are isolated and sequenced according to standard protocols Single Antibodies using the PCR primers, and thereby provided a portion of the 30 Bispecific single chain antibodies comprising either Sclv cDNA sequence coding macaque PSMA from codon3 to the binding domain P1, P2, P3, P4 or P5 against a PSMA-epitope last codon of the mature protein. To generate a construct for of <60° membrane-distance or scFv binding domain D1 or expression of macaque PSMA a cDNA fragment is obtained D2 against a PSMA-epitope of >60 membrane-distance and by gene synthesis according to standard protocols (the cDNA the schv binding domain I2C directed at CD3epsilon on and amino acid sequence of the construct is listed under SEQ 35 human T cells were obained by gene synthesis. The gene ID NO. 238 and 239). In this construct the coding sequence of synthesis fragments were designed as to contain first a Kozak macaque PSMA from amino acid 3 to the last amino acid of site for eukaryotic expression of the construct, followed by a the mature PSMA protein followed by a stop codon is fused in 19 amino acid immunoglobulin leader peptide, followed in frame to the coding sequence of the first two amino acids of frame by the coding sequence of the bispecific single chain the human PSMA protein. The gene synthesis fragment is 40 antibody, followed in frame by the coding sequence of a 6 also designed as to containa Kozak site for eukaryotic expres histidine tag and a stop codon. The variable region arrange sion of the construct and restriction sites at the beginning and ments as well as the SEQ ID Nos of the cDNA- and amino the end of the fragment containing the cDNA. The introduced acid sequences are listed in the table 3 below. restriction sites, Xbal at the 5' end and Sal at the 3' end, are utilised in the following cloning procedures. The gene Syn 45 TABLE 3 thesis fragment is cloned via Xbal and SalI into a plasmid SEQID Formats of protein constructs designated pEF-DHFR (pEF-DHFR is described in Raum et (nucliprot) (N-> C) al. Cancer Immunol Immunother 50 (2001) 141-150) follow ing standard protocols. The aforementioned procedures are 281,280 PSMA-P1 LHXI2CHL 295,294 PSMA-P2 LHXI2CHL 50 carried out according to standard protocols (Sambrook, 309.308 PSMA-P3 LH x I2CHL Molecular Cloning: A Laboratory Manual, 3rd edition, Cold 323,322 PSMA-P4. LH x I2CHL Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. 337,336 PSMA-PSLEH x I2CHL (2001)). A clone with sequence-verified nucleotide sequence 351,350 PSMA-D1 LHXI2CHL is transfected into DHFR deficient CHO cells for eukaryotic 365,364 PSMA-D2 LHXI2CHL expression of the construct. Eukaryotic protein expression in 55 DHFR deficient CHO cells is performed as described by The gene synthesis fragments were also designed as to Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. introduce suitable restriction sites at the beginning (EcoRI) Gene amplification of the construct is induced by increasing and at the end of the fragment (Sal I) for cloning of the gene concentrations of methotrexate (MTX) to a final concentra synthesis fragment into the mammalian cell expression vec tion of up to 20 nM MTX. 60 torpEF-DHFR (pEF-DHFR is described in Raum et al. Can 2.16 Flow Cytometric Analysis of Cross-species Specificity cer Immunol Immunother 50 (2001) 141-150). The afore of Bispecific Antibodies Directed at Membrane-proximal mentioned procedures were carried out according to standard Target Epitopes of Human PSMA protocols (Sambrook, Molecular Cloning: A Laboratory In order to test the cross-species specificity of bispecific Manual, 3rd edition, Cold Spring Harbour Laboratory Press, antibodies directed at membrane-proximal target epitopes of 65 Cold Spring Harbour, N.Y. (2001)). A clone with sequence human PSMA the capability of the constructs to bind to verified nucleotide sequence was transfected into DHFR defi macaque PSMA and macaque CD3, respectively, is investi cient CHO cells for eukaryotic expression of the construct. US 9,260,522 B2 67 68 Eukaryotic protein expression in DHFR deficient CHO cells ul RPMI with 50% FCS for 60 minutes at 37° C. Subse was performed as described by Kaufmann R. J. (1990) Meth quently the labelled target cells were washed 3 times with 5 ods Enzymol. 185, 537-566. Gene amplification of the con ml RPMI and then used in the cytotoxicity assay. The assay struct was induced by increasing concentrations of methotr was performed in a 96 well plate in a total volume of 250 ul exate (MTX) to a final concentration of up to 20 nM MTX. supplemented RPMI (as above) with an E:T cell ratio of 10:1. After two passages of stationary culture cell culture Superna I2C-based PSMA-directed bscAbs were added as culture tant was collected and used in the Subsequent experiments. supernatants from transfected CHO cells at different dilu 3.2. Membrane-distance <60 A or >60 A of PSMA-epitopes tions. The assay time was 18 hours. Cytotoxicity was mea Recognized by I2C-based PSMA-directed bscAbs sured as relative values of released chromium related to the Epitope confirmation of PSMA-directed bispecific single 10 difference of maximum lysis (addition of Triton-X) and spon antibodies was carried out by flowcytometry on CHO cells taneous lysis (without effector cells). All measurements were transfected with (unmutated) human PSMA, unmutated rat carried out in quadruplicates. Measurement of released chro PSMA and rat PSMA mutated to the homologous human mium activity was performed with a Wizard R. 3" gamma amino acid at every mismatched amino acid position with a counter (Perkin Elmer Life Sciences GmbH, Koln, Ger membrane-distance of >60 A as described in Example 2. 15 200,000 cells of each CHO-transfectant were incubated for many). Analysis of the experimental data was performed with 30 min on ice with 50 ul of cell culture supernatant of trans Prism 4 for Windows(R (version 4.02, GraphPad Software fected cells expressing the PSMA-directed bispecific anti Inc., San Diego, Calif., USA). Sigmoidal dose response body constructs. The cells were washed twice in PBS with 2% curves typically had R values >0.90 as determined by the FCS and binding of the construct was detected with a murine software. Penta His antibody (Qiagen; diluted 1:20 in 50 ul PBS with As shown in FIG. 5 all 5 PSMA-directed bscAbs, whose 2% FCS). After washing, bound anti H is antibodies were PSMA-epitopes have a membrane-distance of <60 A are sub detected with an Fc gamma-specific antibody (Dianova) con stantially more potent in redirecting T cell cytotoxicity than jugated to phycoerythrin, diluted 1:100 in PBS with 2% FCS. the other two PSMA-directed bscAbs, whose PSMA Cell culture medium was used as a negative control. Flowcy 25 epitopes have a membrane-distance of >60 A. tometry was performed on a FACS-Calibur apparatus, the 4. Generation of Additional Bispecific Single CellOuest Software was used to acquire and analyze the data Antibodies Directed at CD3 and (Becton Dickinson biosciences, Heidelberg). FACS staining Membrane-proximal Target Epitopes of Human and measuring of the fluorescence intensity were performed Psma as described in Current Protocols in Immunology (Coligan, 30 Kruisbeek, Margulies. Shevach and Strober, Wiley-Inter The human antibody germline VH sequence VH1 1-03 science, 2002). (http://vbase.mrc-cpe.cam.ac.uk/) is chosen as framework FIG. 4 shows, that PSMA-directed bscAbs P1XI2C, context for CDRH1 (SEQID NO 260), CDRH2 (SEQID NO P2xI2C, P3XI2C, P4xI2C and P5xI2C bind to (unmutated) 261) and CDRH3 (SEQ ID NO 262). For VH1 1-03 the human PSMA but neither to unmutated nor to mutated rat 35 following degenerated oligonucleotides have to be synthe PSMA and thus confirms a membrane-distance of <60 A for sized that overlap in a terminal stretch of approximately the PSMA-epitope of each of these bispecific constructs. By 15-20 nucleotides (to this end every second primer is an contrast, PSMA-directed bscAbs DlxI2C and D2xI2C do not antisense primer): bind to unmutated rat PSMA but to (unmutated) human and mutated rat PSMA consistent with PSMA-epitopes of a mem 40 brane-distance 60 A. 3.3. Relative T Cell Cytotoxicity Redirected by I2C-based (SEQ ID NO: 449) PSMA-directed bscAbs with PSMA-epitopes of a Mem CTT GAT CTC GAG TCC SCT GAG STG RWG AAG CCT GGC GCC. TCC GTG AAG TCC TGC AAG GCC. TCC GGC brane-distance <60 A and >60 A TAC T cell cytotoxicity redirected by I2C-based PSMA-di 45 rected bscAbs against CHO cells transfected with human (SEQ ID NO: 45O) PSMA was measured in a chromium 51 (Cr) release assay. CCA TTC CAG CMS CTG GCC GGG TKY CTG TYT CAC CCA As source of effector T cells stimulated human CD4/CD56 GTG CAT CAC GTA GCC GGT GAA GGT GTA GCC GGA. GGC depleted PBMC were used. Stimulated human PBMC were CTT GCA obtained as follows: A Petri dish (145 mm diameter, Greiner 50 bio-one GmbH, Kremsmunster) was coated with a commer (SEQ ID NO: 451) cially available anti-CD3 specific antibody (e.g. OKT3. CCC GGC CAG SKG GAA ATS GGC TAC ATC AAC Orthoclone) in a final concentration of 1 lug/ml for 1 hour at CCT TAC AAC GAC ACC TAC AAC GGC AAG TTC 37°C. Unbound protein was removed by one washing step AAG with PBS. The fresh PBMC were isolated from peripheral 55 blood (30-50 ml human blood) by Ficoll gradient centrifuga (SEQ ID NO: 452) tion according to standard protocols. 3-5x107 PBMC were TTC CAT GTA GGC GGT GGA GGM GKA CKT GTC KCT GGT added to the precoated petridish in 120 ml of RPMI 1640 with AA.K. GGT GRC TYT GCC CTT GAA CTT GCC GTT GTA stabilized glutamine/10% FCS/IL-220U/ml (Proleukin, Chi ron) and stimulated for 2 days. On the third day the cells were 60 (SEQ ID NO: 453 ) collected and washed once with RPMI 1640. IL-2 was added TCC ACC GCC TAC ATG GAA CTG TCC CTG ASG TCT to a final concentration of 20U/ml and the cells were cultured GAG GAC ACC GCC GTG TAC TAC TGC AGG GGC again for one day in the same cell culture medium as above. By depletion of CD4+ T cells and CD56+ NK cells accord (SEQ ID NO: 454) ing to standard protocols CD8+ cytotoxic T lymphocytes 65 CGA TAC GGT GAC CAG AGT GCC TCT CCA. GGA GTC (CTLs) were enriched. Target cells were washed twice with GAA GTA GTA. CCA GTT CTC GCC CCT GCA GTA GTA PBS and labelled with 11.1 MBq'Crina final volume of 100 US 9,260,522 B2 69 70 This primer-set spans over the whole VH sequence. The VL PCR product is then used as a template for a Within this set primers are mixed in equal amounts (e.g. 1 standard PCR reaction using primers that incorporate N-ter ul of each primer (primer stocks 20 to 100 uM) to a 20 ul PCR minal and C-terminal Suitable cloning restriction sites. The reaction) and added to a PCR mix consisting of PCR buffer, DNA fragment of the correct size (for a VL approximately nucleotides and Taq polymerase. This mix is incubated at 94° 330 nucleotides) is isolated by agarose gel electrophoresis C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° according to standard methods. In this way Sufficient VL C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° DNA fragment is amplified. C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. The final VH1 1-03-based VHPCR product (i.e. the rep Subsequently the product is run in an agarose gel electro ertoire of human/humanized VH) is combined with the final 10 VkII A1-based VL PCR product (i.e. the repertoire of human/ phoresis and the product of a size from 200 to 400 isolated humanized VL) in the phage display vector pComb3H5Bhis. from the gel according to standard methods. This VH-VL combination forms a library of functional sch vs The VH PCR product is then used as a template for a from which—after display on filamentous phage—anti standard PCR reaction using primers that incorporate N-ter PSMA binders are selected, screened, identified and con minal and C-terminal Suitable cloning restriction sites. The 15 firmed as described in the following: DNA fragment of the correct size (for a VHapproximately 450 ng of the light chain fragments (SacI-Spel digested) 350 nucleotides) is isolated by agarose gel electrophoresis are ligated with 1400 ng of the phagemid pComb3H5Bhis according to standard methods. In this way sufficient VH (SacI-Spel digested; large fragment). The resulting combina DNA fragment is amplified. torial antibody library is then transformed into 300 ul of The human antibody germline VL sequence Vkll A1 electrocompetent Escherichia coli XL1 Blue cells by elec (http://vbase.mrc-cpe.cam.ac.uk/) is chosen as framework troporation (2.5 kV, 0.2 cm gap cuvette, 25 uFD, 200 Ohm, context for CDRL1 (SEQ ID NO: 255), CDRL2 (SEQ ID Biorad gene-pulser) resulting in a library size of more than NO: 256) and CDRL3 (SEQID NO: 257). For Vkll A1 the 107 independent clones. After one hour of phenotype expres following degenerated oligonucleotides have to be synthe Sion, positive transformants are selected for carbenicilline sized that overlap in a terminal stretch of approximately 25 resistance encoded by the pComb3H5BHis vector in 100 ml 15-20 nucleotides (to this end every second primer is an of liquid super broth (SB)-culture over night. Cells are then antisense primer): harvested by centrifugation and plasmid preparation is car ried out using a commercially available plasmid preparation kit (Qiagen). 5 'P6-WL-A-SacI 30 (SEQ ID NO: 455) 2800 ng of this plasmid-DNA containing the VL-library CTT GAT GAG CTC ATG ACC CAG TCT CCA SYC. TCC (XhoI-BstEII digested; large fragment) are ligated with 900 CTG SCT, GTG ACT GGC CAG CSG GCC. TCC ATC. TCT ng of the heavy chain V-fragments (XhoI-BstEII digested) TGC CGG and again transformed into two 300ul aliquots of electrocom petent E. coli XL 1 Blue cells by electroporation (2.5 kV, 0.2 (SEQ ID NO: 456) 35 cm gap cuvette, 25uFD,200 Ohm) resulting in a total VH-VL CCA GTG CAT GAA GGT GTT GTC GTA GGA GTC GAT G.C.A. schv (single chain variable fragment) library size of more CTC GGA. GGC CCG GCA AGA GAT GGA GGC than 10 independent clones. After phenotype expression and slow adaptation to carbe (SEQ ID NO: 457) nicilline, the E. coli cells containing the antibody library are ACC TTC ATG CAC TWT CAG CAG ARG CCT GGC CAG 40 transferred into SB-carbenicilline (SB with 50 ug/mL carbe YCT CCT MRC CKG ATC TWC CGG GCC TOT ATC CTG nicilline) selection medium. The E. coli cells containing the GAA antibody library is then infected with an infectious dose of 10' particles of helper phage VCSM13 resulting in the pro (SEQ ID NO: 458) duction and secretion of filamentous M13 phage, wherein CAG GGT GAA GTC GGT GCC GGA GCC AGA GCC GGA GAA 45 phage particle contains single stranded pComb3H5BHis CCG GKC AGG GAY GCC GGA TTC CAG GAT AGA. GGC CCG DNA encoding a scFV-fragment and displayed the corre sponding scFV-protein as a translational fusion to phage coat (SEQ ID NO: 459) protein III. This pool of phages displaying the antibody ACC GAC TTC ACC AMA ATC TMC CST GTG GAG GCC library is used for the selection of antigen binding entities. GAS GAC GTG GSC TAC TAC TGC CAC CAG 50 For this purpose the phage library carrying the cloned schv-repertoire is harvested from the respective culture (SEQ ID NO: 460) supernatant by PEG8000/NaCl precipitation and centrifuga ACT CAG ACT AGT CGT ACG CTT GAT TTC CAG CTT GGT tion. Approximately 10' to 10" scFv phage particles are CCC. TCC GCC GAA. GGT GTA AGG GTC CTC GAT GGA CTG GTG GCA GTA GTA resuspended in 0.4 ml of PBS/0.1% BSA and incubated with 55 10 to 107 PSMA-positive human prostate cancer cell line This primer-set spans over the whole corresponding VL LNCaP (ATCC No. CRL-1740) for 1 hour on ice under slow Sequence. agitation. These LNCaP cells are harvested beforehand by Within this set primers are mixed in equal amounts (e.g. 1 centrifugation, washed in PBS and resuspended in PBS/1% ul of each primer (primer stocks 20 to 100 uM) to a 20 ul PCR FCS (containing 0.05% Na Azide). Schv phage which do not reaction) and added to a PCR mix consisting of PCR buffer, 60 specifically bind to LNCaP cells are eliminated by up to five nucleotides and Taq polymerase. This mix is incubated at 94° washing steps with PBS/1% FCS (containing 0.05% Na C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° Azide). After washing, binding entities are eluted from the C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° cells by resuspending the cells in HCl-glycine pH 2.2 (10 min C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. incubation with Subsequent Vortexing) and after neutraliza Subsequently the product is run in an agarose gel electro 65 tion with 2M Tris pH 12, the eluate is used for infection of a phoresis and the product of a size from 200 to 400 isolated fresh uninfected E. coli XL1 Blue culture (OD600>0.5). The from the gel according to standard methods. E. coli culture containing E.coli cells successfully transduced US 9,260,522 B2 71 72 with a phagemid copy, encoding a human/humanized scEv sequence encoding the bispecific single chain antibodies fragment, are again selected for carbenicilline resistance and prior to insertion of the resulting DNA-fragment as obtained subsequently infected with VCMS 13 helperphage to start the by gene synthesis into the multiple cloning site of the expres second round of antibody display and in vitro selection. A sion vectorpFF-DHFR (Raum et al. Cancer Immunol Immu total of 4 to 5 rounds of selections are carried out, normally. nother 50 (2001) 141-150). Transfection of the generated In order to screen for PSMA specific binders plasmid DNA expression plasmids is carried out as described in Example corresponding to 4 and 5 rounds of panning is isolated from E. 3.1. Protein expression and purification of bispecificantibody coli cultures after selection. For the production of soluble constructs, flow cytometric confirmation of binding to CD3 schv-protein, VH-VL-DNA fragments are excised from the and to membrane-proximal target epitopes of human PSMA plasmids (Xho1-Spel). These fragments are cloned via the 10 same restriction sites into the plasmid pComb3H5BFlag? His as well as the analysis of bioactivity by cytotoxicity assay are differing from the original pComb3H5BHis in that the performed as described in Example 2. All other state of the art expression construct (i.e. the ScFV) includes a Flag-tag procedures are carried out according to standard protocols (DYKDDDDK) at its C-terminus before the His6-tag and that (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd phage protein III/N2 domain and protein III/CT domain had 15 edition, Cold Spring Harbour Laboratory Press, Cold Spring been deleted. After ligation, each pool (different rounds of Harbour, N.Y. (2001)). panning) of plasmid DNA is transformed into 100 ul heat Only those bispecific antibody constructs that bind to CD3 shock competent E. coli TG1 or XLI blue and plated onto and to membrane-proximal target epitopes of human PSMA carbenicilline LB-agar. Single colonies are picked into 100 ul and show potent recruitment of cytotoxic activity of effector of LB carb (50 ug/ml carbenicilline). T cells against target cells positive for PSMA are selected for E. coli transformed with pComb3H5BFlag? His containing further use. a VL- and VH-segment produce soluble schv in sufficient amounts after induction with 1 mM IPTG. Due to a suitable Example 4 signal sequence, the Sclv-chain is exported into the peri plasma where it folds into a functional conformation. 25 4.1. Generation of PSMA- and CD3-directed Bispecific Single E. coli TG1 bacterial colonies from the transforma Single Chain Antibodies tion plates are picked for periplasmic Small scale preparations Bispecific single chain antibodies comprising either Sclv and grown in SB-medium (e.g. 10 ml) supplemented with 20 binding domain P6 against a PSMA-epitope of <60 A mem mM MgCl, and carbenicilline 50 ug/ml (and re-dissolved in brane-distance or schv binding domain D3 against a PSMA PBS (e.g. 1 ml) after harvesting. By four rounds of freezing at 30 epitope of >60 A membrane-distance and the sclv binding -70° C. and thawing at 37° C., the outer membrane of the domain I2C directed at CD3epsilon on human T cells were bacteria is destroyed by temperature shock and the soluble obained by gene synthesis. The gene synthesis fragments periplasmic proteins including the ScFVs are released into the were designed as to contain first a Kozak site for eukaryotic supernatant. After elimination of intact cells and cell-debris expression of the construct, followed by a 19 amino acid by centrifugation, the Supernatant containing the anti-PSMA 35 immunoglobulin leader peptide, followed in frame by the scFVs is collected and used for the identification of PSMA coding sequence of the bispecific single chain antibody, fol specific binders as follows: lowed inframe by the coding sequence of a 6 histidine tag and Binding of schvs to PSMA is tested by flow cytometry on a stop codon. The variable region arrangements as well as the the PSMA-positive human prostate cancer cell line LNCaP SEQ ID Nos of the cDNA- and amino acid sequences are (ATCC No. CRL-1740). A periplasmic small scale prepara 40 listed in the table 4 below. tion as described above without any grown bacteria is used as negative control. TABLE 4 For flow cytometry 2.5x10 cells are incubated with 50 ul SEQID Formats of protein constructs of scFv periplasmic preparation or with 5 lug/ml of purified (nucliprot) (N-> C) scEv in 50 ul PBS with 2% FCS. The binding of scFv is 45 detected with an anti-His antibody (Penta-His-Antibody, BSA 267.266 PSMA-P6 LHXI2CHL free, Qiagen GmbH, Hilden, FRG) at 2 ug/ml in 50 lul PBS 253,252 PSMA-D3 LH XI2CHL with 2% FCS. As a second step reagent a R-Phycoerythrin conjugated affinity purified F(ab')2 fragment, goat anti The gene synthesis fragments were also designed as to mouse IgG (Fc-gamma fragment specific), diluted 1:100 in 50 introduce suitable restriction sites at the beginning (EcoRI) 50 ul PBS with 2% FCS (Dianova, Hamburg, FRG) is used. and at the end of the fragment (Sal I) for cloning of the gene The samples are measured on a FACSscan (BD biosciences, synthesis fragment into the mammalian cell expression vec Heidelberg, FRG). torpEF-DHFR (pEF-DHFR is described in Raum et al. Can Single clones are then analyzed for favourable properties cer Immunol Immunother 50 (2001) 141-150). The afore and amino acid sequence. PSMA specific schvs are converted 55 mentioned procedures were carried out according to standard into recombinant bispecific single chain antibodies by joining protocols (Sambrook, Molecular Cloning: A Laboratory them via a Gly Ser-linker with the CD3 specific scFv I2C Manual, 3rd edition, Cold Spring Harbour Laboratory Press, (SEQ ID NO: 185) or any other CD3 specific sclv of the Cold Spring Harbour, N.Y. (2001)). A clone with sequence invention to result in constructs with the domain arrangement verified nucleotide sequence was transfected into DHFR defi VHes-(Gly-Seri)s-VLPsa-Ser Gly-Seri-VHos 60 cient CHO cells for eukaryotic expression of the construct. (Gly-Ser)-VL or alternative domain arrangements such Eukaryotic protein expression in DHFR deficient CHO cells aS VLPsalt-(Gly-Seri)s-VHest-Gly-Ser-VHos was performed as described by Kaufmann R. J. (1990) Meth (Gly-Ser)-VL. For expression in CHO cells the coding ods Enzymol. 185, 537-566. Gene amplification of the con sequences of (i) an N-terminal immunoglobulin heavy chain struct was induced by increasing concentrations of methotr leader comprising a start codon embedded within a Kozak 65 exate (MTX) to a final concentration of up to 20 nM MTX. consensus sequence and (ii) a C-terminal His-tag followed After two passages of stationary culture cell culture Superna by a stop codon are both attached in frame to the nucleotide tant was collected and used in the Subsequent experiments. US 9,260,522 B2 73 74 4.2. Membrane-distance <60 A or >60 A of PSMA-epitopes sured as relative values of released chromium related to the Recognized by I2C-based PSMA-directed bscAbs difference of maximum lysis (addition of Triton-X) and spon Epitope confirmation of PSMA-directed bispecific single taneous lysis (without effector cells). All measurements were antibodies was carried out by flowcytometry on CHO cells carried out in quadruplicates. Measurement of released chro transfected with (unmutated) human PSMA, unmutated rat mium activity was performed with a Wizard R. 3" gamma PSMA and rat PSMA mutated to the homologous human counter (Perkin Elmer Life Sciences GmbH, Koln, Ger amino acid at every mismatched amino acid position with a many). Analysis of the experimental data was performed with membrane-distance of >60 A as described in Example 2. Prism 4 for Windows(R (version 4.02, GraphPad Software 200,000 cells of each CHO-transfectant were incubated for Inc., San Diego, Calif., USA). Sigmoidal dose response 30 min on ice with 50 ul of cell culture supernatant of trans 10 curves typically had R values >0.90 as determined by the fected cells expressing the PSMA-directed bispecific anti Software. As shown in FIG. 7 the PSMA-directed bScAb body constructs. The cells were washed twice in PBS with 2% P6xI2C, whose PSMA-epitope has a membrane-distance of FCS and binding of the construct was detected with a murine <60A is substantially more potent in redirecting T cell cyto Penta His antibody (Qiagen; diluted 1:20 in 50 ul PBS with toxicity than PSMA-directed bscAb D3xI2C, whose PSMA 2% FCS). After washing, bound anti H is antibodies were 15 epitope has a membrane-distance of 60 A. detected with an Fc gamma-specific antibody (Dianova) con jugated to phycoerythrin, diluted 1:100 in PBS with 2% FCS. Example 5 Cell culture medium was used as a negative control. Flowcy tometry was performed on a FACS-Calibur apparatus, the Crossreactive Binding to Human and CellOuest Software was used to acquire and analyze the data Non-chimpanzee Primate PSMA and CD3 of (Becton Dickinson biosciences, Heidelberg). FACS staining I2C-based bscAbs Against Membrane-proximal and measuring of the fluorescence intensity were performed PSMA-Eptitopes as described in Current Protocols in Immunology (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley-Inter 5.1. Cloning and Expression of Cyno PSMA Antigen on CHO science, 2002). 25 Cells FIG. 6 shows, that PSMA-directed bscAb P6xI2C binds to The cDNA sequence of macaque PSMA was obtained as (unmutated) human PSMA but neither to unmutated nor to described in Example 2.15 As described above, these PCRs mutated rat PSMA and thus confirms a membrane-distance of generated five overlapping fragments, which were isolated <60 A for the PSMA-epitope of this bispecific construct. By and sequenced according to standard protocols using the PCR contrast, PSMA-directed bscAb D3xI2C does not bind to 30 primers, and thereby provided a portion of the cDNA unmutated rat PSMA but to (unmutated) human and mutated sequence coding macaque PSMA from codon 3 to the last rat PSMA consistent with a PSMA-epitope of a membrane codon of the mature protein. To generate a construct for distance >60 A. expression of macaque PSMA a cDNA fragment was 4.3. Relative T Cell Cytotoxicity Redirected by I2C-based obtained by gene synthesis according to standard protocols PSMA-directed bscabs with PSMA-epitopes of a Membrane 35 (the cDNA and amino acid sequence of the construct is listed distance <60 A and >60 A under SEQID Nos 238 and 239). In this construct the coding T cell cytotoxicity redirected by I2C-based PSMA-di sequence of macaque PSMA from amino acid 3 to the last rected bscAbs against CHO cells transfected with human amino acid of the mature PSMA protein followed by a stop PSMA was measured in a chromium 51 (Cr) release assay. codon was fused in frame to the coding sequence of the first As source of effector T cells stimulated human CD4/CD56 40 two amino acids of the human PSMA protein. The gene depleted PBMC were used. Stimulated human PBMC were synthesis fragment was also designed as to contain a Kozak obtained as follows: A Petri dish (145 mm diameter, Greiner site for eukaryotic expression of the construct and restriction bio-one GmbH, Kremsmunster) was coated with a commer sites at the beginning and the end of the fragment containing cially available anti-CD3 specific antibody (e.g. OKT3. the cDNA. The introduced restriction sites, Xbal at the 5' end Orthoclone) in a final concentration of 1 lug/ml for 1 hour at 45 and SalI at the 3' end, were utilised in the following cloning 37°C. Unbound protein was removed by one washing step procedures. The gene synthesis fragment was cloned via Xbal with PBS. The fresh PBMC were isolated from peripheral and SalI into a plasmid designated pBF-DHFR following blood (30-50 ml human blood) by Ficoll gradient centrifuga standard protocols. The aforementioned procedures were car tion according to standard protocols. 3-5x107 PBMC were ried out according to standard protocols (Sambrook, Molecu added to the precoated petridish in 120 ml of RPMI 1640 with 50 lar Cloning: A Laboratory Manual, 3rd edition, Cold Spring stabilized glutamine/10% FCS/IL-220U/ml (Proleukin, Chi Harbour Laboratory Press, Cold Spring Harbour, N.Y. ron) and stimulated for 2 days. On the third day the cells were (2001)). A clone with sequence-verified nucleotide sequence collected and washed once with RPMI 1640. IL-2 was added was transfected into DHFR deficient CHO cells for eukary to a final concentration of 20U/ml and the cells were cultured otic expression of the construct. Eukaryotic protein expres again for one day in the same cell culture medium as above. 55 sion in DHFR deficient CHO cells was performed as By depletion of CD4+ T cells and CD56+ NK cells accord described by Kaufmann R.J. (1990) Methods Enzymol. 185, ing to standard protocols CD8+ cytotoxic T lymphocytes 537-566. Gene amplification of the construct was induced by (CTLs) were enriched. Target cells were washed twice with increasing concentrations of methotrexate (MTX) to a final PBS and labelled with 11.1 MBq'Crina final volume of 100 concentration of up to 20 nM MTX. ul RPMI with 50% FCS for 60 minutes at 37° C. Subse 60 5.2. Flow Cytometric Binding Analysis of I2C-based Bscabs quently the labelled target cells were washed 3 times with 5 Against Membrane-proximal Eptitopes of Human PSMA on ml RPMI and then used in the cytotoxicity assay. The assay Non-chimanzee Primate PSMA and on Human and Non was performed in a 96 well plate in a total volume of 250 ul chimpanzee Primate CD3 supplemented RPMI (as above) with an E:T cell ratio of 10:1. Binding of bscAbs P1XI2C, P2xI2C, P3XI2C, P4xI2C, I2C-based PSMA-directed bscAbs were added as culture 65 P5xI2C and P6xI2C directed against membrane-proximal supernatants from transfected CHO cells at different dilu PSMA-epitopes to CHO cells expressing human PSMA is tions. The assay time was 18 hours. Cytotoxicity was mea shown by flowcytometry in Examples 3 and 4. US 9,260,522 B2 75 76 Binding of these bscAbs to macaque PSMA as well as to obtained by gene synthesis according to standard protocols human and macaque CD3 was analysed by flowcytometry (the cDNA and amino acid sequence of the construct is listed using CHO cells transfected with macaque PSMA, the human under SEQID Nos 368369). The gene synthesis fragment is CD3 positive T cell leukemia cell line HPB-ALL (DSMZ, designed as to contain first a Kozak site for eukaryotic expres Braunschweig, ACC483) and the CD3 positive macaque T sion of the construct followed by the coding sequence of the cell line 4119LnPx (kindly provided by Prof Fickenscher, murine Lag3 protein from amino acid 1 to 441 corresponding Hygiene Institute, Virology, Erlangen-Nuernberg; published to the signal peptide and extracellular domains of murine in Knappe A, et al., and Fickenscher H. Blood 2000, 95, Lag3, followed in frame by the coding sequence of an artifi 3256-61). Results are shown in FIG.8. 200,000 cells of the cial Ser-Gly-Ser-linker, followed in frame by the coding respective cell population were incubated for 30 min on ice 10 sequence of the human FAPalpha protein from amino acid 27 with 50 ul of cell culture supernatant of CHO cells transfected to 760 corresponding to the extracellular domains of human with the PSMA-directed bispecific antibody constructs. The FAPalpha, followed in frame by the coding sequence of an cells were washed twice in PBS and binding of the construct artificial Ser-Gly-linker, followed in frame by the coding was detected with an unlabeled murine Penta His antibody sequence of a 6 histidine tag and a stop codon. The gene (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After 15 synthesis fragment is also designed as to introduce restriction washing, bound anti His antibodies were detected with an Fc sites at the beginning and at the end of the fragment. The gamma-specific antibody (Dianova) conjugated to phyco introduced restriction sites, Spel at the 5' end and SalI at the erythrin, diluted 1:100 in 50 ul PBS with 2% FCS. Fresh 3' end, are utilized in the following cloning procedures. The culture medium was used as a negative control. gene synthesis fragment is cloned via Spel and Sal into a Flow cytometry was performed on a FACS-Calibur appa plasmid designated pEF-DHFR (pEF-DHFR is described in ratus, the CellOuest software was used to acquire and analyze Raum et al. Cancer Immuno) Immunother 50 (2001) 141 the data (Becton Dickinson biosciences, Heidelberg). FACS 150) following standard protocols. The aforementioned pro staining and measuring of the fluorescence intensity were cedures are all carried out according to standard protocols performed as described in Current Protocols in Immunology (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley 25 edition, Cold Spring Harbour Laboratory Press, Cold Spring Interscience, 2002). Harbour, N.Y. (2001)). A clone with sequence-verified nucle otide sequence is transfected into DHFR deficient CHO cells 6. Generation of Bispecific Single Chain Antibodies for eukaryotic expression of the construct. Eukaryotic protein Directed at Membrane-proximal Target Epitopes of expression in DHFR deficient CHO cells is performed as Human FAPalpha 30 described by Kaufmann R.J. (1990) Methods Enzymol. 185, 537-566. Gene amplification of the construct is induced by 6.1 Generation of CHO Cells Expressing Human FAPalpha increasing concentrations of methotrexate (MTX) to a final The coding sequence of human FAPalpha as published in concentration of up to 20 nM MTX. After two passages of GenBank (Accession number NM 004460) is obtained by stationary culture the cells are grown in roller bottles with gene synthesis according to standard protocols. The gene 35 nucleoside-free HyO PF CHO liquid soy medium (with 4.0 synthesis fragment is designed as to contain first a Kozak site mM L-Glutamine with 0.1% Pluronic F-68; HyClone) for 7 for eukaryotic expression of the construct followed by the days before harvest. The cells are removed by centrifugation coding sequence of the human FAPalpha protein and a stop and the Supernatant containing the expressed protein is stored codon (the cDNA and amino acid sequence of the construct is at -20° C. Alternatively a clone of the expression plasmid listed under SEQID Nos 366 and 367). The gene synthesis 40 with sequence-verified nucleotide sequence is used for trans fragment is also designed as to introduce restriction sites at fection and protein expression in the FreeStyle 293 Expres the beginning and at the end of the fragment. The introduced sion System (Invitrogen GmbH, Karlsruhe, Germany) restriction sites, Xmal at the 5' end and Sal at the 3' end, are according to the manufacturer's protocol. Supernatant con utilized in the following cloning procedures. The gene Syn taining the expressed protein is obtained, cells are removed by thesis fragment is cloned via Xmal and Sal into a plasmid 45 centrifugation and the Supernatant is stored at -20°C. designated pEF-DHFR (pEF-DHFR is described in Raum et Purification of the soluble human FAPalpha fusion protein al. Cancer Immunol Immunother 50 (2001) 141-150) follow is performed as follows: Akta R. Explorer System (GE Health ing standard protocols. The aforementioned procedures are Systems) and Unicorn R Software are used for chromatogra carried out according to standard protocols (Sambrook, phy. Immobilized metal affinity chromatography (“IMAC) Molecular Cloning: A Laboratory Manual, 3rd edition, Cold 50 is performed using a Fractogel EMD Chelate R (Merck) Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. which is loaded with ZnCl2 according to the protocol pro (2001)). A clone with sequence-verified nucleotide sequence vided by the manufacturer. The column is equilibrated with is transfected into DHFR deficient CHO cells for eukaryotic buffer A (20 mM sodium phosphate buffer pH 7.2, 0.1 M expression of the construct. Eukaryotic protein expression in NaCl) and the cell culture supernatant (500 ml) is applied to DHFR deficient CHO cells is performed as described by 55 the column (10 ml) at a flow rate of 3 ml/min. The column is Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. washed with buffer A to remove unbound sample. Bound Gene amplification of the construct is induced by increasing protein is eluted using a two step gradient of buffer B (20 mM concentrations of methotrexate (MTX) to a final concentra sodium phosphate buffer pH 7.2, 0.1 M NaCl, 0.5 MImida tion of up to 20 nM MTX. Zole) according to the following procedure: 6.2 Generation of a Soluble Human FAPalpha Fusion Protein 60 Step 1: 20% buffer B in 6 column volumes The coding sequence of human FAPalpha as described in Step 2: 100% buffer B in 6 column volumes Example 6.1 and the coding sequence of murine Lag3 as Eluted protein fractions from step 2 are pooled for further published in GenBank (Accession number NM 008479) are purification. All chemicals are of research grade and pur used for the construction of an artificial clNA sequence chased from Sigma (Deisenhofen) or Merck (Darmstadt). encoding a soluble fusion protein of human FAPalpha and 65 Gel filtration chromatography is performed on a HiLoad murine Lag3. To generate a construct for expression of the 16/60 Superdex 200 prep grade column (GE/Amersham) soluble human FAPalpha fusion protein a cDNA fragment is equilibrated with Equi-buffer (25 mM Citrate, 200 mM US 9,260,522 B2 77 78 Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow gene synthesis fragment is also designed as to introduce rate 1 ml/min) are subjected to standard SDS-PAGE and restriction sites at the 5' end (EcoRI) and at the 3' end (Sal I) Western Blot for detection. Prior to purification, the columnis of the cDNA fragment for cloning into the mammalian cell calibrated for molecular weight determination (molecular expression vector pEF-DHFR (pEF-DHFR is described in weight marker kit, Sigma MW GF-200). Protein concentra Raum et al. Cancer Immunol Immunother 50 (2001) 141 tions are determined using OD280 nm. 150). Internal restriction sites are removed by silent mutation 6.3 Generation of CHO Cells Expressing Murine Fapalpha of the coding sequence in the gene synthesis fragment. The The sequence of murine FAPalpha (NM 007986, Mus aforementioned procedures are carried out according to stan musculus fibroblast activation protein (Fap), mRNA, dard protocols (Sambrook, Molecular Cloning: A Laboratory National Center for Biotechnology Information, http://ww 10 Manual, 3rd edition, Cold Spring Harbour Laboratory Press, w.ncbi.nlm.nih.gov/entrez) is used to obtain a synthetic Cold Spring Harbour, N.Y. (2001)). A clone with sequence cDNA molecule by gene synthesis according to standard verified nucleotide sequence is transfected into DHFR defi protocols. The gene synthesis fragment is designed as to cient CHO cells for eukaryotic expression of the construct. contain first a Kozak site for eukaryotic expression of the Eukaryotic protein expression in DHFR deficient CHO cells construct followed by the coding sequence of the complete 15 is performed as described by Kaufmann R.J. (1990) Methods murine FAPalpha antigen, followed in frame by the coding Enzymol. 185, 537-566. Gene amplification for increased sequence of a FLAG-tag and a stop codon (the cDNA and antigen expression is induced by increasing concentrations of amino acid sequence of the construct is listed under SEQID methotrexate (MTX) to a final concentration of up to 20 nM Nos 370 and 371). An alternative construct identical to the MTX. aforementioned construct except for a C-terminal 6 Histi Cell surface expression of mutated human FAPalpha with dine-tag instead of the FLAG-tag is also generated. The gene murine membrane-distal epitopes by the generated transfec synthesis fragment is also designed as to introduce restriction tants is confirmed by flow cytometric binding analysis per sites at the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA formed as described herein using an anti-FLAG M2 antibody fragment for cloning into the mammalian cell expression (Sigma-Aldrich, Inc.; diluted 1:900 in 50 ul PBS with 2% vector pEF-DHFR (pEF-DHFR is described in Raum et al. 25 FCS) detected with an Fc gamma-specific antibody (Di Cancer Immunol Immunother 50 (2001) 141-150). The afore anova) conjugated to phycoerythrin. Expression of mutated mentioned procedures are carried out according to standard human FAPalpha with murine membrane-distal epitopes was protocols (Sambrook, Molecular Cloning: A Laboratory confirmed as shown in FIG. 10. Manual, 3rd edition, Cold Spring Harbour Laboratory Press, 6.5 Generation of CHO Cells Expressing a Mutated Murine Cold Spring Harbour, N.Y. (2001)). A clone with sequence 30 FAPalpha Antigen with Human Membrane-distal Epitopes verified nucleotide sequence is transfected into DHFR defi The coding sequence of a mutated murine FAPalpha anti cient CHO cells for eukaryotic expression of the construct. gen with human membrane-distal epitopes is obtained by Eukaryotic protein expression in DHFR deficient CHO cells gene synthesis according to standard protocols. The gene is performed as described by Kaufmann R.J. (1990) Methods synthesis fragment is designed as to contain first a Kozak site Enzymol. 185, 537-566. Gene amplification for increased 35 for eukaryotic expression of the construct followed by the antigen expression is induced by increasing concentrations of coding sequence of the complete murine FAPalpha antigen methotrexate (MTX) to a final concentration of up to 20 nM mutated at the same 15 specific extracellular amino acid MTX. Cell surface expression of murine FAPalpha by the positions as identified in the foregoing Example 6.4 to the generated transfectants is confirmed by flow cytometric bind respective homologous human amino acid, followed in frame ing analysis performed as described herein. In the case of the 40 by the coding sequence of a FLAG tag and a stop codon (the construct with the 6 Histidine tagamurine Penta His antibody cDNA and amino acid sequence of the construct is listed (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS) was used under SEQ ID Nos 374 and 375). The gene synthesis frag and detected with an Fc gamma-specific antibody (Dianova) ment is also designed as to introduce restriction sites at the 5' conjugated to phycoerythrin. Expression of murine FAPalpha end (EcoRI) and at the 3' end (Sal I) of the cDNA fragment for was confirmed as shown in FIG. 10. 45 cloning into the mammalian cell expression vector pEF 6.4 Generation of CHO Cells Expressing a Mutated Human DHFR(pEF-DHFR is described in Raumetal. Cancer Immu FAPalpha Antigen with Murine Membrane-distal Epitopes nol Immunother 50 (2001) 141-150). The aforementioned The coding sequence of a mutated human FAPalpha anti procedures are carried out according to standard protocols gen with murine membrane-distal epitopes is obtained by (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd gene synthesis according to standard protocols. The gene 50 edition, Cold Spring Harbour Laboratory Press, Cold Spring synthesis fragment is designed as to contain first a Kozak site Harbour, N.Y. (2001)). A clone with sequence-verified nucle for eukaryotic expression of the construct followed by the otide sequence is transfected into DHFR deficient CHO cells coding sequence of the complete human FAPalpha antigen for eukaryotic expression of the construct. Eukaryotic protein mutated at 15 specific amino acid positions as explained expression in DHFR deficient CHO cells is performed as below, followed in frame by the coding sequence of a FLAG 55 described by Kaufmann R.J. (1990) Methods Enzymol. 185, tag and a stop codon (the cDNA and amino acid sequence of 537-566. Gene amplification for increased antigen expres the construct is listed under SEQID Nos 372 and 373). All sion is induced by increasing concentrations of methotrexate extracellular amino acids of human FAPalpha mismatching (MTX) to a final concentration of up to 20 nM MTX. with the homologous murine sequence, whose alpha C-atoms 6.6 Immunization of Mice Using a Soluble Human FAPalpha have a distance of >60 A from the alpha C-atom of the 60 Fusion Protein thirteenth extracellular amino acid (i.e. the reference aa) as Twelve weeks old F1 mice from BALB? cxC57BL/6 cross counted from the junction of transmembrane and extracellu ings are immunized with the soluble human FAPalpha fusion lar region, are mutated to the homologous mismatched protein as described in Example 6.2. To this end for each murine amino acid. This applies to the 15 amino acids that are animal 40 ug of the soluble human FAPalpha fusion protein listed in table 2 and marked in bold. The homologous mis 65 are mixed with 10 nmol of a thioate-modified CpG-Oligo matched amino acids between human and murine FAPalpha nucleotide (5'-tccatgacgttcctgatgct-3') in 300 ul PBS and are are identified by sequence alignment as shown in FIG.9. The injected intraperitoneally. Mice receive booster immuniza US 9,260,522 B2 79 80 tions after 21, 42 and optionally 63 days in the same way. Ten fusion to phage coat protein III. This pool of phages display days after the first booster immunization, blood samples are ing the antibody library is later used for the selection of taken and antibody serum titers against human FAPalpha are antigen binding entities. tested by flow cytometry according to standard protocols. To 6.8 Phage Display Based Selection of Membrane-proximal this end 200,000 cells of the human FAPalpha transfected 5 Target Binders on CHO Cells Expressing the Mutated Human CHO cells as described in Example 6.1 are incubated for 30 FAPalpha Antigen with Murine Membrane-distal Epitopes min on ice with 50 ul of serum of the immunized animals The phage library carrying the cloned schv-repertoire is diluted 1:1000 in PBS with 2% FCS. The cells are washed harvested from the respective culture supernatant by twice in PBS with 2% FCS and binding of serum antibodies is PEG8000/NaCl precipitation and centrifugation. Approxi detected with a mouse Fc gamma-specific antibody (Di- 10 mately 10' to 10" scFv phage particles are resuspended in anova) conjugated to phycoerythrin, diluted 1:100 in PBS 0.4 ml of PBS/0.1% BSA and incubated with 10 to 107 CHO with 2% FCS. Serum of the animals obtained prior to immu cells expressing the mutated human FAPalpha antigen with nization is used as a negative control. murine membrane-distal epitopes as described in example 6.4 Flow cytometry is performed on a FACS-Calibur appara for 1 hour on ice under slow agitation. These CHO cells are tus, the CellOuest Software is used to acquire and analyze the 15 grown beforehand, harvested by centrifugation, washed in data (Becton Dickinson biosciences, Heidelberg). FACS PBS and resuspended in PBS/1% FCS (containing Na Azide). staining and measuring of the fluorescence intensity are per schv phage which do not specifically bind to the CHO cells formed as described in Current Protocols in Immunology are eliminated by up to five washing steps with PBS/1% FCS (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley (containing Na Azide). After washing, binding entities are Interscience, 2002). 2O eluted from the cells by resuspending the cells in HCl-glycine Animals demonstrating significant serum reactivity pH 2.2 (10 min incubation with Subsequent Vortexing) and against human FAPalpha as determined by the FACS analysis after neutralization with 2M Tris pH 12, the eluate is used for are used in the Subsequent experiment. infection of a fresh uninfected E. coli XL1 Blue culture 6.7 Generation of an Immune Murine Antibody scFV (OD600>0.5). The E. coli culture containing E. coli cells Library: Construction of a Combinatorial Antibody Library 25 Successfully transduced with a phagemid copy, encoding a and Phage Display murine ScFV-fragment, are again selected for carbenicillin Three days after the last booster immunization spleen cells resistance and subsequently infected with VCMS13 helper of reactive animals are harvested for the preparation of total phage to start the second round of antibody display and in RNA according to standard protocols. vitro selection. Typically a total of 4 to 5 rounds of selections A library of murine immunoglobulin (Ig) light chain 30 are carried out. (kappa) variable region (VK) and Ig heavy chain variable 6.9 Screening for Membrane-proximal Target Binders on region (VH) DNA-fragments is constructed by RT-PCR on CHO Cells Expressing the Human Fapalpha Antigen, the murine spleen RNA using VK- and VH specific primers. Murine Fapalpha Antigen and the Mutated Murine Fapalpha cDNA is synthesized according to standard protocols, see Antigen with Human Membrane-distal Epitopes example 2.7. 35 Plasmid DNA corresponding to 4 and 5 rounds of panning 450 ng of the kappa light chain fragments (SacI-Spel is isolated from E. coli cultures after selection. For the pro digested) are ligated with 1400 ng of the phagemid duction of soluble schv-protein, VH-VL-DNA fragments are pComb3H5Bhis (SacI-Spel digested; large fragment). The excised from the plasmids (XhoI-SpeI). These fragments are resulting combinatorial antibody library is then transformed cloned via the same restriction sites in the plasmid into 300 ul of electrocompetent Escherichia coli XL 1 Blue 40 pComb3H5BFlag? His differing from the original cells by electroporation (2.5 kV, 0.2 cm gap cuvette, 25uFD, pComb3H5BHis in that the expression construct (e.g. scFv) 200 Ohm, Biorad gene-pulser) resulting in a library size of includes a Flag-tag (TGDYKDDDDK) between the scFv and more than 107 independent clones. After one hour of pheno the His6-tag and the additional phage proteins are deleted. type expression, positive transformants are selected for car After ligation, each pool (different rounds of panning) of benicillin resistance encoded by the pComb3H5BHis vector 45 plasmid DNA is transformed into 100 ul heat shock compe in 100 ml of liquid super broth (SB)-culture over night. Cells tent E. coli TG1 or XLI blue and plated onto carbenicillin are then harvested by centrifugation and plasmid preparation LB-agar. Single colonies are picked into 100 ul of LB carb is carried out using a commercially available plasmid prepa (LB with 50 lug/ml carbenicillin). ration kit (Qiagen). After induction with 1 mMIPTG. E. coli transformed with 2800 ng of this plasmid-DNA containing the VK-library 50 pComb3H5BFlag? His containing a VL- and VH-segment (XhoI-BstEII digested; large fragment) are ligated with 900 produce soluble scFv in sufficient amounts. Due to a suitable ng of the heavy chain V-fragments (XhoI-BstEII digested) signal sequence, the Sclv is exported into the periplasma and again transformed into two 300 ulaliquots of electrocom where it folds into a functional conformation. petent E. coli XL1 Blue cells by electroporation (2.5 kV, 0.2 Single E. coli bacterial colonies from the transformation cm gap cuvette, 25uFD, 200Ohm) resulting in a totalVH-VK 55 plates are picked for periplasmic Small scale preparations and schv (single chain variable fragment) library size of more grown in SB-medium (e.g. 10 ml) supplemented with 20 mM than 107 independent clones. MgCl, and carbenicillin 50 ug/ml (and re-dissolved in PBS After phenotype expression and slow adaptation to carbe (e.g. 1 ml) after harvesting. A temperature shock is applied by nicillin, the E. coli cells containing the antibody library are four rounds of freezing at -70° C. and thawing at 37° C. transferred into SB-Carbenicillin (SB with 50 lug/mL carbe- 60 whereby the outer membrane of the bacteria is destroyed and nicillin) selection medium. The E. coli cells containing the the soluble periplasmic proteins including the Sclvs are antibody library are then infected with an infectious dose of released into the supernatant. After elimination of intact cells 10' particles of helper phage VCSM13 resulting in the pro and cell-debris by centrifugation, the Supernatant containing duction and secretion of filamentous M13 phage, wherein the murine anti-human FAPalpha-sch vs is collected and used each phage particle contains single stranded 65 for further examination. pComb3H5BHis-DNA encoding a murine schv-fragment Screening of the isolated schvs for membrane-proximal and displays the corresponding scFV-proteinas a translational target binders is performed by flow cytometry on CHO cells US 9,260,522 B2 81 82 expressing the human FAPalpha antigen as described in that have each one a different amount of human and murine Example 6.1, the murine FAPalpha antigen as described in residues at the respective differing framework positions (pool Example 6.3 and the mutated murine FAPalpha antigen with of humanized VH). The same procedure is performed for the human membrane-distal epitopes as described in Example VL region of the murine anti-FAPalpha scFv to a membrane 6.5. proximal target epitope of human FAPalpha (pool of human For flow cytometry 2.5x10 cells of the respective cell lines ized VL). The pool of humanized VH is then combined with are incubated with 50 ul supernatant. The binding of the the pool of humanized VL in the phage display vector constructs is detected with an anti-His antibody (Penta-His pComb3H5Bhis to form a library of functional sclvs from Antibody, BSA free, Qiagen GmbH, Hilden, FRG) at 2 ug/ml which—after display on filamentous phage—anti-FAPalpha in 50 ul PBS with 2% FCS. As a second step reagent an 10 R-Phycoerythrin-conjugated affinity purified F(ab')2 frag binders to membrane-proximal target epitopes of human ment, goat anti-mouse IgG (Fc-gamma fragment specific), FAPalpha are selected, screened, identified and confirmed as diluted 1:100 in 50 ul PBS with 2% FCS (Dianova, Hamburg, described above for the parental non-human (murine) anti FRG) is used. The samples are measured on a FACSscan (BD FAPalpha schv. Single clones are then analyzed for favorable biosciences, Heidelberg, FRG). 15 properties and amino acid sequence. Those scFVs, which are Only constructs which show binding to CHO cells express closest in amino acid sequence homology to human germline ing the human FAPalpha antigen and do not show binding to V-segments, are preferred. CHO cells expressing the murine FAPalpha antigen and also Human/humanized anti-FAPalpha schvs to membrane do not show binding to CHO cells expressing the mutated proximal target epitopes of human FAPalpha are converted murine FAPalpha antigen with human membrane-distal into recombinant bispecific single chain antibodies and fur epitopes are selected for further use. ther characterized as follows. 6.10 Generation of Human/Humanized Equivalents of Non 6.11 Generation of I2C-based Bispecific Single Chain Anti human schvs to Membrane-proximal Target Epitopes of bodies Directed at Membrane-proximal Target Epitopes of Human FAPalpha Human FAPalpha The VH region of a murine anti-FAPalpha scFv to a mem 25 Anti-FAPalpha schvs to membrane-proximal target brane-proximal target epitope of human FAPalpha is aligned epitopes of human FAPalpha with favorable properties and against human antibody germline amino acid sequences. The amino acid sequence are converted into recombinant bispe human antibody germline VH sequence is chosen which has cific single chain antibodies by joining them via a Gly-Ser the closest homology to the non-human VH and a direct linker with the CD3 specific sclv 120 (SEQID NO: 185) to alignment of the two amino acid sequences is performed. 30 result in constructs with the domain arrangement VHeart There are a number of framework residues of the non-human (Gly-Seri)s-VL4-Ser Gly-Ser-VHos-(Gly, Ser)- VH that differ from the human VH framework regions ("dif VL-ps. ferent framework positions'). Some of these residues may I2C-based bispecific single chain antibodies directed at contribute to the binding and activity of the antibody to its membrane-proximal target epitopes of human FAPalpha target. 35 were designed as set out in the following Table 5: To construct a library that contains the murine CDRs and at every framework position that differs from the chosen human VH sequence both possible residues (the human and the TABLE 5 maternal murine amino acid residue), degenerated oligo Formats of I2C-based bispecific single chain antibodies directed nucleotides are synthesized. These oligonucleotides incorpo 40 at membrane-proximal target epitopes of human FAPalpha rate at the differing positions the human residue with a prob SEQID Formats of protein constructs ability of 75% and the murine residue with a probability of (nucliprot) (N-> C) 25%. For one human VH e.g. six of these oligonucleotides have to be synthesized that overlap in a terminal stretch of 820,819 FA19D12HLXI2CHL approximately 20 nucleotides. To this end every second 45 primer is an antisense primer. Restriction sites within the 764,763 FA22C11HLXI2CHL oligonucleotides needed for later cloning are deleted. These primers may have a length of 60 to 90 nucleotides, depending on the number of primers that are needed to span over the whole V sequence. 50 These e.g. six primers are mixed in equal amounts (e.g. 1 ul Alternatively further constructs with different domain of each primer (primer stocks 20 to 100 uM) to a 20 lul PCR arrangements can be generated according to standard proto reaction) and added to a PCR mix consisting of PCR buffer, colls. For expression in CHO cells the coding sequences of (i) nucleotides and Taq polymerase. This mix is incubated at 94° an N-terminal immunoglobulin heavy chain leader compris C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° 55 ing a start codon embedded within a Kozak consensus C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° sequence and (ii) a C-terminal His6-tag followed by a stop C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. codon are both attached in frame to the nucleotide sequence Subsequently the product is run in an agarose gel electro encoding the bispecific single chain antibodies prior to inser phoresis and the product of a size from 200 to 400 base pairs tion of the resulting DNA-fragment as obtained by gene Syn isolated from the gel according to standard methods. 60 thesis into the multiple cloning site of the expression vector This PCR product is then used as a template for a standard pEF-DHFR (Raum et al. Cancer Immunol Immunother 50 PCR reaction using primers that incorporate suitable N-ter (2001) 141-150). A clone with sequence-verified nucleotide minal and C-terminal cloning restriction sites. The DNA frag sequence is transfected into DHFR deficient CHO cells for ment of the correct size (for a VH approximately 350 nucle eukaryotic expression of the construct. Eukaryotic protein otides) is isolated by agarose gel electrophoresis according to 65 expression in DHFR deficient CHO cells is performed as standard methods. In this way sufficientVHDNA fragment is described by Kaufmann R.J. (1990) Methods Enzymol. 185, amplified. This VH fragment is now a pool of VH fragments 537-566. Gene amplification of the construct is induced by US 9,260,522 B2 83 84 increasing concentrations of methotrexate (MTX) to a final Western Blot is performed using an Optitran?R BA-S83 concentration of up to 20 nM MTX. membrane and the Invitrogen Blot Module according to the 6.12 Expression and Purification of Bispecific Single Chain protocol provided by the manufacturer. The antibody used is Antibody Molecules Directed at Membrane-proximal Target directed against the His Tag (Penta His, Qiagen) and a Goat Epitopes of Human FAPalpha anti-mouse Ig labeled with alkaline phosphatase (AP) Bispecific single chain antibody molecules are expressed (Sigma) is used as second step reagent, and BCIP/NBT in Chinese hamster ovary cells (CHO). Eukaryotic protein (Sigma) as substrate. A band detected at 52 kD corresponds to expression in DHFR deficient CHO cells is performed as purified bispecific single chain antibodies. described by Kaufmann R.J. (1990) Methods Enzymol. 185, 6.13 Flow Cytometric Binding Analysis of Bispecific Anti 537-566. Gene amplification of the constructs is induced by 10 bodies Directed at Membrane-proximal Target Epitopes of addition of increasing concentrations of MTX up to final Human FAPalpha concentrations of 20 nMMTX. After two passages of station In order to test the functionality of bispecific antibody ary culture cell culture Supernatant is collected and used in the constructs regarding the capability to bind to CD3 and to Subsequent experiments. To generate Supernatant for purifi human FAPalpha, respectively, a FACS analysis is per cation after two passages of Stationary culture the cells are 15 formed. For this purpose CHO cells transfected with human grown in roller bottles with nucleoside-free HyO PF CHO FAPalpha as described in Example 6.1 and the human CD3 liquid soy medium (with 4.0 mM L-Glutamine with 0.1% positive T cell leukemia cell line HPB-ALL (DSMZ, Braun Pluronic F-68; HyClone) for 7 days before harvest. The cells schweig, ACC483) are used. 200,000 cells of the respective are removed by centrifugation and the Supernatant containing cell lines are incubated for 30 min on ice with 50 ul of cell the expressed protein is stored at -20°C. Alternatively, con culture Supernatant of transfected cells expressing the bispe structs are transiently expressed in HEK 293 cells. Transfec cific antibody constructs. The cells are washed twice in PBS tion is performed with 293fectin reagent (Invitrogen, #12347 with 2% FCS and binding of the construct is detected with a 019) according to the manufacturer's protocol. Furthermore murine Penta His antibody (Qiagen; diluted 1:20 in 50 ul PBS the constructs are alternatively expressed in transiently trans with 2% FCS). After washing, bound anti His antibodies are fected DHFR deficient CHO cells using for example 25 detected with an Fc gamma-specific antibody (Dianova) con FuGENER HD Transfection Reagent (Roche Diagnostics jugated to phycoerythrin, diluted 1:100 in PBS with 2% FCS. GmbH, Cat. No. 04709691001) according to the manufactur Supernatant of untransfected cells is used as a negative con er's protocol. trol. Akta R. Explorer System (GE Health Systems) and Uni Flow cytometry is performed on a FACS-Calibur appara corn R Software are used for chromatography. Immobilized 30 tus, the CellOuest Software is used to acquire and analyze the metal affinity chromatography (“IMAC) is performed using data (Becton Dickinson biosciences, Heidelberg). FACS a Fractogel EMD ChelateR (Merck) which is loaded with staining and measuring of the fluorescence intensity are per ZnCl2 according to the protocol provided by the manufac formed as described in Current Protocols in Immunology turer. The column is equilibrated with buffer A (20 mM (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley sodium phosphate buffer pH 7.2, 0.1 M NaCl) and the cell 35 Interscience, 2002). culture supernatant (500 ml) is applied to the column (10 ml) Only those constructs that show bispecific binding to at a flow rate of 3 ml/min. The column is washed with buffer human CD3 as well as to human FAPalpha are selected for A to remove unbound sample. Bound protein is eluted using further use. a two step gradient of buffer B (20 mM sodium phosphate The bispecific binding of the single chain molecules listed buffer pH 7.2, 0.1 MNaCl, 0.5 MImidazole) according to the 40 above was clearly detectable as shown in FIG. 11. In the following procedure: FACS analysis all constructs showed binding to human CD3 Step 1: 20% buffer B in 6 column volumes and human FAPA compared to the negative control. Step 2: 100% buffer B in 6 column volumes 6.14 Bioactivity of Bispecific Antibodies Directed at Mem Eluted protein fractions from step 2 are pooled for further brane-proximal Target Epitopes of Human FAPalpha purification. All chemicals are of research grade and pur 45 Bioactivity of generated bispecific single chain antibodies chased from Sigma (Deisenhofen) or Merck (Darmstadt). is analyzed by chromium 51 (Cr) release in vitro cytotox Gel filtration chromatography is performed on a HiLoad icity assays using the CHO cells transfected with human 16/60 Superdex 200 prep grade column (GE/Amersham) FAPalpha described in Example 6.1. To confirm that signifi equilibrated with Equi-buffer (25 mM Citrate, 200 mM cant bioactivity is only recruited by binding to membrane Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow 50 proximal target epitopes of human FAPalpha in addition— rate 1 ml/min) are subjected to standard SDS-PAGE and CHO cells expressing the murine FAPalpha antigen as Western Blot for detection. Prior to purification, the columnis described in Example 6.3 and CHO cells expressing the calibrated for molecular weight determination (molecular mutated human FAPalpha antigen with murine membrane weight marker kit, Sigma MW GF-200). Protein concentra distal epitopes as described in Example 6.4 are used. As tions are determined using OD280 nm. 55 effector cells stimulated human CD4/CD56 depleted PBMC Purified bispecific single chain antibody protein is ana are used. lyzed in SDS PAGE under reducing conditions performed Stimulated human PBMC are obtained as follows: with pre-cast 4-12% Bis Tris gels (Invitrogen). Sample prepa A Petri dish (145 mm diameter, Greiner bio-one GmbH, ration and application are performed according to the proto Kremsmünster) is coated with a commercially available anti col provided by the manufacturer. The molecular weight is 60 CD3 specific antibody (e.g. OKT3. Orthoclone) in a final determined with MultiMark protein standard (Invitrogen). concentration of 1 lug/ml for 1 hour at 37°C. Unbound protein The gel is stained with colloidal Coomassie (Invitrogen pro is removed by one washing step with PBS. The fresh PBMC tocol). The purity of the isolated protein is typically >95% as are isolated from peripheral blood (30-50 ml human blood) determined by SDS-PAGE. by Ficoll gradient centrifugation according to standard pro The bispecific single chain antibody has a molecular 65 tocols. 3-5x107 PBMC are added to the precoated petridishin weight of about 52 kDa under native conditions as determined 120 ml of RPMI 1640 with stabilized glutamine/10% FCS/ by gel filtration in PBS. IL-2 20 U/ml (Proleukin, Chiron) and stimulated for 2 days. US 9,260,522 B2 85 86 On the third day the cells are collected and washed once with - Continued RPMI 1640. IL2 is added to a final concentration of 20U/ml reverse prlmer: 5' - cacatttgaaaagaccagttccagatgc-3' SEQ ID and the cells are cultivated again for one day in the same cell NO: 381 culture medium as above. By depletion of CD4+ T cells and CD56+ NK cells accord 4. forward primer: 5'-agattacagotgtcagaaaatt catagaaatgg-3 '' SEQ ID ing to standard protocols CD8+ cytotoxic T lymphocytes NO: 382 (CTLs) are enriched. reverse primer: Target cells are washed twice with PBS and labeled with 5'-atata aggttitt cagattctgatacaggc-3' SEQ ID 11.1 MBq 'Cr in a final volume of 100 ul RPMI with 50% NO: 383 FCS for 60 minutes at 37° C. Subsequently the labeled target 10 These PCRs generate four overlapping fragments, which cells are washed 3 times with 5 ml RPMI and then used in the cytotoxicity assay. The assay is performed in a 96 well plate are isolated and sequenced according to standard protocols in a total volume of 250 ul supplemented RPMI (as above) using the PCR primers, and thereby provided the cDNA with an E:T ratio of 10:1. Supernatant of cells expressing the sequence coding macaque FAPalpha. To generate a construct bispecific single chain antibody molecules in a final concen 15 for expression of macaque FAPalpha a cDNA fragment is tration of 6.6% and 14 threefold dilutions thereofare applied. obtained by gene synthesis according to standard protocols The assay time is 18 hours. Cytotoxicity is measured as (the cDNA and amino acid sequence of the construct is listed relative values of released chromium in the supernatant under SEQID Nos 384 and 385). This construct contains the related to the difference of maximum lysis (addition of Tri complete coding sequence of macaque FAPalpha followed by ton-X) and spontaneous lysis (without effector cells). All a stop codon. The gene synthesis fragment is also designed as measurements are done in quadruplicates. Measurement of to contain a Kozak site for eukaryotic expression of the con chromium activity in the supernatants is performed with a struct and restriction sites at the beginning and the end of the Wizard R. 3" gammacounter (Perkin Elmer Life Sciences fragment containing the cDNA. The introduced restriction GmbH, Koln, Germany). Analysis of the experimental data is sites, EcoRI at the 5' end and Sall at the 3' end, are utilised in performed with Prism 4 for Windows(R (version 4.02, Graph 25 the following cloning procedures. The gene synthesis frag Pad Software Inc., San Diego, Calif., USA). Sigmoidal dose ment is cloned via EcoRI and Sal into a plasmid designated response curves typically have R values >0.90 as determined pEF-DHFR (pEF-DHFR is described in Raum et al. Cancer by the software. Immunol Immunother 50 (2001) 141-150) following stan Only those constructs showing potent recruitment of cyto dard protocols. The aforementioned procedures are carried toxic activity of effector T cells against target cells positive 30 out according to standard protocols (Sambrook, Molecular for FAPalpha are selected for further use. As shown in FIG. 12 Cloning: A Laboratory Manual, 3rd edition, Cold Spring all of the generated bispecific antibodies directed at mem Harbour Laboratory Press, Cold Spring Harbour, N.Y. brane-proximal target epitopes of human FAPalpha demon (2001)). A clone with sequence-verified nucleotide sequence strated cytotoxic activity against human FAPA positive target is transfected into DHFR deficient CHO cells for eukaryotic cells and target cells positive for the mutated human FAPal 35 expression of the construct. Eukaryotic protein expression in pha antigen with murine membrane-distal epitopes elicited DHFR deficient CHO cells is performed as described by by stimulated human CD4/CD56 depleted PBMC but did not Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. recruit significant cytotoxic activity against murine FAPalpha Gene amplification of the construct is induced by increasing positive target cells. Thereby specific recruitment of cyto concentrations of methotrexate (MTX) to a final concentra toxic activity via binding to membrane-proximal target 40 tion of up to 20 nM MTX. epitopes of human FAPalpha was confirmed. 6.16 Flow Cytometric Analysis of Cross-species Specificity 6.15 Generation of CHO Cells Expressing Macaque FAPal of Bispecific Antibodies Directed at Membrane-proximal pha Target Epitopes of Human FAPalpha The cDNA sequence of macaque FAPalpha is obtained by In order to test the cross-species specificity of bispecific a set of four PCRs on cDNA from macaque monkey skin 45 antibodies directed at membrane-proximal target epitopes of prepared according to standard protocols. The following human FAPalpha the capability of the constructs to bind to reaction conditions: 1 cycle at 94° C. for 3 minutes followed macaque FAPalpha and macaque CD3, respectively, is inves by 40 cycles with 94° C. for 0.5 minutes, 56° C. for 0.5 tigated by FACS analysis. For this purpose the macaque minutes and 72°C. for 3 minutes followed by a terminal cycle FAPalpha transfected CHO cells as described in example 50 6.15 and the macaque T cell line 4119LnPx (kindly provided of 72° C. for 3 minutes and the following primers are used: by Prof Fickenscher, Hygiene Institute, Virology, Erlangen Nuernberg; published in Knappe A, et al., and Fickenscher 1. forward primer: H. Blood 2000, 95, 3256-61) are used. 200,000 cells of the 5'-cagott C caactacaaagacagac-3' SEQ ID respective cell lines are incubated for 30 minonice with 50 ul NO : 376 55 reverse primer: of cell culture supernatant of transfected cells expressing the 5'-tttcctcitt cataaaccoagtctgg-3' SEQ ID cross-species specific bispecific antibody constructs. The NO : 377 cells are washed twice in PBS with 2% FCS and binding of the construct is detected with a murine Penta His antibody 2. forward primer: 5'-ttgaaacaaagaccaggagatccacc-3' SEQ ID (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After 60 NO : 378 washing, bound anti His antibodies are detected with an Fc reverse primer: gamma-specific antibody (Dianova) conjugated to phyco 5'-agatggcaagtaa.ca cacttcttgc-3' SEQ ID erythrin, diluted 1:100 in PBS with 2% FCS. Supernatant of NO : 379 untransfected cells is used as a negative control. 3. forward primer: Flow cytometry is performed on a FACS-Calibur appara 5'-gaagaaa Catctacaga attagcattgg-3' SEQ ID 65 tus, the CellOuest software is used to acquire and analyze the NO: 380 data (Becton Dickinson biosciences, Heidelberg). FACS staining and measuring of the fluorescence intensity are per US 9,260,522 B2 87 88 formed as described in Current Protocols in Immunology described in Raum et al. Cancer Immunol Immunother 50 (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley (2001) 141-150) following standard protocols. The afore Interscience, 2002). mentioned procedures are all carried out according to stan The cross-species specific binding of the single chain mol dard protocols (Sambrook, Molecular Cloning: A Laboratory ecules listed above was clearly detectable as shown in FIG. Manual, 3rd edition, Cold Spring Harbour Laboratory Press, 11. In the FACS analysis all constructs showed binding to Cold Spring Harbour, N.Y. (2001)). A clone with sequence macaque CD3 and macaque FAPA compared to the negative verified nucleotide sequence is transfected into DHFR defi control. cient CHO cells for eukaryotic expression of the construct. Eukaryotic protein expression in DHFR deficient CHO cells 7. Generation of Bispecific Single Chain Antibodies 10 is performed as described by Kaufmann R.J. (1990) Methods Directed at Membrane-proximal Target Epitopes of Enzymol. 185, 537-566. Gene amplification of the construct Human c-MET is induced by increasing concentrations of methotrexate (MTX) to a final concentration of up to 20 nM MTX. After 7.1 Generation of CHO Cells Expressing Human c-MET two passages of stationary culture the cells are grown in roller The coding sequence of human c-MET as published in 15 bottles with nucleoside-free HyOPF CHO liquid soy medium GenBank (Accession number NM 000245) is obtained by (with 4.0 mM L-Glutamine with 0.1% Pluronic F-68; gene synthesis according to standard protocols. The gene HyClone) for 7 days before harvest. The cells are removed by synthesis fragment is designed as to contain first a Kozak site centrifugation and the Supernatant containing the expressed for eukaryotic expression of the construct followed by the protein is stored at -20°C. Alternatively a clone of the expres coding sequence of the human c-MET protein and a stop sion plasmid with sequence-verified nucleotide sequence is codon (the cDNA and amino acid sequence of the construct is used for transfection and protein expression in the FreeStyle listed under SEQID Nos 368 and 387). The gene synthesis 293 Expression System (Invitrogen GmbH, Karlsruhe, Ger fragment is also designed as to introduce restriction sites at many) according to the manufacturer's protocol. Supernatant the beginning and at the end of the fragment. The introduced containing the expressed protein is obtained, cells are restriction sites, EcoRI at the 5' end and Sal at the 3' end, are 25 removed by centrifugation and the Supernatant is stored at utilized in the following cloning procedures. Internal restric -20° C. tion sites are removed by silent mutation of the coding For purification of the soluble human c-MET fusion pro sequence in the gene synthesis fragment. The gene synthesis tein a goat anti-mouse Fc affinity column is prepared accord fragment is cloned via EcoRI and SalI into a plasmid desig ing to standard protocols using a commercially available nated pEF-DHFR (pEF-DHFR is described in Raum et al. 30 affinity purified goat anti-mouse IgG Fc fragment specific Cancer Immunol Immunother 50 (2001) 141-150) following antibody with minimal cross-reaction to human, bovine and standard protocols. The aforementioned procedures are car horse serum proteins (Jackson ImmunoResearch Europe ried out according to standard protocols (Sambrook, Molecu Ltd.). Using this affinity column the fusion protein is isolated lar Cloning: A Laboratory Manual, 3rd edition, Cold Spring out of cell culture supernatant on an Akta Explorer System Harbour Laboratory Press, Cold Spring Harbour, N.Y. 35 (GE Amersham) and eluted by citric acid. The eluate is neu (2001)). A clone with sequence-verified nucleotide sequence tralized and concentrated. is transfected into DHFR deficient CHO cells for eukaryotic 7.3 Generation of CHO Cells Expressing Murine c-MET expression of the construct. Eukaryotic protein expression in The sequence of murine c-MET (NM 008591 Mus mus DHFR deficient CHO cells is performed as described by culus met proto-oncogene (Met), mRNA, National Center for Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. 40 Biotechnology Information, http://www.ncbi.nlm.nih.gov/ Gene amplification of the construct is induced by increasing entrez) is used to obtain a synthetic cDNA molecule by gene concentrations of methotrexate (MTX) to a final concentra synthesis according to standard protocols. The gene synthesis tion of up to 20 nM MTX. fragment is designed as to contain the coding sequence of an 7.2 Generation of a Soluble Human c-MET Fusion Protein immunoglobulin heavy chain leader comprising a start codon The modified coding sequence of human c-MET as 45 embedded within a Kozak consensus sequence, followed in described in Example 7.1 is used for the construction of an frame by the coding sequence of a FLAG tag, followed in artificial cDNA sequence encoding a soluble fusion protein of frame by the complete coding sequence of the mature murine human c-MET and murine IgG1 Fc. To generate a construct c-MET (the cDNA and amino acid sequence of the construct for expression of the soluble human c-MET fusion protein a is listed under SEQID Nos 390 and 391). The gene synthesis cDNA fragment is obtained by gene synthesis according to 50 fragment is also designed as to introduce restriction sites at standard protocols (the cDNA and amino acid sequence of the the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA construct is listed under SEQID Nos 388 and 389). The gene fragment for cloning into the mammalian cell expression synthesis fragment is designed as to contain first a Kozak site vector pEF-DHFR (pEF-DHFR is described in Raum et al. for eukaryotic expression of the construct followed by the Cancer Immunol Immunother 50 (2001) 141-150) Internal coding sequence of the human c-MET protein from amino 55 restriction sites are removed by silent mutation of the coding acid 1 to 932 corresponding to the signal peptide and extra sequence in the gene synthesis fragment. The aforementioned cellular domains of human c-MET, followed in frame by the procedures are carried out according to standard protocols coding sequence of an artificial Ser-Gly-Ser-linker, fol (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd lowed in frame by the coding sequence of the hinge region edition, Cold Spring Harbour Laboratory Press, Cold Spring and Fc gamma portion of murine IgG1, followed in frame by 60 Harbour, N.Y. (2001)). A clone with sequence-verified nucle the coding sequence of a 6 histidine tag and a stop codon. The otide sequence is transfected into DHFR deficient CHO cells gene synthesis fragment is also designed as to introduce for eukaryotic expression of the construct. Eukaryotic protein restriction sites at the beginning and at the end of the frag expression in DHFR deficient CHO cells is performed as ment. The introduced restriction sites, EcoRI at the 5' end and described by Kaufmann R.J. (1990) Methods Enzymol. 185, SalI at the 3' end, are utilized in the following cloning proce 65 537-566. Gene amplification for increased antigen expres dures. The gene synthesis fragment is cloned via EcoRI and sion is induced by increasing concentrations of methotrexate Sall into a plasmid designated pEF-DHFR (pEF-DHFR is (MTX) to a final concentration of up to 20 nM MTX. US 9,260,522 B2 89 90 7.4 Generation of CHO Cells Expressing a Mutated Human Twelve weeks old F1 mice from BALB? cxC57BL/6 cross c-MET Antigen with Murine Membrane-distal Epitopes ings are immunized with the soluble human c-MET fusion The coding sequence of a mutated human c-MET antigen protein as described in Example 7.2. To this end for each with murine membrane-distal epitopes is obtained by gene animal 40 ug of the soluble human c-MET fusion protein are synthesis according to standard protocols. The gene synthesis mixed with 10 nmol of a thioate-modified CpG-Oligonucle fragment is designed as to contain the coding sequence of an otide (5'-tccatgacgttcctgatgct-3") in 300 ul PBS and are immunoglobulin heavy chain leader comprising a start codon injected intraperitoneally. Mice receive booster immuniza embedded within a Kozak consensus sequence, followed in tions after 21, 42 and optionally 63 days in the same way. Ten frame by the coding sequence of a FLAG tag, followed in days after the first booster immunization, blood samples are 10 taken and antibody serum titers against human c-MET are frame by the coding sequence of the alpha-chain of the sema tested by flow cytometry according to standard protocols. To domain of mature murine c-MET followed in frame by this end 200,000 cells of the human c-MET transfected CHO human c-MET from the beta-chain of the sema domain to the cells as described in Example 7.17 are incubated for 30 min stop codon (the cDNA and amino acid sequence of the con on ice with 50 ul of serum of the immunized animals diluted struct is listed under SEQ ID Nos 392 and 393). The gene 15 1:1000 in PBS with 2% FCS. The cells are washed twice in synthesis fragment is also designed as to introduce restriction PBS with 2% FCS and binding of serum antibodies is sites at the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA detected with an mouse Fc gamma-specific antibody (Di fragment for cloning into the mammalian cell expression anova) conjugated to phycoerythrin, diluted 1:100 in PBS vector pEF-DHFR (pEF-DHFR is described in Raum et al. with 2% FCS. Serum of the animals obtained prior to immu Cancer Immunol Immunother 50 (2001) 141-150). Internal nization is used as a negative control. Flow cytometry is restriction sites are removed by silent mutation of the coding performed on a FACS-Calibur apparatus, the CellOuest soft sequence in the gene synthesis fragment. The aforementioned ware is used to acquire and analyze the data (Becton Dickin procedures are carried out according to standard protocols son biosciences, Heidelberg). FACS staining and measuring (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd of the fluorescence intensity are performed as described in edition, Cold Spring Harbour Laboratory Press, Cold Spring 25 Current Protocols in Immunology (Coligan, Kruisbeek, Mar Harbour, N.Y. (2001)). A clone with sequence-verified nucle gullies. Shevach and Strober, Wiley-Interscience, 2002). otide sequence is transfected into DHFR deficient CHO cells Animals demonstrating significant serum reactivity for eukaryotic expression of the construct. Eukaryotic protein against human c-MET as determined by the FACS analysis expression in DHFR deficient CHO cells is performed as are used in the Subsequent experiment. described by Kaufmann R.J. (1990) Methods Enzymol. 185, 30 7.7 Generation of an Immune Murine Antibody schv Library: 537-566. Gene amplification for increased antigen expres Construction of a Combinatorial Antibody Library and Phage sion is induced by increasing concentrations of methotrexate Display (MTX) to a final concentration of up to 20 nM MTX. Three days after the last booster immunization spleen cells 7.5 Generation of CHO Cells Expressing a Mutated Murine of reactive animals are harvested for the preparation of total c-MET Antigen with Human Membrane-distal Epitopes 35 RNA according to standard protocols. The coding sequence of a mutated murine c-MET antigen A library of murine immunoglobulin (Ig) light chain with human membrane-distal epitopes is obtained by gene (kappa) variable region (VK) and Ig heavy chain variable synthesis according to standard protocols. The gene synthesis region (VH) DNA-fragments is constructed by RT-PCR on fragment is designed as to contain the coding sequence of an murine spleen RNA using VK- and VH specific primers. immunoglobulin heavy chain leader comprising a start codon 40 cDNA is synthesized according to standard protocols, see embedded within a Kozak consensus sequence, followed in example 2.7. frame by the coding sequence of a FLAG tag, followed in 450 ng of the kappa light chain fragments (SacI-Spel frame by the coding sequence of the alpha-chain of the sema digested) are ligated with 1400 ng of the phagemid domain of mature human c-MET, followed in frame by pComb3H5Bhis (SacI-Spel digested; large fragment). The murine c-MET from the beta-chain of the sema domain to the 45 resulting combinatorial antibody library is then transformed stop codon (the cDNA and amino acid sequence of the con into 300 ul of electrocompetent Escherichia coli XL 1 Blue struct is listed under SEQ ID Nos 394 and 395). The gene cells by electroporation (2.5 kV, 0.2 cm gap cuvette, 25uFD, synthesis fragment is also designed as to introduce restriction 200 Ohm, Biorad gene-pulser) resulting in a library size of sites at the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA more than 107 independent clones. After one hour of pheno fragment for cloning into the mammalian cell expression 50 type expression, positive transformants are selected for car vector pEF-DHFR (pEF-DHFR is described in Raum et al. benicillin resistance encoded by the pComb3H5BHis vector Cancer Immunol Immunother 50 (2001) 141-150). Internal in 100 ml of liquid super broth (SB)-culture over night. Cells restriction sites are removed by silent mutation of the coding are then harvested by centrifugation and plasmid preparation sequence in the gene synthesis fragment. The aforementioned is carried out using a commercially available plasmid prepa procedures are carried out according to standard protocols 55 ration kit (Qiagen). (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd 2800 ng of this plasmid-DNA containing the VK-library edition, Cold Spring Harbour Laboratory Press, Cold Spring (XhoI-BstEII digested; large fragment) are ligated with 900 Harbour, N.Y. (2001)). A clone with sequence-verified nucle ng of the heavy chain V-fragments (XhoI-BstEII digested) otide sequence is transfected into DHFR deficient CHO cells and again transformed into two 300 ulaliquots of electrocom for eukaryotic expression of the construct. Eukaryotic protein 60 petent E. coli XL 1 Blue cells by electroporation (2.5 kV, 0.2 expression in DHFR deficient CHO cells is performed as cm gap cuvette, 25uFD, 200Ohm) resulting in a totalVH-VK described by Kaufmann R.J. (1990) Methods Enzymol. 185, schv (single chain variable fragment) library size of more 537-566. Gene amplification for increased antigen expres than 107 independent clones. sion is induced by increasing concentrations of methotrexate After phenotype expression and slow adaptation to carbe (MTX) to a final concentration of up to 20 nM MTX. 65 nicillin, the E. coli cells containing the antibody library are 7.6 Immunization of Mice Using a Soluble Human c-MET transferred into SB-Carbenicillin (SB with 50 lug/mL carbe Fusion Protein nicillin) selection medium. The E. coli cells containing the US 9,260,522 B2 91 92 antibody library are then infected with an infectious dose of whereby the outer membrane of the bacteria is destroyed and 10' particles of helper phage VCSM 13 resulting in the the soluble periplasmic proteins including the Sclvs are production and secretion of filamentous M13 phage, wherein released into the supernatant. After elimination of intact cells each phage particle contains single stranded and cell-debris by centrifugation, the Supernatant containing pComb3H5BHis-DNA encoding a murine schv-fragment the murine anti-human c-MET-scFVs is collected and used for and displays the corresponding scFV-proteinas a translational further examination. fusion to phage coat protein III. This pool of phages display Screening of the isolated schvs for membrane-proximal ing the antibody library is later used for the selection of target binders is performed by flow cytometry on CHO cells antigen binding entities. expressing the human c-MET antigen as described in 7.8 Phage Display Based Selection of Membrane-proximal 10 Example 7.17, the murine c-MET antigen as described in Target Binders on CHO Cells Expressing the Mutated Human Example 7.17, the mutated human c-MET antigen with c-MET Antigen with Murine Membrane-distal Epitopes murine membrane-distal epitopes as described in Example The phage library carrying the cloned schv-repertoire is 7.17 and the mutated murine c-MET antigen with human harvested from the respective culture supernatant by membrane-distal epitopes as described in Example 7.17. PEG8000/NaCl precipitation and centrifugation. Approxi 15 For flow cytometry 2.5x10 cells of the respective cell lines mately 10' to 10' sclv phage particles are resuspended in are incubated with 50 ul supernatant. The binding of the 0.4 ml of PBS/0.1% BSA and incubated with 10 to 107 CHO constructs is detected with an anti-His antibody (Penta-His cells expressing the mutated human c-MET antigen with Antibody, BSA free, Qiagen GmbH, Hilden, FRG) at 2 g/ml murine membrane-distal epitopes as described in example in 50 lul PBS with 2% FCS. As a second step reagent a 7.17 for 1 hour on ice under slow agitation. These CHO cells R-Phycoerythrin-conjugated affinity purified F(ab')2 frag are grown beforehand, harvested by centrifugation, washed in ment, goat anti-mouse IgG (Fc-gamma fragment specific), PBS and resuspended in PBS/1% FCS (containing Na Azide). diluted 1:100 in 50 ul PBS with 2% FCS (Dianova, Hamburg, schv phage which do not specifically bind to the CHO cells FRG) is used. The samples are measured on a FACSscan (BD are eliminated by up to five washing steps with PBS/1% FCS biosciences, Heidelberg, FRG). (containing Na Azide). After washing, binding entities are 25 Only constructs which show binding to CHO cells express eluted from the cells by resuspending the cells in HCl-glycine ing the human c-MET antigen and show binding to CHO cells pH 2.2 (10 min incubation with Subsequent Vortexing) and expressing the mutated human c-MET antigen with murine after neutralization with 2M Tris pH 12, the eluate is used for membrane-distal epitopes and do not show binding to CHO infection of a fresh uninfected E. coli XL1 Blue culture cells expressing the murine c-MET antigen and also do not (OD600>0.5). The E. coli culture containing E. coli cells 30 show binding to CHO cells expressing the mutated murine Successfully transduced with a phagemid copy, encoding a c-MET antigen with human membrane-distal epitopes are murine sclv-fragment, are again selected for carbenicillin selected for further use. resistance and subsequently infected with VCMS13 helper ScFv specific for membrane proximal epitopes of human phage to start the second round of antibody display and in cMET were generated as described above and designated as vitro selection. Typically a total of 4 to 5 rounds of selections 35 set out in the following Table 6: are carried out. 7.9 Screening for Membrane-proximal Target Binders on TABLE 6 CHO Cells Expressing the Human c-MET Antigen, the Murine c-MET Antigen, the Mutated Human c-MET Antigen Designation of single chain antibody fragments with Murine Membrane-distal Epitopes and the Mutated 40 SEQ ID Murine c-MET Antigen with Human Membrane-distal (nucliprot) Designation Epitopes 734,733 MEO6F2HL Plasmid DNA corresponding to 4 and 5 rounds of panning 720,719 MEO6E1 OHL is isolated from E. coli cultures after selection. For the pro 706,705 MEO6D2HL duction of soluble schv-protein, VH-VL-DNA fragments are 45 692,691 MEO6D1HL excised from the plasmids (XhoI-SpeI). These fragments are 664f663 MEO6C7HL 6SOf 649 MEO6C6HL cloned via the same restriction sites in the plasmid 678,677 MEO6B7EHL pComb3H5BFlag? His differing from the original 636,635 MEOSF6HL pComb3H5BHis in that the expression construct (e.g. scFv) 622,621 MEOSB7EHL includes a Flag-tag (TGDYKDDDDK) between the scFv and 50 608,607 ME99B1HL the His6-tag and the additional phage proteins are deleted. 594,593 ME7SH6HL After ligation, each pool (different rounds of panning) of plasmid DNA is transformed into 100 ul heat shock compe Membrane-proximal target binding of the single chain tent E. coli TG1 or XLI blue and plated onto carbenicillin molecules listed above was clearly detectable as shown in LB-agar. Single colonies are picked into 100 ul of LB carb 55 FIG. 16. In the FACS analysis all constructs showed binding (LB with 50 lug/ml carbenicillin). to the human c-MET antigen and showed binding to the After induction with 1 mMIPTG. E. coli transformed with mutated human c-MET antigen with murine membrane-dis pComb3H5BFlag? His containing a VL- and VH-segment tal epitopes and did not show binding to the murine c-MET produce soluble scFv in sufficient amounts. Due to a suitable antigen and did also not show binding to the mutated murine signal sequence, the Sclv is exported into the periplasma 60 c-MET antigen with human membrane-distal epitopes as where it folds into a functional conformation. compared to the negative control. Single E. coli bacterial colonies from the transformation 7.10 Generation of Human/Humanized Equivalents of Non plates are picked for periplasmic Small scale preparations and human schvs to Membrane-proximal Target Epitopes of grown in SB-medium (e.g. 10 ml) supplemented with 20 mM Human c-Met MgCl, and carbenicillin 50 ug/ml (and re-dissolved in PBS 65 The VH region of a murine anti-c-MET scFv to a mem (e.g. 1 ml) after harvesting. A temperature shock is applied by brane-proximal target epitope of human c-MET is aligned four rounds of freezing at -70° C. and thawing at 37° C. against human antibody germline amino acid sequences. The US 9,260,522 B2 93 94 human antibody germline VH sequence is chosen which has chain antibodies by joining them via a Gly-Ser-linker with the closest homology to the non-human VH and a direct the CD3 specific scFv 120 (SEQ ID NO: 185) to result in alignment of the two amino acid sequences is performed. constructs with the domain arrangement VH-(Gly There are a number of framework residues of the non-human Ser)-VL-7-Ser Gly-Ser-VH-(Gly Ser)--VLs. VH that differ from the human VH framework regions (“dif 5 I2C-based bispecific single chain antibodies directed at ferent framework positions'). Some of these residues may membrane-proximal target epitopes of human c-MET were contribute to the binding and activity of the antibody to its designed as set out in the following Table 7: target. To construct a library that contains the murine CDRs and at TABLE 7 every framework position that differs from the chosen human 10 VH sequence both possible residues (the human and the Formats of I2C-based bispecific single chain antibodies directed maternal murine amino acid residue), degenerated oligo at membrane-proximal target epitopes of human C-MET nucleotides are synthesized. These oligonucleotides incorpo SEQID Formats of protein constructs rate at the differing positions the human residue with a prob (nucliprot) (N-> C) ability of 75% and the murine residue with a probability of 15 512,511 ME86H11HLXI2CHL 25%. For one human VH e.g. six of these oligonucleotides 526,525 ME62A12HLXI2CHL have to be synthesized that overlap in a terminal stretch of 540,539 ME63F2HLXI2CHL approximately 20 nucleotides. To this end every second 554,553 ME62D11HLXI2CHL primer is an antisense primer. Restriction sites within the 568,567 ME62C1 OHLXI2CHL oligonucleotides needed for later cloning are deleted. S82,581 ME62A4HLXI2CHL These primers may have a length of 60 to 90 nucleotides, depending on the number of primers that are needed to span Alternatively further constructs with different domain over the whole V sequence. arrangements can be generated according to standard proto These e.g. six primers are mixed in equal amounts (e.g. 1 ul colls. For expression in CHO cells the coding sequences of (i) of each primer (primer stocks 20 to 100 uM) to a 20 lul PCR 25 an N-terminal immunoglobulin heavy chain leader compris reaction) and added to a PCR mix consisting of PCR buffer, ing a start codon embedded within a Kozak consensus nucleotides and Taq polymerase. This mix is incubated at 94° sequence and (ii) a C-terminal His6-tag followed by a stop C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° codon are both attached in frame to the nucleotide sequence C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° encoding the bispecific single chain antibodies prior to inser C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. 30 Subsequently the product is run in an agarose gel electro tion of the resulting DNA-fragment as obtained by gene Syn phoresis and the product of a size from 200 to 400 base pairs thesis into the multiple cloning site of the expression vector isolated from the gel according to standard methods. pEF-DHFR (Raum et al. Cancer Immunol Immunother 50 This PCR product is then used as a template for a standard (2001) 141-150). A clone with sequence-verified nucleotide PCR reaction using primers that incorporate suitable N-ter 35 sequence is transfected into DHFR deficient CHO cells for minal and C-terminal cloning restriction sites. The DNA frag eukaryotic expression of the construct. Eukaryotic protein ment of the correct size (for a VH approximately 350 nucle expression in DHFR deficient CHO cells is performed as otides) is isolated by agarose gel electrophoresis according to described by Kaufmann R.J. (1990) Methods Enzymol. 185, standard methods. In this way sufficientVHDNA fragment is 537-566. Gene amplification of the construct is induced by amplified. This VH fragment is now a pool of VH fragments 40 increasing concentrations of methotrexate (MTX) to a final that have each one a different amount of human and murine concentration of up to 20 nM MTX. residues at the respective differing framework positions (pool 7.12 Expression and Purification of Bispecific Single Chain of humanized VH). The same procedure is performed for the Antibody Molecules Directed at Membrane-proximal Target VL region of the murine anti-c-MET scFv to a membrane Epitopes of Human c-MET proximal target epitope of human c-MET (pool of humanized 45 Bispecific single chain antibody molecules are expressed VL). in Chinese hamster ovary cells (CHO). Eukaryotic protein The pool of humanized VH is then combined with the pool expression in DHFR deficient CHO cells is performed as of humanized VL in the phage display vectorpComb3H5Bhis described by Kaufmann R.J. (1990) Methods Enzymol. 185, to form a library of functional scFVs from which—after dis 537-566. Gene amplification of the constructs is induced by play on filamentous phage—anti-c-MET binders to mem 50 addition of increasing concentrations of MTX up to final brane-proximal target epitopes of human c-MET are selected, concentrations of 20 nMMTX. After two passages of station screened, identified and confirmed as described above for the ary culture cell culture Supernatantis collected and used in the parental non-human (murine) anti-c-MET scFv. Single Subsequent experiments. To generate Supernatant for purifi clones are then analyzed for favorable properties and amino cation after two passages of Stationary culture the cells are acid sequence. Those Sclvs, which are closest in amino acid 55 grown in roller bottles with nucleoside-free HyO PF CHO sequence homology to human germline V-segments, are pre liquid soy medium (with 4.0 mM L-Glutamine with 0.1% ferred. Pluronic F-68; HyClone) for 7 days before harvest. The cells Human/humanized anti-c-MET scFvs to membrane-proxi are removed by centrifugation and the Supernatant containing mal target epitopes of human c-MET are converted into the expressed protein is stored at -20°C. Alternatively, con recombinant bispecific single chain antibodies and further 60 structs are transiently expressed in HEK 293 cells. Transfec characterized as follows. tion is performed with 293fectin reagent (Invitrogen, #12347 7.11 Generation of I2C-based Bispecific Single Chain Anti 019) according to the manufacturer's protocol. Furthermore bodies Directed at Membrane-proximal Target Epitopes of the constructs are alternatively expressed in transiently trans Human c-MET fected DHFR deficient CHO cells using for example Anti-c-MET scFvs to membrane-proximal target epitopes 65 FuGENER HD Transfection Reagent (Roche Diagnostics of human c-MET with favorable properties and amino acid GmbH, Cat. No. 04709691001) according to the manufactur sequence are converted into recombinant bispecific single er's protocol. US 9,260,522 B2 95 96 Akta R. Explorer System (GE Health Systems) and Uni expressing the bispecific antibody constructs. The cells are corn R Software are used for chromatography. Immobilized washed twice in PBS with 2% FCS and binding of the con metal affinity chromatography (“IMAC) is performed using struct is detected with a murine Penta His antibody (Qiagen; a Fractogel EMD ChelateR (Merck) which is loaded with diluted 1:20 in 50 ul PBS with 2% FCS). After washing, ZnCl2 according to the protocol provided by the manufac 5 bound anti His antibodies are detected with an Fc gamma turer. The column is equilibrated with buffer A (20 mM specific antibody (Dianova) conjugated to phycoerythrin, sodium phosphate buffer pH 7.2, 0.1 M NaCl) and the cell diluted 1:100 in PBS with 2% FCS. Supernatant of untrans culture supernatant (500 ml) is applied to the column (10 ml) fected cells is used as a negative control. at a flow rate of 3 ml/min. The column is washed with buffer Flow cytometry is performed on a FACS-Calibur appara A to remove unbound sample. Bound protein is eluted using 10 a two step gradient of buffer B (20 mM sodium phosphate tus, the CellOuest Software is used to acquire and analyze the buffer pH 7.2, 0.1 MNaCl, 0.5 MImidazole) according to the data (Becton Dickinson biosciences, Heidelberg). FACS following procedure: staining and measuring of the fluorescence intensity are per Step 1: 20% buffer B in 6 column volumes formed as described in Current Protocols in Immunology Step 2: 100% buffer B in 6 column volumes 15 (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley Eluted protein fractions from step 2 are pooled for further Interscience, 2002). purification. All chemicals are of research grade and pur Only those constructs that show bispecific binding to chased from Sigma (Deisenhofen) or Merck (Darmstadt). human CD3 as well as to human c-MET and neither bind to Gel filtration chromatography is performed on a HiLoad the murine c-MET antigen nor to the mutated murine c-MET 16/60 Superdex 200 prep grade column (GE/Amersham) antigen with human membrane-distal epitopes are selected equilibrated with Equi-buffer (25 mM Citrate, 200 mM for further use. Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow The bispecific binding of the single chain molecules listed rate 1 ml/min) are subjected to standard SDS-PAGE and above was clearly detectable as shown in FIG. 13. In the Western Blot for detection. Prior to purification, the columnis FACS analysis all constructs showed binding to human CD3 calibrated for molecular weight determination (molecular 25 and human c-MET compared to the negative control. Mem weight marker kit, Sigma MW GF-200). Protein concentra brane-proximal target binding of the single chain molecules tions are determined using OD280 nm. listed above was clearly detectable as shown in FIG. 14. In the Purified bispecific single chain antibody protein is ana FACS analysis all constructs showed binding to the mutated lyzed in SDS PAGE under reducing conditions performed human c-MET antigen with murine membrane-distal with pre-cast 4-12% Bis Tris gels (Invitrogen). Sample prepa 30 ration and application are performed according to the proto epitopes and did not show binding to the murine c-MET col provided by the manufacturer. The molecular weight is antigen and did also not show binding to the mutated murine determined with MultiMark protein standard (Invitrogen). c-MET antigen with human membrane-distal epitopes as The gel is stained with colloidal Coomassie (Invitrogen pro compared to the negative control. Expression of the c-MET tocol). The purity of the isolated protein is typically >95% as 35 antigens was confirmed by detection with an anti-FLAG M2 determined by SDS-PAGE. antibody as described herein. In the FACS analysis also The bispecific single chain antibody has a molecular shown in FIG. 14 CHO cells transfected with the murine weight of about 52 kDa under native conditions as determined c-MET antigen, the mutated murine c-MET antigen with by gel filtration in PBS. human membrane-distal epitopes and the mutated human Western Blot is performed using an Optitran?R) BA-S83 40 c-MET antigen with murine membrane-distal epitopes, membrane and the Invitrogen Blot Module according to the respectively, showed comparable expression of the antigens protocol provided by the manufacturer. The antibody used is as detected with the anti-FLAG antibody. directed against the His Tag (Penta His, Qiagen) and a Goat 7.14 Bioactivity of Bispecific Antibodies Directed at Mem anti-mouse Ig labeled with alkaline phosphatase (AP) brane-proximal Target Epitopes of Human c-MET (Sigma) is used as second step reagent, and BCIP/NBT 45 Bioactivity of generated bispecific single chain antibodies (Sigma) as substrate. Aband detected at 52 kD corresponds to is analyzed by chromium 51 (Cr) release in vitro cytotox purified bispecific single chain antibodies. icity assays using the CHO cells transfected with human 7.13 Flow Cytometric Binding Analysis of Bispecific Anti c-MET described in Example 7.17. To confirm that signifi bodies Directed at Membrane-proximal Target Epitopes of cant bioactivity is only recruited by binding to membrane Human c-MET 50 proximal target epitopes of human c-MET in addition— In order to test the functionality of bispecific antibody CHO cells expressing murine c-MET and the mutated murine constructs regarding the capability to bind to CD3 and to c-MET antigen with human membrane-distal epitopes, membrane-proximal target epitopes of human c-MET, respectively, both as described in Example 7.17 are used. As respectively, a FACS analysis is performed. For this purpose effector cells stimulated human CD4/CD56 depleted PBMC CHO cells transfected with human c-MET as described in 55 are used. Example 7.17 and the human CD3 positive T cell leukemia Stimulated human PBMC are obtained as follows: cell line HPB-ALL (DSMZ. Braunschweig, ACC483) are A Petri dish (145 mm diameter, Greiner bio-one GmbH, used. For confirmation of binding to membrane-proximal Kremsmünster) is coated with a commercially available anti target epitopes of human c-MET in addition CHO cells CD3 specific antibody (e.g. OKT3. Orthoclone) in a final expressing the murine c-MET antigen as described in 60 concentration of 1 lug/ml for 1 hour at 37°C. Unbound protein Example 7.17, CHO cells expressing the mutated human is removed by one washing step with PBS. The fresh PBMC c-MET antigen with murine membrane-distal epitopes as are isolated from peripheral blood (30-50 ml human blood) described in Example 7.17 and CHO cells expressing the by Ficoll gradient centrifugation according to standard pro mutated murine c-MET antigen with human membrane-dis tocols. 3-5x107 PBMC are added to the precoated petridishin tal epitopes as described in Example 7.17 are used. 200,000 65 120 ml of RPMI 1640 with stabilized glutamine/10% FCS/ cells of the respective cell lines are incubated for 30 min on IL-2 20 U/ml (Proleukin, Chiron) and stimulated for 2 days. ice with 50 ul of cell culture supernatant of transfected cells On the third day the cells are collected and washed once with US 9,260,522 B2 97 98 RPMI 1640. IL2 is added to a final concentration of 20U/ml - Continued and the cells are cultivated again for one day in the same cell ggtgg-3 reverse primer: culture medium as above. 5'-gacitt cattgaaatgcacaatcagg-3' SEO ID NO: 403 By depletion of CD4+ T cells and CD56+ NK cells accord ing to standard protocols CD8+ cytotoxic T lymphocytes 5. forward primer: 5'-tgctictaaatccagagctgg to c-3' SEO ID NO: 404 (CTLs) are enriched. reverse primer: Target cells are washed twice with PBS and labeled with 5'-gtcagataagaaattic cittagaatcc -3 SEQ ID NO: 405 11.1 MBq 'Cr in a final volume of 100 ul RPMI with 50% FCS for 60 minutes at 37° C. Subsequently the labeled target These PCRs generate five overlapping fragments, which cells are washed 3 times with 5 ml RPMI and then used in the 10 are isolated and sequenced according to standard protocols cytotoxicity assay. The assay is performed in a 96 well plate using the PCR primers, and thereby provide a portion of the in a total volume of 250 ul supplemented RPMI (as above) cDNA sequence coding macaque c-MET from codon 10 of with an E:T ratio of 10:1. Supernatant of cells expressing the the leader peptide to the last codon of the mature protein. To bispecific single chain antibody molecules in a final concen generate a construct for expression of macaque c-MET a tration of 50% and 20 threefold dilutions thereofare applied. 15 cDNA fragment is obtained by gene synthesis according to The assay time is 18 hours. Cytotoxicity is measured as standard protocols (the cDNA and amino acid sequence of the relative values of released chromium in the Supernatant construct is listed under SEQ ID Nos 406 and 407). In this related to the difference of maximum lysis (addition of Tri construct the coding sequence of macaque c-MET from ton-X) and spontaneous lysis (without effector cells). All amino acid 10 of the leader peptide to the last amino acid of measurements are done in quadruplicates. Measurement of the mature c-MET protein followed by a stop codon is fused chromium activity in the Supernatants is performed with a in frame to the coding sequence of the amino acids 1 to 9 of Wizard R. 3" gammacounter (Perkin Elmer Life Sciences the leader peptide of the human c-MET protein. The gene GmbH, Koln, Germany). Analysis of the experimental data is synthesis fragment is also designed as to contain a Kozak site performed with Prism 4 for Windows(R (version 4.02, Graph 25 for eukaryotic expression of the construct and restriction sites Pad Software Inc., San Diego, Calif., USA). Sigmoidal dose at the beginning and the end of the fragment containing the response curves typically have R values >0.90 as determined cDNA. The introduced restriction sites, EcoRI at the 5' end by the software. and SalI at the 3' end, are utilised in the following cloning Only those constructs showing potent recruitment of cyto procedures. Internal restriction sites are removed by silent toxic activity of effector T cells against target cells positive 30 mutation of the coding sequence in the gene synthesis frag for c-MET are selected for further use. ment. The gene synthesis fragment is cloned via EcoRI and As shown in FIG. 15 all of the generated bispecific anti Sal I into a plasmid designated pEF-DHFR (pEF-DHFR is bodies directed at membrane-proximal target epitopes of described in Raum et al. Cancer Immunol Immunother 50 human c-MET demonstrated cytotoxic activity against (2001) 141-150) following standard protocols. The afore human c-MET positive target cells elicited by stimulated 35 mentioned procedures are carried out according to standard human CD4/CD56 depleted PBMC but did not recruit sig protocols (Sambrook, Molecular Cloning: A Laboratory nificant cytotoxic activity against murine c-MET positive Manual, 3rd edition, Cold Spring Harbour Laboratory Press, target cells and target cells positive for the mutated murine Cold Spring Harbour, N.Y. (2001)). A clone with sequence c-MET antigen with human membrane-distal epitopes. verified nucleotide sequence is transfected into DHFR defi Thereby specific recruitment of cytotoxic activity viabinding 40 cient CHO cells for eukaryotic expression of the construct. to membrane-proximal target epitopes of human c-MET was Eukaryotic protein expression in DHFR deficient CHO cells confirmed. is performed as described by Kaufmann R.J. (1990) Methods 7.15 Generation of CHO Cells Expressing Macaque c-MET Enzymol. 185, 537-566. Gene amplification of the construct The cDNA sequence of macaque c-MET is obtained by a is induced by increasing concentrations of methotrexate set of 5 PCRs on cNA from macaque monkey Liver pre 45 (MTX) to a final concentration of up to 20 nM MTX. pared according to standard protocols. The following reaction 7.16 Flow Cytometric Analysis of Cross-species Specificity conditions: 1 cycle at 94° C. for 2 minutes followed by 40 of Bispecific Antibodies Directed at Membrane-proximal cycles with 94° C. for 1 minute, 56°C. for 1 minute and 72° Target Epitopes of Human c-MET C. for 3 minutes followed by a terminal cycle of 72°C. for 3 In order to test the cross-species specificity of bispecific minutes and the following primers are used: 50 antibodies directed at membrane-proximal target epitopes of human c-MET the capability of the constructs to bind to 1. forward primer: macaque c-MET and macaque CD3, respectively, is investi 5'-aggaatt caccatgaaggcc.ccc SEO ID NO: 396 gated by FACS analysis. For this purpose the macaque c-MET transfected CHO cells as described in example 7.17 reverse primer: 55 and the macaque T cell line 4119LnPx (kindly provided by 5' - ct coagaggcatttic catgtagg-3 SEO ID NO: 397 Prof Fickenscher, Hygiene Institute, Virology, Erlangen-Nu 2. forward primer: ernberg; published in Knappe A, et al., and Fickenscher H., 5'-gtccaaagggaaactictagatgc-3' SEO ID NO: 398 Blood 2000, 95, 3256-61) are used. 200,000 cells of the reverse primer: 5'-ggagacactggatgggagtic Cagg-3 SEO ID NO: 399 respective cell lines are incubated for 30 minonice with 50 ul 60 of cell culture Supernatant of transfected cells expressing the 3. forward primer: cross-species specific bispecific antibody constructs. The 5'-catcagagggit cqctt catgcagg-3' SEO ID NO: 4 OO cells are washed twice in PBS with 2% FCS and binding of reverse primer: the construct is detected with a murine Penta His antibody 5'-gctittggittitt cagggggagttgc-3' SEO ID NO: 401 (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After 4. forward primer: 65 washing, bound anti His antibodies are detected with an Fc 5'-atccaac caaatc.ttitt attagt SEO ID NO: 4 O2 gamma-specific antibody (Dianova) conjugated to phyco erythrin, diluted 1:100 in PBS with 2% FCS. Supernatant of US 9,260,522 B2 99 100 untransfected cells is used as a negative control. Flow cytom Gene amplification of the constructs is induced by increasing etry is performed on a FACS-Calibur apparatus, the concentrations of methotrexate (MTX) to a final concentra CellOuest software is used to acquire and analyze the data tion of up to 20 nM MTX. (Becton Dickinson biosciences, Heidelberg). FACS staining and measuring of the fluorescence intensity are performed as 5 8. Generation of Bispecific Single Chain Antibodies described in Current Protocols in Immunology (Coligan, Directed at Membrane-proximal Target Epitopes of Kruisbeek, Margulies. Shevach and Strober, Wiley-Inter Human Endosialin science, 2002). The cross-species specific binding of the single chain mol 8.1 Generation of CHO Cells Expressing Human Endosialin ecules listed above was clearly detectable as shown in FIG. 10 The coding sequence of human Endosialin as published in 13. In the FACS analysis all constructs showed binding to GenBank (Accession number NM 020404) is obtained by macaque CD3 and macaque c-MET compared to the negative gene synthesis according to standard protocols. The gene control. synthesis fragment is designed as to contain first a Kozak site 7.17 Generation of CHO Cells with Enhanced Expression of for eukaryotic expression of the construct followed by the Extracellular Domains of Human c-MET, Macaque c-MET, 15 coding sequence of the human Endosialin protein, followed Murine c-MET, Mutated Murine c-MET with Human Mem inframe by the coding sequence of a Flagtaganda stop codon brane-distal Epitopes and Mutated Human c-MET with (the cDNA and amino acid sequence of the construct is listed Murine Membrane-distal Epitopes, Respectively under SEQ ID Nos 408 and 409). The gene synthesis frag The modified coding sequences of human c-MET, ment is also designed as to introduce restriction sites at the macaque c-MET, murine c-MET, mutated murine c-MET beginning and at the end of the fragment. The introduced with human membrane-distal epitopes and mutated human restriction sites, EcoRI at the 5' end and Xbal at the 3' end, are c-MET with murine membrane-distal epitopes as described utilized in the following cloning procedures. The gene Syn above are used for the construction of artificial cDNA thesis fragment is cloned via EcoRI and Xbal into a plasmid sequences encoding fusion proteins of the extracellular designated pEF-DHFR (pEF-DHFR is described in Raum et domains of human c-MET, macaque c-MET, murine c-MET, 25 al. Cancer Immunol Immunother 50 (2001) 141-150) follow mutated murine c-MET with human membrane-distal ing standard protocols. The aforementioned procedures are epitopes and mutated human c-MET with murine membrane carried out according to standard protocols (Sambrook, distal epitopes, respectively, with a truncated variant of Molecular Cloning: A Laboratory Manual, 3rd edition, Cold human EpCAM. To generate constructs for expression of Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. these c-MET fusion proteins cloNA fragments are obtained 30 (2001)). A clone with sequence-verified nucleotide sequence by gene synthesis according to standard protocols (the cDNA is transfected into DHFR deficient CHO cells for eukaryotic and amino acid sequences of the constructs are listed under expression of the construct. Eukaryotic protein expression in SEQID Nos 489 and 490 for human c-MET, 491 and 492 for DHFR deficient CHO cells is performed as described by macaque c-MET, 493 and 494 for murine c-MET, 495 and Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. 496 formutated murine c-MET with human membrane-distal 35 Gene amplification of the construct is induced by increasing epitopes and 497 and 498 for mutated human c-MET with concentrations of methotrexate (MTX) to a final concentra murine membrane-distal epitopes). The gene synthesis frag tion of up to 20 nM MTX. ments are designed as to contain first a Kozak site for eukary 8.2 Generation of a Soluble Human Endosialin Fusion Pro otic expression of the construct, followed by the coding tein sequence of a 19 amino acid immunoglobulin leader peptide, 40 The coding sequence of human Endosialin as described in followed in frame by the coding sequence of a FLAG tag Example 8.1 is used for the construction of an artificial cDNA (only in the case of the murine and the mutated human and sequence encoding a soluble fusion protein of human Endo mutated murine constructs), followed in frame by the coding sialin and murine IgG1 Fc. To generate a construct for expres sequence of the extracellular domains of human c-MET, sion of the soluble human Endosialin fusion protein a cDNA macaque c-MET, murine c-MET, mutated murine c-MET 45 fragment is obtained by gene synthesis according to standard with human membrane-distal epitopes and mutated human protocols (the cDNA and amino acid sequence of the con c-MET with murine membrane-distal epitopes, respectively, struct is listed under SEQ ID Nos 410 and 411). The gene followed in frame by the coding sequence of an artificial synthesis fragment is designed as to contain first a Kozak site Ser-Gly-Ser-Gly-linker, followed in frame by the coding for eukaryotic expression of the construct followed by the sequence of the transmembrane domain and intracellular 50 coding sequence of the human Endosialin protein from amino domain of human EpCAM (as published in GenBank; Acces acid 1 to 685 corresponding to the signal peptide and extra sion number NM 002354; amino acids 266 to 314 as cellular domains of human Endosialin, followed in frame by counted from the start codon except for a point mutation at the coding sequence of an artificial Thr-Gly-Ser-linker, position 279 with isoleucine instead of valine) and a stop followed in frame by the coding sequence of the hinge region codon. The gene synthesis fragments are also designed as to 55 and Fc gamma portion of murine IgG1, followed in frame by introduce restriction sites at the beginning and at the end of the coding sequence of a 6 histidine tag and a stop codon. The the fragment. The introduced restriction sites, EcoRI at the 5' gene synthesis fragment is also designed as to introduce end and SalI at the 3' end, are utilized in the following cloning restriction sites at the beginning and at the end of the frag procedures. The gene synthesis fragments are cloned via ment. The introduced restriction sites, EcoRI at the 5' end and EcoRI and SalI into a plasmid designated pEF-DHFR (pEF 60 Xbal at the 3' end, are utilized in the following cloning pro DHFR is described in Raum et al. Cancer Immunol Immu cedures. The gene synthesis fragment is cloned via EcoRI and nother 50 (2001) 141-150) following standard protocols. Xbal into a plasmid designated pEF-DHFR (pEF-DHFR is Clones with sequence-verified nucleotide sequence are trans described in Raum et al. Cancer Immunol Immunother 50 fected into DHFR deficient CHO cells for eukaryotic expres (2001) 141-150) following standard protocols. The afore sion of the constructs. Eukaryotic protein expression in 65 mentioned procedures are all carried out according to stan DHFR deficient CHO cells is performed as described by dard protocols (Sambrook, Molecular Cloning: A Laboratory Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. Manual, 3rd edition, Cold Spring Harbour Laboratory Press, US 9,260,522 B2 101 102 Cold Spring Harbour, N.Y. (2001)). A clone with sequence synthesis fragment is designed as to contain the coding verified nucleotide sequence is transfected into DHFR defi sequence of an immunoglobulin heavy chain leader compris cient CHO cells for eukaryotic expression of the construct. ing a start codon embedded within a Kozak consensus Eukaryotic protein expression in DHFR deficient CHO cells sequence, followed in frame by the coding sequence of a is performed as described by Kaufmann R.J. (1990) Methods FLAG tag, followed in frame by the coding sequence of the Enzymol. 185, 537-566. Gene amplification of the construct C-type lectin domain of mature murine endosialin followed in is induced by increasing concentrations of methotrexate frame by human endosialin from the Sushi/SCR/CCP domain (MTX) to a final concentration of up to 20 nM MTX. After to the stop codon (the cDNA and amino acid sequence of the two passages of stationary culture the cells are grown in roller construct is listed under SEQID Nos 414 and 415). The gene bottles with nucleoside-free HyOPF CHO liquid soy medium 10 synthesis fragment is also designed as to introduce restriction (with 4.0 mM L-Glutamine with 0.1% Pluronic F-68; sites at the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA HyClone) for 7 days before harvest. The cells are removed by fragment for cloning into the mammalian cell expression centrifugation and the Supernatant containing the expressed vector pEF-DHFR (pEF-DHFR is described in Raum et al. protein is stored at -20°C. Alternatively a clone of the expres Cancer Immunol Immunother 50 (2001) 141-150). The afore sion plasmid with sequence-verified nucleotide sequence is 15 mentioned procedures are carried out according to standard used for transfection and protein expression in the FreeStyle protocols (Sambrook, Molecular Cloning: A Laboratory 293 Expression System (Invitrogen GmbH, Karlsruhe, Ger Manual, 3rd edition, Cold Spring Harbour Laboratory Press, many) according to the manufacturer's protocol. Supernatant Cold Spring Harbour, N.Y. (2001)). A clone with sequence containing the expressed protein is obtained, cells are verified nucleotide sequence is transfected into DHFR defi removed by centrifugation and the Supernatant is stored at cient CHO cells for eukaryotic expression of the construct. -20° C. Eukaryotic protein expression in DHFR deficient CHO cells For purification of the soluble human Endosialin fusion is performed as described by Kaufmann R.J. (1990) Methods protein a goat anti-mouse Fc affinity column is prepared Enzymol. 185, 537-566. Gene amplification for increased according to standard protocols using a commercially avail antigen expression is induced by increasing concentrations of able affinity purified goat anti-mouse IgG Fc fragment spe 25 methotrexate (MTX) to a final concentration of up to 20 nM cific antibody with minimal cross-reaction to human, bovine MTX. and horse serum proteins (Jackson ImmunoResearch Europe 8.5 Generation of CHO Cells Expressing a Mutated Murine Ltd.). Using this affinity column the fusion protein is isolated Endosialin Antigen with Human Membrane-distal Epitopes out of cell culture supernatant on an Akta Explorer System The coding sequence of a mutated murine Endosialin anti (GE Amersham) and eluted by citric acid. The eluate is neu 30 gen with human membrane-distal epitopes is obtained by tralized and concentrated. gene synthesis according to standard protocols. The gene 8.3 Generation of CHO Cells Expressing the Murine Endo synthesis fragment is designed as to contain the coding sialin Antigen sequence of an immunoglobulin heavy chain leader compris The sequence of murine Endosialin (NM 054042 Mus ing a start codon embedded within a Kozak consensus musculus Endosialin antigen, endosialin (Cd248), mRNA: 35 sequence, followed in frame by the coding sequence of a National Center for Biotechnology Information, http://ww FLAG tag, followed in frame by the coding sequence of the w.ncbi.nlm.nih.gov/entrez) is used to obtain a synthetic C-type lectin domain of mature human endosialin followed in cDNA molecule by gene synthesis according to standard frame by murine endosialin from the Sushi/SCR/CCP protocols. The gene synthesis fragment is designed as to domain to the stop codon (the cDNA and amino acid contain the coding sequence of an immunoglobulin heavy 40 sequence of the construct is listed under SEQID Nos 416 and chain leader comprising a start codon embedded within a 417). The gene synthesis fragment is also designed as to Kozak consensus sequence, followed in frame by the coding introduce restriction sites at the 5' end (EcoRI) and at the 3' sequence of a FLAG tag, followed in frame by the complete end (Sal I) of the cDNA fragment for cloning into the mam coding sequence of mature murine endosialin (the cDNA and malian cell expression vector pEF-DHFR (pEF-DHFR is amino acid sequence of the construct is listed under SEQID 45 described in Raum et al. Cancer Immunol Immunother 50 Nos 412 and 413). The gene synthesis fragment is also (2001) 141-150). Internal restriction sites are removed by designed as to introduce restriction sites at the 5' end (EcoRI) silent mutation of the coding sequence in the gene synthesis and at the 3' end (Sal I) of the cDNA fragment for cloning into fragment. The aforementioned procedures are carried out the mammalian cell expression vector pEF-DHFR (pEF according to standard protocols (Sambrook, Molecular Clon DHFR is described in Raum et al. Cancer Immunol Immu 50 ing: A Laboratory Manual, 3rd edition, Cold Spring Harbour nother 50 (2001) 141-150). The aforementioned procedures Laboratory Press, Cold Spring Harbour, N.Y. (2001)). A are carried out according to standard protocols (Sambrook, clone with sequence-verified nucleotide sequence is trans Molecular Cloning: A Laboratory Manual, 3rd edition, Cold fected into DHFR deficient CHO cells for eukaryotic expres Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. sion of the construct. Eukaryotic protein expression in DHFR (2001)). A clone with sequence-verified nucleotide sequence 55 deficient CHO cells is performed as described by Kaufmann is transfected into DHFR deficient CHO cells for eukaryotic R. J. (1990) Methods Enzymol. 185, 537-566. Gene amplifi expression of the construct. Eukaryotic protein expression in cation for increased antigen expression is induced by increas DHFR deficient CHO cells is performed as described by ing concentrations of methotrexate (MTX) to a final concen Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. tration of up to 20 nM MTX. Gene amplification for increased antigen expression is 60 8.6 Immunization of Mice Using a Soluble Human Endosia induced by increasing concentrations of methotrexate (MTX) lin Fusion Protein to a final concentration of up to 20 nM MTX. Twelve weeks old F1 mice from BALB? cx C57BL/6 cross 8.4 Generation of CHO Cells Expressing a Mutated Human ings are immunized with the soluble human Endosialin fusion Endosialin Antigen with Murine Membrane-distal Epitopes protein as described in Example 8.2. To this end for each The coding sequence of a mutated human Endosialin anti 65 animal 40 ug of the soluble human Endosialin fusion protein gen with murine membrane-distal epitopes is obtained by are mixed with 10 nmol of a thioate-modified CpG-Oligo gene synthesis according to standard protocols. The gene nucleotide (5'-tccatgacgttcctgatgct-3') in 300 ul PBS and are US 9,260,522 B2 103 104 injected intraperitoneally. Mice receive booster immuniza and displays the corresponding scFV-proteinas a translational tions after 21, 42 and optionally 63 days in the same way. Ten fusion to phage coat protein III. This pool of phages display days after the first booster immunization, blood samples are ing the antibody library is later used for the selection of taken and antibody serum titers against human Endosialin are antigen binding entities. tested by flow cytometry according to standard protocols. To 5 8.8 Phage Display Based Selection of Membrane-proximal this end 200,000 cells of the human Endosialin transfected Target Binders on CHO Cells Expressing the Mutated Human CHO cells as described in Example 8.1 are incubated for 30 Endosialin Antigen with Murine Membrane-distal Epitopes min on ice with 50 ul of serum of the immunized animals The phage library carrying the cloned schv-repertoire is diluted 1:1000 in PBS with 2% FCS. The cells are washed harvested from the respective culture supernatant by twice in PBS with 2% FCS and binding of serum antibodies is 10 detected with an mouse Fc gamma-specific antibody (Di PEG8000/NaCl precipitation and centrifugation. Approxi anova) conjugated to phycoerythrin, diluted 1:100 in PBS mately 10' to 10' sclv phage particles are resuspended in with 2% FCS. Serum of the animals obtained prior to immu 0.4 ml of PBS/0.1% BSA and incubated with 10 to 107 CHO nization is used as a negative control. cells expressing the mutated human Endosialin antigen with Flow cytometry is performed on a FACS-Calibur appara- 15 murine membrane-distal epitopes as described in example 8.4 tus, the CellOuest Software is used to acquire and analyze the for 1 hour on ice under slow agitation. These CHO cells are data (Becton Dickinson biosciences, Heidelberg). FACS grown beforehand, harvested by centrifugation, washed in staining and measuring of the fluorescence intensity are per PBS and resuspended in PBS/1% FCS (containing Na Azide). formed as described in Current Protocols in Immunology schv phage which do not specifically bind to the CHO cells (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley- 20 are eliminated by up to five washing steps with PBS/1% FCS Interscience, 2002). (containing Na Azide). After washing, binding entities are Animals demonstrating significant serum reactivity eluted from the cells by resuspending the cells in HCl-glycine against human Endosialin as determined by the FACS analy pH 2.2 (10 min incubation with Subsequent Vortexing) and sis are used in the Subsequent experiment. after neutralization with 2M Tris pH 12, the eluate is used for 8.7 Generation of an Immune Murine Antibody schv Library: 25 infection of a fresh uninfected E. coli XL1 Blue culture Construction of a Combinatorial Antibody Library and Phage (OD600>0.5). The E. coli culture containing E. coli cells Display Successfully transduced with a phagemid copy, encoding a Three days after the last booster immunization spleen cells murine ScFV-fragment, are again selected for carbenicillin of reactive animals are harvested for the preparation of total resistance and subsequently infected with VCMS 13 helper RNA according to standard protocols. 30 phage to start the second round of antibody display and in A library of murine immunoglobulin (Ig) light chain vitro selection. Typically a total of 4 to 5 rounds of selections (kappa) variable region (VK) and Ig heavy chain variable are carried out. region (VH) DNA-fragments is constructed by RT-PCR on 8.9 Screening for Membrane-proximal Target Binders on murine spleen RNA using VK- and VH specific primers. CHO Cells Expressing the Human Endosialin Antigen, the cDNA is synthesized according to standard protocols, see 35 Murine Endosialin Antigen and the Mutated Murine Endo example 2.7. sialin Antigen with Human Membrane-distal Epitopes 450 ng of the kappa light chain fragments (SacI-Spel Plasmid DNA corresponding to 4 and 5 rounds of panning digested) are ligated with 1400 ng of the phagemid is isolated from E. coli cultures after selection. For the pro pComb3H5Bhis (SacI-Spel digested; large fragment). The duction of soluble schv-protein, VH-VL-DNA fragments are resulting combinatorial antibody library is then transformed 40 excised from the plasmids (XhoI-SpeI). These fragments are into 300 ul of electrocompetent Escherichia coli XL 1 Blue cloned via the same restriction sites in the plasmid cells by electroporation (2.5 kV, 0.2 cm gap cuvette, 25uFD, pComb3H5BFlag? His differing from the original 200 Ohm, Biorad gene-pulser) resulting in a library size of pComb3H5BHis in that the expression construct (e.g. scFv) more than 107 independent clones. After one hour of pheno includes a Flag-tag (TGDYKDDDDK) between the scFv and type expression, positive transformants are selected for car- 45 the His6-tag and the additional phage proteins are deleted. benicillin resistance encoded by the pComb3H5BHis vector After ligation, each pool (different rounds of panning) of in 100 ml of liquid super broth (SB)-culture over night. Cells plasmid DNA is transformed into 100 ul heat shock compe are then harvested by centrifugation and plasmid preparation tent E. coli TG1 or XLI blue and plated onto carbenicillin is carried out using a commercially available plasmid prepa LB-agar. Single colonies are picked into 100 ul of LB carb ration kit (Qiagen). 50 (LB with 50 lug/ml carbenicillin). 2800 ng of this plasmid-DNA containing the VK-library After induction with 1 mMIPTG. E. coli transformed with (XhoI-BstEII digested; large fragment) are ligated with 900 pComb3H5BFlag? His containing a VL- and VH-segment ng of the heavy chain V-fragments (XhoI-BstEII digested) produce soluble scFv in sufficient amounts. Due to a suitable and again transformed into two 300 ulaliquots of electrocom signal sequence, the Sclv is exported into the periplasma petent E. coli XL 1 Blue cells by electroporation (2.5 kV, 0.2 55 where it folds into a functional conformation. cm gap cuvette, 25uFD, 200Ohm) resulting in a totalVH-VK Single E. coli bacterial colonies from the transformation schv (single chain variable fragment) library size of more plates are picked for periplasmic Small scale preparations and than 107 independent clones. grown in SB-medium (e.g. 10 ml) supplemented with 20 mM After phenotype expression and slow adaptation to carbe MgCl, and carbenicillin 50 ug/ml (and re-dissolved in PBS nicillin, the E. coli cells containing the antibody library are 60 (e.g. 1 ml) after harvesting. A temperature shock is applied by transferred into SB-Carbenicillin (SB with 50 lug/mL carbe four rounds of freezing at -70° C. and thawing at 37° C. nicillin) selection medium. The E. coli cells containing the whereby the outer membrane of the bacteria is destroyed and antibody library are then infected with an infectious dose of the soluble periplasmic proteins including the Sclvs are 10' particles of helper phage VCSM13 resulting in the pro released into the supernatant. After elimination of intact cells duction and secretion of filamentous M13 phage, wherein 65 and cell-debris by centrifugation, the Supernatant containing each phage particle contains single stranded the murine anti-human Endosialin-SclvS is collected and pComb3H5BHis-DNA encoding a murine schv-fragment used for further examination. US 9,260,522 B2 105 106 Screening of the isolated schvs for membrane-proximal standard methods. In this way sufficientVHDNA fragment is target binders is performed by flow cytometry on CHO cells amplified. This VH fragment is now a pool of VH fragments expressing the human Endosialin antigen as described in that have each one a different amount of human and murine Example 8.1, the murine Endosialin antigen as described in residues at the respective differing framework positions (pool Example 8.3 and the mutated murine Endosialin antigen with 5 of humanized VH). The same procedure is performed for the human membrane-distal epitopes as described in Example VL region of the murine anti-Endosialin schv to a membrane 85. proximal target epitope of human Endosialin (pool of human For flow cytometry 2.5x10 cells of the respective cell lines ized VL). The pool of humanized VH is then combined with are incubated with 50 ul supernatant. The binding of the the pool of humanized VL in the phage display vector constructs is detected with an anti-His antibody (Penta-His 10 pComb3H5Bhis to form a library of functional sclvs from Antibody, BSA free, Qiagen GmbH, Hilden, FRG) at 2 ug/ml which—after display on filamentous phage—anti-Endosialin in 50 lul PBS with 2% FCS. As a second step reagent a binders to membrane-proximal target epitopes of human R-Phycoerythrin-conjugated affinity purified F(ab')2 frag Endosialinare selected, screened, identified and confirmed as ment, goat anti-mouse IgG (Fc-gamma fragment specific), described above for the parental non-human (murine) anti diluted 1:100 in 50 ul PBS with 2% FCS (Dianova, Hamburg, 15 Endosialin schv. Single clones are then analyzed for favor FRG) is used. The samples are measured on a FACSscan (BD able properties and amino acid sequence. Those Sclvs, which biosciences, Heidelberg, FRG). are closest in amino acid sequence homology to human ger Only constructs which show binding to CHO cells express mline V-segments, are preferred. ing the human Endosialin antigen and do not show binding to Human/humanized anti-Endosialin Scvs to membrane CHO cells expressing the murine Endosialin antigen and also proximal target epitopes of human Endosialin are converted do not show binding to CHO cells expressing the mutated into recombinant bispecific single chain antibodies and fur murine Endosialin antigen with human membrane-distal ther characterized as follows. epitopes are selected for further use. 8.11 Generation of I2C-based Bispecific Single Chain Anti 8.10 Generation of Human/Humanized Equivalents of Non bodies Directed at Membrane-proximal Target Epitopes of human schvs to Membrane-proximal Target Epitopes of 25 Human Endosialin Human Endosialin Anti-Endosialin ScFVS to membrane-proximal target The VH region of a murine anti-Endosialin schv to a mem epitopes of human Endosialin with favorable properties and brane-proximal target epitope of human Endosialin is aligned amino acid sequence are converted into recombinant bispe against human antibody germline amino acid sequences. The cific single chain antibodies by joining them via a Gly-Ser human antibody germline VH sequence is chosen which has 30 linker with the CD3 specific sclv I2C (SEQID NO: 185) to the closest homology to the non-human VH and a direct result in constructs with the domain arrangement VH alignment of the two amino acid sequences is performed. ahn-(Gly-Seri)s-VLEast-Ser Gly-Ser-VHops There are a number of framework residues of the non-human (Gly-Ser)-VL. Alternatively further constructs with dif VH that differ from the human VH framework regions (“dif ferent domain arrangements can be generated according to ferent framework positions'). Some of these residues may 35 standard protocols. For expression in CHO cells the coding contribute to the binding and activity of the antibody to its sequences of (i) an N-terminal immunoglobulin heavy chain target. leader comprising a start codon embedded within a Kozak To construct a library that contains the murine CDRs and at consensus sequence and (ii) a C-terminal His6-tag followed every framework position that differs from the chosen human by a stop codon are both attached in frame to the nucleotide VH sequence both possible residues (the human and the 40 sequence encoding the bispecific single chain antibodies maternal murine amino acid residue), degenerated oligo prior to insertion of the resulting DNA-fragment as obtained nucleotides are synthesized. These oligonucleotides incorpo by gene synthesis into the multiple cloning site of the expres rate at the differing positions the human residue with a prob sion vectorpFF-DHFR (Raum et al. Cancer Immunol Immu ability of 75% and the murine residue with a probability of nother 50 (2001) 141-150). A clone with sequence-verified 25%. For one human VH e.g. six of these oligonucleotides 45 nucleotide sequence is transfected into DHFR deficient CHO have to be synthesized that overlap in a terminal stretch of cells for eukaryotic expression of the construct. Eukaryotic approximately 20 nucleotides. To this end every second protein expression in DHFR deficient CHO cells is performed primer is an antisense primer. Restriction sites within the as described by Kaufmann R. J. (1990) Methods Enzymol. oligonucleotides needed for later cloning are deleted. 185, 537-566. Gene amplification of the construct is induced These primers may have a length of 60 to 90 nucleotides, 50 by increasing concentrations of methotrexate (MTX) to a depending on the number of primers that are needed to span final concentration of up to 20 nM MTX. over the whole V sequence. 8.12 Expression and Purification of Bispecific Single Chain These e.g. six primers are mixed in equal amounts (e.g. 1 ul Antibody Molecules Directed at Membrane-proximal Target of each primer (primer stocks 20 to 100 uM) to a 20 lul PCR Epitopes of Human Endosialin reaction) and added to a PCR mix consisting of PCR buffer, 55 Bispecific single chain antibody molecules are expressed nucleotides and Taq polymerase. This mix is incubated at 94° in Chinese hamster ovary cells (CHO). Eukaryotic protein C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° expression in DHFR deficient CHO cells is performed as C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° described by Kaufmann R.J. (1990) Methods Enzymol. 185, C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. 537-566. Gene amplification of the constructs is induced by Subsequently the product is run in an agarose gel electro 60 addition of increasing concentrations of MTX up to final phoresis and the product of a size from 200 to 400 base pairs concentrations of 20 nMMTX. After two passages of station isolated from the gel according to standard methods. ary culture cell culture Supernatantis collected and used in the This PCR product is then used as a template for a standard Subsequent experiments. To generate Supernatant for purifi PCR reaction using primers that incorporate suitable N-ter cation after two passages of Stationary culture the cells are minal and C-terminal cloning restriction sites. The DNA frag 65 grown in roller bottles with nucleoside-free HyO PF CHO ment of the correct size (for a VH approximately 350 nucle liquid soy medium (with 4.0 mM L-Glutamine with 0.1% otides) is isolated by agarose gel electrophoresis according to Pluronic F-68; HyClone) for 7 days before harvest. The cells US 9,260,522 B2 107 108 are removed by centrifugation and the Supernatant containing For confirmation of binding to membrane-proximal target the expressed protein is stored at -20°C. Alternatively, con epitopes of human Endosialin in addition CHO cells structs are transiently expressed in HEK 293 cells. Transfec expressing the murine Endosialin antigen as described in tion is performed with 293fectin reagent (Invitrogen, #12347 Example 8.3 and CHO cells expressing the mutated murine 019) according to the manufacturer's protocol. Furthermore Endosialin antigen with human membrane-distal epitopes as the constructs are alternatively expressed in transiently trans described in Example 8.5 are used. 200,000 cells of the fected DHFR deficient CHO cells using for example respective cell lines are incubated for 30 minonice with 50 ul FuGENER HD Transfection Reagent (Roche Diagnostics of cell culture Supernatant of transfected cells expressing the GmbH, Cat. No. 04709691001) according to the manufactur bispecific antibody constructs. The cells are washed twice in er's protocol. 10 PBS with 2% FCS and binding of the construct is detected Akta R. Explorer System (GE Health Systems) and Uni with a murine Penta His antibody (Qiagen; diluted 1:20 in 50 corn R Software are used for chromatography. Immobilized ul PBS with 2% FCS). After washing, bound anti H is anti metal affinity chromatography (“IMAC) is performed using bodies are detected with an Fc gamma-specific antibody (Di a Fractogel EMD ChelateR (Merck) which is loaded with anova) conjugated to phycoerythrin, diluted 1:100 in PBS ZnCl2 according to the protocol provided by the manufac 15 with 2% FCS. Supernatant of untransfected cells is used as a turer. The column is equilibrated with buffer A (20 mM negative control. Flow cytometry is performed on a FACS sodium phosphate buffer pH 7.2, 0.1 M NaCl) and the cell Calibur apparatus, the CellOuest software is used to acquire culture supernatant (500 ml) is applied to the column (10 ml) and analyze the data (Becton Dickinson biosciences, Heidel at a flow rate of 3 ml/min. The column is washed with buffer berg). FACS staining and measuring of the fluorescence A to remove unbound sample. Bound protein is eluted using intensity are performed as described in Current Protocols in a two step gradient of buffer B (20 mM sodium phosphate Immunology (Coligan, Kruisbeek, Margulies. Shevach and buffer pH 7.2, 0.1 MNaCl, 0.5 MImidazole) according to the Strober, Wiley-Interscience, 2002). following procedure: Only those constructs that show bispecific binding to Step 1: 20% buffer B in 6 column volumes human CD3 as well as to human Endosialin and neither bind Step 2: 100% buffer B in 6 column volumes 25 to the murine Endosialin antigen nor to the mutated murine Eluted protein fractions from step 2 are pooled for further Endosialin antigen with human membrane-distal epitopes are purification. All chemicals are of research grade and pur selected for further use. chased from Sigma (Deisenhofen) or Merck (Darmstadt). 8.14 Bioactivity of Bispecific Antibodies Directed at Mem Gel filtration chromatography is performed on a HiLoad brane-proximal Target Epitopes of Human Endosialin 16/60 Superdex 200 prep grade column (GE/Amersham) 30 Bioactivity of generated bispecific single chain antibodies equilibrated with Equi-buffer (25 mM Citrate, 200 mM is analyzed by chromium 51 (Cr) release in vitro cytotox Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow icity assays using the CHO cells transfected with human rate 1 ml/min) are subjected to standard SDS-PAGE and Endosialin described in Example 8.1. As effector cells stimu Western Blot for detection. Prior to purification, the columnis lated human CD4/CD56 depleted PBMC are used. calibrated for molecular weight determination (molecular 35 Stimulated human PBMC are obtained as follows: weight marker kit, Sigma MW GF-200). Protein concentra A Petri dish (145 mm diameter, Greiner bio-one GmbH, tions are determined using OD280 nm. Kremsmunster) is coated with a commercially available anti Purified bispecific single chain antibody protein is ana CD3 specific antibody (e.g. OKT3. Orthoclone) in a final lyzed in SDS PAGE under reducing conditions performed concentration of 1 lug/ml for 1 hour at 37°C. Unbound protein with pre-cast 4-12% Bis Tris gels (Invitrogen). Sample prepa 40 is removed by one washing step with PBS. The fresh PBMC ration and application are performed according to the proto are isolated from peripheral blood (30-50 ml human blood) col provided by the manufacturer. The molecular weight is by Ficoll gradient centrifugation according to standard pro determined with MultiMark protein standard (Invitrogen). tocols. 3-5x107 PBMC are added to the precoated petridishin The gel is stained with colloidal Coomassie (Invitrogen pro 120 ml of RPMI 1640 with stabilized glutamine/10% FCS/ tocol). The purity of the isolated protein is typically >95% as 45 IL-2 20 U/ml (Proleukin, Chiron) and stimulated for 2 days. determined by SDS-PAGE. On the third day the cells are collected and washed once with The bispecific single chain antibody has a molecular RPMI 1640. IL2 is added to a final concentration of 20U/ml weight of about 52 kDa under native conditions as determined and the cells are cultivated again for one day in the same cell by gel filtration in PBS. culture medium as above. Western Blot is performed using an Optitran?R) BA-S83 50 By depletion of CD4+ T cells and CD56+ NK cells accord membrane and the Invitrogen Blot Module according to the ing to standard protocols CD8+ cytotoxic T lymphocytes protocol provided by the manufacturer. The antibody used is (CTLs) are enriched. directed against the His Tag (Penta His, Qiagen) and a Goat Target cells are washed twice with PBS and labeled with anti-mouse Ig labeled with alkaline phosphatase (AP) 11.1 MBq 'Cr in a final volume of 100 ul RPMI with 50% (Sigma) is used as second step reagent, and BCIP/NBT 55 FCS for 60 minutes at 37° C. Subsequently the labeled target (Sigma) as substrate. Aband detected at 52 kD corresponds to cells are washed 3 times with 5 ml RPMI and then used in the purified bispecific single chain antibodies. cytotoxicity assay. The assay is performed in a 96 well plate 8.13 Flow Cytometric Binding Analysis of Bispecific Anti in a total volume of 250 ul supplemented RPMI (as above) bodies Directed at Membrane-proximal Target Epitopes of with an E:T ratio of 10:1. 1 ug/ml of purified bispecific single Human Endosialin 60 chain antibody molecule and 20 threefold dilutions thereof In order to test the functionality of bispecific antibody are applied. The assay time is 18 hours. Cytotoxicity is mea constructs regarding the capability to bind to CD3 and to Sured as relative values of released chromium in the Superna membrane-proximal target epitopes of human Endosialin, tant related to the difference of maximum lysis (addition of respectively, a FACS analysis is performed. For this purpose Triton-X) and spontaneous lysis (without effector cells). All CHO cells transfected with human Endosialinas described in 65 measurements are done in quadruplicates. Measurement of Example 8.1 and the human CD3 positive T cell leukemia cell chromium activity in the Supernatants is performed with a line HPB-ALL (DSMZ. Braunschweig, ACC483) are used. Wizard R. 3" gammacounter (Perkin Elmer Life Sciences US 9,260,522 B2 109 110 GmbH, Koln, Germany). Analysis of the experimental data is carried out according to standard protocols (Sambrook, performed with Prism 4 for Windows(R (version 4.02, Graph Molecular Cloning: A Laboratory Manual, 3rd edition, Cold Pad Software Inc., San Diego, Calif., USA). Sigmoidal dose Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. responsecurves typically have R values >0.90 as determined (2001)). A clone with sequence-verified nucleotide sequence by the software. EC50 values calculated by the analysis pro 5 is transfected into DHFR deficient CHO cells for eukaryotic gram are used for comparison of bioactivity. expression of the construct. Eukaryotic protein expression in Only those constructs showing potent recruitment of cyto DHFR deficient CHO cells is performed as described by toxic activity of effector T cells against target cells positive Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. for Endosialin are selected for further use. Gene amplification of the construct is induced by increasing 8.15 Generation of CHO Cells Expressing Macaque Endo 10 concentrations of methotrexate (MTX) to a final concentra sialin tion of up to 20 nM MTX. The cDNA sequence of macaque Endosialin is obtained by 8.16 Flow Cytometric Analysis of Cross-species Specificity a set of 2 PCRs on cDNA from macaque monkey colon of Bispecific Antibodies Directed at Membrane-proximal prepared according to standard protocols. The following Target Epitopes of Human Endosialin reaction conditions: 1 cycle at 95°C. for 5 minutes followed 15 In order to test the cross-species specificity of bispecific by 40 cycles with 95°C. for 45 seconds, 50° C. for 45 seconds antibodies directed at membrane-proximal target epitopes of and 72°C. for 2 minutes followed by a terminal cycle of 72 human Endosialin the capability of the constructs to bind to C. for 5 minutes and the following primers are used for the macaque Endosialin and macaque CD3, respectively, is first PCR: investigated by FACS analysis. For this purpose the macaque Endosialin transfected CHO cells as described in example 8.15 and the macaque T cell line 4119LnPx (kindly provided forward primer: by Prof Fickenscher, Hygiene Institute, Virology, Erlangen 5'-atatgaatt cqccaccatgctgctg.cgcct SEQ ID NO: 418 Nuernberg; published in Knappe A, et al., and Fickenscher gttgctggcc-3' H., Blood 2000, 95, 3256-61) are used. 200,000 cells of the reverse primer: 25 respective cell lines are incubated for 30 minonice with 50 ul 5'-gtott catctt cotcatcct cocc-3' SEO ID NO: 419 of cell culture Supernatant of transfected cells expressing the cross-species specific bispecific antibody constructs. The The following reaction conditions: 1 cycle at 95°C. for 5 cells are washed twice in PBS with 2% FCS and binding of minutes followed by 40 cycles with 95°C. for 45 seconds, 58° the construct is detected with a murine Penta His antibody C. for 45 seconds and 72° C. for 2 minutes followed by a 30 (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After terminal cycle of 72° C. for 5 minutes and the following washing, bound anti His antibodies are detected with an Fc primers are used for the second PCR: gamma-specific antibody (Dianova) conjugated to phyco erythrin, diluted 1:100 in PBS with 2% FCS. Supernatant of forward primer: untransfected cells is used as a negative control. 5'-gtcaactacgttggtggctt coagtg-3 SEO ID NO: 42O 35 Flow cytometry is performed on a FACS-Calibur appara tus, the CellOuest Software is used to acquire and analyze the reverse primer: 5'-ggtctagat cactitat cqt catcatctttgt SEQ ID NO: 421 data (Becton Dickinson biosciences, Heidelberg). FACS agtic cacgctggttctgcaggtotgc-3' staining and measuring of the fluorescence intensity are per formed as described in Current Protocols in Immunology The PCR reactions are performed under addition of PCR 40 (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley grade betain to a final concentration of 1 M. Those PCRs Interscience, 2002). generate two overlapping fragments, which are isolated and sequenced according to standard protocols using the PCR 9. Generation of Bispecific Single Chain Antibodies primers, and thereby provided a portion of the cDNA Directed at Membrane-proximal Target Epitopes of sequence coding macaque Endosialin from codon 9 of the 45 Human IGF-1R leaderpeptide to codon 733 of the mature protein. To generate a construct for expression of macaque Endosialin a cDNA 9.1 Generation of CHO Cells Expressing Human IGF-1R fragment is obtained by gene synthesis according to standard The coding sequence of human IGF-1R as published in protocols (the cDNA and amino acid sequence of the con GenBank (Accession number NM 000875) is obtained by struct is listed under SEQ ID Nos 422 and 423). In this 50 gene synthesis according to standard protocols. The gene construct the coding sequence of macaque Endosialin from synthesis fragment is designed as to contain first a Kozak site amino acid 9 of the leader peptide to amino acid 733 of the for eukaryotic expression of the construct followed by the mature protein, followed in frame by the coding sequence of coding sequence of the human IGF-1R protein and a stop amino acid 734 to the last amino acid of the mature human codon (the cDNA and amino acid sequence of the construct is Endosialin protein, followed inframe by the coding sequence 55 listed under SEQ ID Nos 424 and 425). The gene synthesis ofa FLAG tag and a stop codon is fused inframe to the coding fragment is also designed as to introduce restriction sites at sequence of the amino acids 1 to 8 of the leader peptide of the the beginning and at the end of the fragment. The introduced human Endosialin protein. The gene synthesis fragment is restriction sites, EcoRI at the 5' end and Sal at the 3' end, are also designed as to containa Kozak site for eukaryotic expres utilized in the following cloning procedures. An internal sion of the construct and restriction sites at the beginning and 60 restriction site is removed by silent mutation of the coding the end of the fragment containing the cDNA. The introduced sequence in the gene synthesis fragment (BspEl: nucleotide restriction sites, EcoRI at the 5' end and Xbal at the 3' end, are 18 from A to C). The gene synthesis fragment is cloned via utilised in the following cloning procedures. The gene Syn EcoRI and SalI into a plasmid designated pEF-DHFR (pEF thesis fragment is cloned via EcoRI and Xbal into a plasmid DHFR is described in Raum et al. Cancer Immunol Immu designated pEF-DHFR (pEF-DHFR is described in Raum et 65 nother 50 (2001) 141-150) following standard protocols. The al. Cancer Immunol Immunother 50 (2001) 141-150) follow aforementioned procedures are carried out according to stan ing standard protocols. The aforementioned procedures are dard protocols (Sambrook, Molecular Cloning: A Laboratory US 9,260,522 B2 111 112 Manual, 3rd edition, Cold Spring Harbour Laboratory Press, out of cell culture supernatant on an Akta Explorer System Cold Spring Harbour, N.Y. (2001)). A clone with sequence (GE Amersham) and eluted by citric acid. The eluate is neu verified nucleotide sequence is transfected into DHFR defi tralized and concentrated. cient CHO cells for eukaryotic expression of the construct. 9.3 Generation of CHO Cells Expressing Murine IGF-1R Eukaryotic protein expression in DHFR deficient CHO cells The sequence of murine IGF-1R(NM 010513 Mus mus is performed as described by Kaufmann R.J. (1990) Methods culus insulin-like growth factor I receptor (Igf1 r), mRNA, Enzymol. 185, 537-566. Gene amplification of the construct National Center for Biotechnology Information, http://ww is induced by increasing concentrations of methotrexate w.ncbi.nlm.nih.gov/entrez) is used to obtain a synthetic (MTX) to a final concentration of up to 20 nM MTX. cDNA molecule by gene synthesis according to standard 9.2 Generation of a Soluble Human IGF-1R Fusion Protein 10 protocols. The gene synthesis fragment is designed as to The modified coding sequence of human IGF-1R as contain the coding sequence of an immunoglobulin heavy described in Example 9.1 is used for the construction of an chain leader comprising a start codon embedded within a artificial cDNA sequence encoding a soluble fusion protein of Kozak consensus sequence, followed in frame by the coding human IGF-1R and murine IgG1 Fc. To generate a construct 15 sequence of a FLAG tag, followed in frame by the complete for expression of the soluble human IGF-1Rfusion protein a coding sequence of mature murine IGF-1R (the cDNA and cDNA fragment is obtained by gene synthesis according to amino acid sequence of the construct is listed under SEQID standard protocols (the cDNA and amino acid sequence of the Nos 428 and 429). The gene synthesis fragment is also construct is listed under SEQID Nos 426 and 427). The gene designed as to introduce restriction sites at the 5' end (EcoRI) synthesis fragment is designed as to contain first a Kozak site and at the 3' end (Sal I) of the cDNA fragment for cloning into for eukaryotic expression of the construct followed by the the mammalian cell expression vector pEF-DHFR (pEF coding sequence of the human IGF-1R protein from amino DHFR is described in Raum et al. Cancer Immunol Immu acid 1 to 935 corresponding to the signal peptide and extra nother 50 (2001) 141-150). The aforementioned procedures cellular domains of human IGF-1R, followed in frame by the are carried out according to standard protocols (Sambrook, coding sequence of an artificial Ser-Gly-Ser-linker, fol 25 Molecular Cloning: A Laboratory Manual, 3rd edition, Cold lowed in frame by the coding sequence of the hinge region Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. and Fc gamma portion of murine IgG1, followed in frame by (2001)). A clone with sequence-verified nucleotide sequence the coding sequence of a 6 histidine tag and a stop codon. The is transfected into DHFR deficient CHO cells for eukaryotic gene synthesis fragment is also designed as to introduce expression of the construct. Eukaryotic protein expression in restriction sites at the beginning and at the end of the frag 30 DHFR deficient CHO cells is performed as described by ment. The introduced restriction sites, EcoRI at the 5' end and Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. SalI at the 3' end, are utilized in the following cloning proce Gene amplification for increased antigen expression is dures. The gene synthesis fragment is cloned via EcoRI and induced by increasing concentrations of methotrexate (MTX) Sall into a plasmid designated pEF-DHFR (pEF-DHFR is to a final concentration of up to 20 nM MTX. 35 9.4 Generation of CHO Cells Expressing a Mutated Human described in Raum et al. Cancer Immunol Immunother 50 IGF-1R Antigen with Murine Membrane-distal Epitopes (2001) 141-150) following standard protocols. The afore The coding sequence of a mutated human IGF-1 Rantigen mentioned procedures are all carried out according to stan with murine membrane-distal epitopes is obtained by gene dard protocols (Sambrook, Molecular Cloning: A Laboratory synthesis according to standard protocols. The gene synthesis Manual, 3rd edition, Cold Spring Harbour Laboratory Press, 40 fragment is designed as to contain the coding sequence of an Cold Spring Harbour, N.Y. (2001)). A clone with sequence immunoglobulin heavy chain leader comprising a start codon verified nucleotide sequence is transfected into DHFR defi embedded within a Kozak consensus sequence, followed in cient CHO cells for eukaryotic expression of the construct. frame by the coding sequence of a FLAG tag, followed in Eukaryotic protein expression in DHFR deficient CHO cells frame by the coding sequence of the L1 domain and the is performed as described by Kaufmann R.J. (1990) Methods 45 cysteine-rich domain of mature murine IGF-1R followed in Enzymol. 185, 537-566. Gene amplification of the construct frame by human IGF-1R from the L2 domain to the stop is induced by increasing concentrations of methotrexate codon (the cDNA and amino acid sequence of the construct is (MTX) to a final concentration of up to 20 nM MTX. After listed under SEQ ID Nos 430 and 431). The gene synthesis two passages of stationary culture the cells are grown in roller fragment is also designed as to introduce restriction sites at bottles with nucleoside-free HyOPF CHO liquid soy medium 50 the 5' end (EcoRI) and at the 3' end (Sal I) of the cDNA (with 4.0 mM L-Glutamine with 0.1% Pluronic F-68; fragment for cloning into the mammalian cell expression HyClone) for 7 days before harvest. The cells are removed by vector pEF-DHFR (pEF-DHFR is described in Raum et al. centrifugation and the Supernatant containing the expressed Cancer Immunol Immunother 50 (2001) 141-150). The afore protein is stored at -20°C. Alternatively a clone of the expres mentioned procedures are carried out according to standard sion plasmid with sequence-verified nucleotide sequence is 55 protocols (Sambrook, Molecular Cloning: A Laboratory used for transfection and protein expression in the FreeStyle Manual, 3rd edition, Cold Spring Harbour Laboratory Press, 293 Expression System (Invitrogen GmbH, Karlsruhe, Ger Cold Spring Harbour, N.Y. (2001)). A clone with sequence many) according to the manufacturer's protocol. Supernatant verified nucleotide sequence is transfected into DHFR defi containing the expressed protein is obtained, cells are cient CHO cells for eukaryotic expression of the construct. removed by centrifugation and the Supernatant is stored at 60 Eukaryotic protein expression in DHFR deficient CHO cells -20°C. For purification of the soluble human IGF-1Rfusion is performed as described by Kaufmann R.J. (1990) Methods protein a goat anti-mouse Fc affinity column is prepared Enzymol. 185, 537-566. Gene amplification for increased according to standard protocols using a commercially avail antigen expression is induced by increasing concentrations of able affinity purified goat anti-mouse IgG Fc fragment spe methotrexate (MTX) to a final concentration of up to 20 nM cific antibody with minimal cross-reaction to human, bovine 65 MTX. and horse serum proteins (Jackson ImmunoResearch Europe 9.5 Generation of CHO Cells Expressing a Mutated Murine Ltd.). Using this affinity column the fusion protein is isolated IGF-1R Antigen with Human Membrane-distal Epitopes US 9,260,522 B2 113 114 The coding sequence of a mutated murine IGF-1 Rantigen region (VH) DNA-fragments is constructed by RT-PCR on with human membrane-distal epitopes is obtained by gene murine spleen RNA using VK- and VH specific primers. synthesis according to standard protocols. The gene synthesis cDNA is synthesized according to standard protocols, see fragment is designed as to contain the coding sequence of an example 2.7. immunoglobulin heavy chain leader comprising a start codon 450 ng of the kappa light chain fragments (SacI-Spel embedded within a Kozak consensus sequence, followed in digested) are ligated with 1400 ng of the phagemid frame by the coding sequence of a FLAG tag, followed in pComb3H5Bhis (SacI-Spel digested; large fragment). The frame by the L1 domain and the cysteine-rich domain of resulting combinatorial antibody library is then transformed mature human IGF-1R followed in frame by murine IGF-1R into 300 ul of electrocompetent Escherichia coli XL 1 Blue from the L2 domain to the stop codon (the cDNA and amino 10 acid sequence of the construct is listed under SEQID Nos 432 cells by electroporation (2.5 kV, 0.2 cm gap cuvette, 25uFD, and 433). The gene synthesis fragment is also designed as to 200 Ohm, Biorad gene-pulser) resulting in a library size of introduce restriction sites at the 5' end (EcoRI) and at the 3' more than 107 independent clones. After one hour of pheno end (Sal I) of the cDNA fragment for cloning into the mam type expression, positive transformants are selected for car malian cell expression vector pEF-DHFR (pEF-DHFR is 15 benicillin resistance encoded by the pComb3H5BHis vector described in Raum et al. Cancer Immunol Immunother 50 in 100 ml of liquid super broth (SB)-culture over night. Cells (2001) 141-150). The aforementioned procedures are carried are then harvested by centrifugation and plasmid preparation out according to standard protocols (Sambrook, Molecular is carried out using a commercially available plasmid prepa Cloning: A Laboratory Manual, 3rd edition, Cold Spring ration kit (Qiagen). Harbour Laboratory Press, Cold Spring Harbour, N.Y. 2800 ng of this plasmid-DNA containing the VK-library (2001)). A clone with sequence-verified nucleotide sequence (XhoI-BstEII digested; large fragment) are ligated with 900 is transfected into DHFR deficient CHO cells for eukaryotic ng of the heavy chain V-fragments (XhoI-BstEII digested) expression of the construct. Eukaryotic protein expression in and again transformed into two 300 ulaliquots of electrocom DHFR deficient CHO cells is performed as described by petent E. coli XL 1 Blue cells by electroporation (2.5 kV, 0.2 Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. 25 cm gap cuvette, 25uFD, 200Ohm) resulting in a totalVH-VK Gene amplification for increased antigen expression is schv (single chain variable fragment) library size of more induced by increasing concentrations of methotrexate (MTX) than 10 independent clones. to a final concentration of up to 20 nM MTX. After phenotype expression and slow adaptation to carbe 9.6 Immunization of Mice Using a Soluble Human IGF-1R nicillin, the E. coli cells containing the antibody library are Fusion Protein 30 transferred into SB-Carbenicillin (SB with 50 lug/mL carbe Twelve weeks old F1 mice from BALB? cxC57BL/6 cross nicillin) selection medium. The E. coli cells containing the ings are immunized with the soluble human IGF-1R fusion antibody library are then infected with an infectious dose of protein as described in Example 9.2. To this end for each 10' particles of helper phage VCSM13 resulting in the pro animal 40 ug of the soluble human IGF-1Rfusion protein are duction and secretion of filamentous M13 phage, wherein mixed with 10 nmol of a thioate-modified CpG-Oligonucle 35 each phage particle contains single stranded otide (5'-tccatgacgttcctgatgct-3") in 300 ul PBS and are pComb3H5BHis-DNA encoding a murine sclv-fragment injected intraperitoneally. Mice receive booster immuniza and displays the corresponding scFV-proteinas a translational tions after 21, 42 and optionally 63 days in the same way. Ten fusion to phage coat protein III. This pool of phages display days after the first booster immunization, blood samples are ing the antibody library is later used for the selection of taken and antibody serum titers against human IGF-1R are 40 antigen binding entities. tested by flow cytometry according to standard protocols. To 9.8 Phage Display Based Selection of Membrane-proximal this end 200,000 cells of the human IGF-1R transfected CHO Target Binders on CHO Cells Expressing the Mutated Human cells as described in Example 9.1 are incubated for 30 min on IGF-1R Antigen with Murine Membrane-distal Epitopes ice with 50 ul of serum of the immunized animals diluted The phage library carrying the cloned schv-repertoire is 1:1000 in PBS with 2% FCS. The cells are washed twice in 45 harvested from the respective culture supernatant by PBS with 2% FCS and binding of serum antibodies is PEG8000/NaCl precipitation and centrifugation. Approxi detected with an mouse Fc gamma-specific antibody (Di mately 10' to 10' sclv phage particles are resuspended in anova) conjugated to phycoerythrin, diluted 1:100 in PBS 0.4 ml of PBS/0.1% BSA and incubated with 10 to 107 CHO with 2% FCS. Serum of the animals obtained prior to immu cells expressing the mutated human IGF-1R antigen with nization is used as a negative control. Flow cytometry is 50 murine membrane-distal epitopes as described in example 9.4 performed on a FACS-Calibur apparatus, the CellOuest soft for 1 hour on ice under slow agitation. These CHO cells are ware is used to acquire and analyze the data (Becton Dickin grown beforehand, harvested by centrifugation, washed in son biosciences, Heidelberg). FACS staining and measuring PBS and resuspended in PBS/1% FCS (containing Na Azide). of the fluorescence intensity are performed as described in schv phage which do not specifically bind to the CHO cells Current Protocols in Immunology (Coligan, Kruisbeek, Mar 55 are eliminated by up to five washing steps with PBS/1% FCS gullies. Shevach and Strober, Wiley-Interscience, 2002). (containing Na Azide). After washing, binding entities are Animals demonstrating significant serum reactivity eluted from the cells by resuspending the cells in HCl-glycine against human IGF-1R as determined by the FACS analysis pH 2.2 (10 min incubation with Subsequent Vortexing) and are used in the Subsequent experiment. after neutralization with 2M Tris pH 12, the eluate is used for 9.7 Generation of an Immune Murine Antibody Scfv Library: 60 infection of a fresh uninfected E. coli XL1 Blue culture Construction of a Combinatorial Antibody Library and Phage (OD600>0.5). The E. coli culture containing E. coli cells Display Successfully transduced with a phagemid copy, encoding a Three days after the last booster immunization spleen cells murine ScFV-fragment, are again selected for carbenicillin of reactive animals are harvested for the preparation of total resistance and subsequently infected with VCMS13 helper RNA according to standard protocols. 65 phage to start the second round of antibody display and in A library of murine immunoglobulin (Ig) light chain vitro selection. Typically a total of 4 to 5 rounds of selections (kappa) variable region (VK) and Ig heavy chain variable are carried out. US 9,260,522 B2 115 116 9.9 Screening for Membrane-proximal Target Binders on There are a number of framework residues of the non-human CHO Cells Expressing the Human IGF-1R Antigen, the VH that differ from the human VH framework regions (“dif Murine IGF-1R Antigen and the Mutated Murine IGF-1R ferent framework positions'). Some of these residues may Antigen with Human Membrane-distal Epitopes contribute to the binding and activity of the antibody to its Plasmid DNA corresponding to 4 and 5 rounds of panning target. is isolated from E. coli cultures after selection. For the pro To construct a library that contains the murine CDRs and at duction of soluble schv-protein, VH-VL-DNA fragments are every framework position that differs from the chosen human excised from the plasmids (XhoI-SpeI). These fragments are VH sequence both possible residues (the human and the cloned via the same restriction sites in the plasmid maternal murine amino acid residue), degenerated oligo pComb3H5BFlag? His differing from the original 10 pComb3H5BHis in that the expression construct (e.g. scFv) nucleotides are synthesized. These oligonucleotides incorpo includes a Flag-tag (TGDYKDDDDK) between the scFv and rate at the differing positions the human residue with a prob the His6-tag and the additional phage proteins are deleted. ability of 75% and the murine residue with a probability of After ligation, each pool (different rounds of panning) of 25%. For one human VH e.g. six of these oligonucleotides plasmid DNA is transformed into 100 ul heat shock compe 15 have to be synthesized that overlap in a terminal stretch of tent E. coli TG1 or XLI blue and plated onto carbenicillin approximately 20 nucleotides. To this end every second LB-agar. Single colonies are picked into 100 ul of LB carb primer is an antisense primer. Restriction sites within the (LB with 50 lug/ml carbenicillin). oligonucleotides needed for later cloning are deleted. After induction with 1 mMIPTG. E. coli transformed with These primers may have a length of 60 to 90 nucleotides, pComb3H5BFlag? His containing a VL- and VH-segment depending on the number of primers that are needed to span produce soluble scFv in sufficient amounts. Due to a suitable over the whole V sequence. signal sequence, the Sclv is exported into the periplasma These e.g. six primers are mixed in equal amounts (e.g. 1 ul where it folds into a functional conformation. of each primer (primer stocks 20 to 100 uM) to a 20 lul PCR Single E. coli bacterial colonies from the transformation reaction) and added to a PCR mix consisting of PCR buffer, plates are picked for periplasmic Small scale preparations and 25 nucleotides and Taq polymerase. This mix is incubated at 94° grown in SB-medium (e.g. 10 ml) supplemented with 20 mM C. for 3 minutes, 65° C. for 1 minute, 62° C. for 1 minute, 59° MgCl, and carbenicillin 50 ug/ml (and re-dissolved in PBS C. for 1 minute, 56°C. for 1 minute, 52°C. for 1 minute, 50° (e.g. 1 ml) after harvesting. A temperature shock is applied by C. for 1 minute and at 72°C. for 10 minutes in a PCR cycler. four rounds of freezing at -70° C. and thawing at 37° C. Subsequently the product is run in an agarose gel electro whereby the outer membrane of the bacteria is destroyed and 30 phoresis and the product of a size from 200 to 400 base pairs the Soluble periplasmic proteins including the Sclvs are isolated from the gel according to standard methods. released into the supernatant. After elimination of intact cells This PCR product is then used as a template for a standard and cell-debris by centrifugation, the Supernatant containing PCR reaction using primers that incorporate suitable N-ter the murine anti-human IGF-1R-scFVs is collected and used minal and C-terminal cloning restriction sites. The DNA frag for further examination. 35 ment of the correct size (for a VH approximately 350 nucle Screening of the isolated schvs for membrane-proximal otides) is isolated by agarose gel electrophoresis according to target binders is performed by flow cytometry on CHO cells standard methods. In this way sufficientVHDNA fragment is expressing the human IGF-1R antigen as described in amplified. This VH fragment is now a pool of VH fragments Example 9.1, the murine IGF-1R antigen as described in that have each one a different amount of human and murine Example 9.3 and the mutated murine IGF-1R antigen with 40 residues at the respective differing framework positions (pool human membrane-distal epitopes as described in Example of humanized VH). The same procedure is performed for the 9.5. VL region of the murine anti-IGF-1R scFv to a membrane For flow cytometry 2.5x10 cells of the respective cell lines proximal target epitope of human IGF-1R (pool of human are incubated with 50 ul supernatant. The binding of the ized VL). constructs is detected with an anti-His antibody (Penta-His 45 The pool of humanized VH is then combined with the pool Antibody, BSA free, Qiagen GmbH, Hilden, FRG) at 2 ug/ml of humanized VL in the phage display vectorpComb3H5Bhis in 50 lul PBS with 2% FCS. As a second step reagent a to form a library of functional scFVs from which—after dis R-Phycoerythrin-conjugated affinity purified F(ab')2 frag play on filamentous phage—anti-IGF-1R binders to mem ment, goat anti-mouse IgG (Fc-gamma fragment specific), brane-proximal target epitopes of human IGF-1R are diluted 1:100 in 50 ul PBS with 2% FCS (Dianova, Hamburg, 50 selected, screened, identified and confirmed as described FRG) is used. The samples are measured on a FACSscan (BD above for the parental non-human (murine) anti-IGF-1R biosciences, Heidelberg, FRG). sch v. Single clones are then analyzed for favorable properties Only constructs which show binding to CHO cells express and amino acid sequence. Those scFVs, which are closest in ing the human IGF-1R antigen and do not show binding to amino acid sequence homology to human germline V-seg CHO cells expressing the murine IGF-1 Rantigen and also do 55 ments, are preferred. not show binding to CHO cells expressing the mutated Human/humanized anti-IGF-1R scFvs to membrane murine IGF-1R antigen with human membrane-distal proximal target epitopes of human IGF-1R are converted into epitopes are selected for further use. recombinant bispecific single chain antibodies and further 9.10 Generation of Human/Humanized Equivalents of Non characterized as follows. human schvs to Membrane-proximal Target Epitopes of 60 9.11 Generation of I2C-based Bispecific Single Chain Anti Human Igf-1R bodies Directed at Membrane-proximal Target Epitopes of The VH region of a murine anti-IGF-1R scFv to a mem Human IGF-1R brane-proximal target epitope of human IGF-1R is aligned Anti-IGF-1R scFvs to membrane-proximal target epitopes against human antibody germline amino acid sequences. The of human IGF-1R with favorable properties and amino acid human antibody germline VH sequence is chosen which has 65 sequence are converted into recombinant bispecific single the closest homology to the non-human VH and a direct chain antibodies by joining them via a Gly-Ser-linker with alignment of the two amino acid sequences is performed. the CD3 specific schv 120 (SEQID) to result in constructs US 9,260,522 B2 117 118 with the domain arrangement VH-1-(Gly Ser)- sodium phosphate buffer pH 7.2, 0.1 M NaCl) and the cell VLoir-ir-Ser Gly-Ser-VHos-(Gly-Seri)s-VLDs. culture supernatant (500 ml) is applied to the column (10 ml) I2C-based bispecific single chain antibodies directed at at a flow rate of 3 ml/min. The column is washed with buffer membrane-proximal target epitopes of human IGF-1R were A to remove unbound sample. Bound protein is eluted using designed as set out in the following Table 8: a two step gradient of buffer B (20 mM sodium phosphate buffer pH 7.2, 0.1 MNaCl, 0.5 MImidazole) according to the TABLE 8 following procedure: Step 1: 20% buffer B in 6 column volumes Formats of I2C-based bispecific single chain antibodies directed Step 2: 100% buffer B in 6 column volumes at membrane-proximal target epitopes of human IGF-1R 10 Eluted protein fractions from step 2 are pooled for further SEQ ID Formats of protein constructs purification. All chemicals are of research grade and pur (nucliprot) (N -> C) chased from Sigma (Deisenhofen) or Merck (Darmstadt). Gel filtration chromatography is performed on a HiLoad 848,847 IGF1R12HLXI2CHL 16/60 Superdex 200 prep grade column (GE/Amersham) 15 equilibrated with Equi-buffer (25 mM Citrate, 200 mM Lysine, 5% Glycerol, pH 7.2). Eluted protein samples (flow Alternatively further constructs with different domain rate 1 ml/min) are subjected to standard SDS-PAGE and arrangements can be generated according to standard proto Western Blot for detection. Prior to purification, the columnis colls. For expression in CHO cells the coding sequences of (i) calibrated for molecular weight determination (molecular an N-terminal immunoglobulin heavy chain leader compris weight marker kit, Sigma MW GF-200). Protein concentra ing a start codon embedded within a Kozak consensus tions are determined using OD280 nm. sequence and (ii) a C-terminal His6-tag followed by a stop Purified bispecific single chain antibody protein is ana codon are both attached in frame to the nucleotide sequence lyzed in SDS PAGE under reducing conditions performed encoding the bispecific single chain antibodies prior to inser with pre-cast 4-12% Bis Tris gels (Invitrogen). Sample prepa tion of the resulting DNA-fragment as obtained by gene Syn 25 ration and application are performed according to the proto thesis into the multiple cloning site of the expression vector col provided by the manufacturer. The molecular weight is pEF-DHFR (Raum et al. Cancer Immunol Immunother 50 determined with MultiMark protein standard (Invitrogen). (2001) 141-150). A clone with sequence-verified nucleotide The gel is stained with colloidal Coomassie (Invitrogen pro sequence is transfected into DHFR deficient CHO cells for tocol). The purity of the isolated protein is typically >95% as eukaryotic expression of the construct. Eukaryotic protein 30 determined by SDS-PAGE. expression in DHFR deficient CHO cells is performed as The bispecific single chain antibody has a molecular described by Kaufmann R.J. (1990) Methods Enzymol. 185, weight of about 52 kDa under native conditions as determined 537-566. Gene amplification of the construct is induced by by gel filtration in PBS. increasing concentrations of methotrexate (MTX) to a final Western Blot is performed using an Optitran?R BA-S83 concentration of up to 20 nM MTX. 35 membrane and the Invitrogen Blot Module according to the 9.12 Expression and Purification of Bispecific Single Chain protocol provided by the manufacturer. The antibody used is Antibody Molecules Directed at Membrane-proximal Target directed against the His Tag (Penta His, Qiagen) and a Goat Epitopes of Human IGF-1R anti-mouse Ig labeled with alkaline phosphatase (AP) Bispecific single chain antibody molecules are expressed (Sigma) is used as second step reagent, and BCIP/NBT in Chinese hamster ovary cells (CHO). Eukaryotic protein 40 (Sigma) as substrate. A band detected at 52 kD corresponds to expression in DHFR deficient CHO cells is performed as purified bispecific single chain antibodies. described by Kaufmann R.J. (1990) Methods Enzymol. 185, 9.13 Flow Cytometric Binding Analysis of Bispecific Anti 537-566. Gene amplification of the constructs is induced by bodies Directed at Membrane-proximal Target Epitopes of addition of increasing concentrations of MTX up to final Human IGF-1R concentrations of 20 nMMTX. After two passages of station 45 In order to test the functionality of bispecific antibody ary culture cell culture Supernatant is collected and used in the constructs regarding the capability to bind to CD3 and to Subsequent experiments. To generate Supernatant for purifi membrane-proximal target epitopes of human IGF-1R, cation after two passages of Stationary culture the cells are respectively, a FACS analysis is performed. For this purpose grown in roller bottles with nucleoside-free HyO PF CHO CHO cells transfected with human IGF-1R as described in liquid soy medium (with 4.0 mM L-Glutamine with 0.1% 50 Example 9.1 and the human CD3 positive T cell leukemia cell Pluronic F-68; HyClone) for 7 days before harvest. The cells line HPB-ALL (DSMZ. Braunschweig, ACC483) are used. are removed by centrifugation and the Supernatant containing For confirmation of binding to membrane-proximal target the expressed protein is stored at -20°C. Alternatively, con epitopes of human IGF-1R in addition—CHO cells structs are transiently expressed in HEK 293 cells. Transfec expressing the murine IGF-1R antigen as described in tion is performed with 293fectin reagent (Invitrogen, #12347 55 Example 9.3, CHO cells expressing the mutated human IGF 019) according to the manufacturer's protocol. Furthermore 1R antigen with murine membrane-distal epitopes as the constructs are alternatively expressed in transiently trans described in Example 9.4 and CHO cells expressing the fected DHFR deficient CHO cells using for example mutated murine IGF-1R antigen with human membrane-dis FuGENER HD Transfection Reagent (Roche Diagnostics tal epitopes as described in Example 9.5 are used. 200,000 GmbH, Cat. No. 04709691001) according to the manufactur 60 cells of the respective cell lines are incubated for 30 min on er's protocol. ice with 50 ul of cell culture supernatant of transfected cells Akta R. Explorer System (GE Health Systems) and Uni expressing the bispecific antibody constructs. The cells are corn R Software are used for chromatography. Immobilized washed twice in PBS with 2% FCS and binding of the con metal affinity chromatography (“IMAC) is performed using struct is detected with a murine Penta His antibody (Qiagen; a Fractogel EMD ChelateR (Merck) which is loaded with 65 diluted 1:20 in 50 ul PBS with 2% FCS). After washing, ZnCl2 according to the protocol provided by the manufac bound anti His antibodies are detected with an Fc gamma turer. The column is equilibrated with buffer A (20 mM specific antibody (Dianova) conjugated to phycoerythrin, US 9,260,522 B2 119 120 diluted 1:100 in PBS with 2% FCS. Supernatant of untrans GmbH, Koln, Germany). Analysis of the experimental data is fected cells is used as a negative control. performed with Prism 4 for Windows(R (version 4.02, Graph Flow cytometry is performed on a FACS-Calibur appara Pad Software Inc., San Diego, Calif., USA). Sigmoidal dose tus, the CellOuest Software is used to acquire and analyze the responsecurves typically have R values >0.90 as determined data (Becton Dickinson biosciences, Heidelberg). FACS 5 by the software. staining and measuring of the fluorescence intensity are per Only those constructs showing potent recruitment of cyto formed as described in Current Protocols in Immunology toxic activity of effector T cells against target cells positive (Coligan, Kruisbeek, Margulies. Shevach and Strober, Wiley for IGF-1R are selected for further use. Interscience, 2002). 9.15 Generation of CHO Cells Expressing Macaque IGF-1R Only those constructs that show bispecific binding to 10 human CD3 as well as to human IGF-1R and neither bind to The coding sequence of macaque IGF-1R as published in the murine IGF-1 Rantigen nor to the mutated murine IGF-1R GenBank (Accession number XM 001100407) is obtained antigen with human membrane-distal epitopes are selected by gene synthesis according to standard protocols. The gene for further use. synthesis fragment is designed as to contain first a Kozak site The bispecific binding of the single chain molecules listed 15 for eukaryotic expression of the construct followed by the above was clearly detectable as shown in FIG. 17. In the coding sequence of the macaque IGF-1R protein and a stop FACS analysis all constructs showed binding to human CD3 codon (the cDNA and amino acid sequence of the construct is and human IGF-1R compared to the negative control. Mem listed under SEQ ID Nos 434 and 435). The gene synthesis brane-proximal target binding of the single chain molecules fragment is also designed as to introduce restriction sites at listed above was clearly detectable as shown in FIG. 19. In the the beginning and at the end of the fragment. The introduced FACS analysis all constructs showed binding to the mutated restriction sites, EcoRI at the 5' end and Sal at the 3' end, are human IGF-1R antigen with murine membrane-distal utilized in the following cloning procedures. The gene Syn epitopes and did not show binding to the murine IGF-1R thesis fragment is cloned via EcoRI and SalI into a plasmid antigen and did also not show binding to the mutated murine designated pEF-DHFR (pEF-DHFR is described in Raum et IGF-1R antigen with human membrane-distal epitopes as 25 al. Cancer Immunol Immunother 50 (2001) 141-150) follow compared to the negative control. ing standard protocols. The aforementioned procedures are 9.14 Bioactivity of Bispecific Antibodies Directed at Mem carried out according to standard protocols (Sambrook, brane-proximal Target Epitopes of Human IGF-1R Molecular Cloning: A Laboratory Manual, 3rd edition, Cold Bioactivity of generated bispecific single chain antibodies Spring Harbour Laboratory Press, Cold Spring Harbour, N.Y. is analyzed by chromium 51 (51Cr) release in vitro cytotox 30 (2001)). A clone with sequence-verified nucleotide sequence icity assays using the CHO cells transfected with human is transfected into DHFR deficient CHO cells for eukaryotic IGF-1R described in Example 9.1. As effector cells stimu expression of the construct. Eukaryotic protein expression in lated human CD4/CD56 depleted PBMC are used. DHFR deficient CHO cells is performed as described by Stimulated human PBMC are obtained as follows: Kaufmann R. J. (1990) Methods Enzymol. 185, 537-566. A Petri dish (145 mm diameter, Greiner bio-one GmbH, 35 Gene amplification of the construct is induced by increasing Kremsmünster) is coated with a commercially available anti concentrations of methotrexate (MTX) to a final concentra CD3 specific antibody (e.g. OKT3. Orthoclone) in a final tion of up to 20 nM MTX. concentration of 1 lug/ml for 1 hour at 37°C. Unbound protein 9.16 Flow Cytometric Analysis of Cross-species Specificity is removed by one washing step with PBS. The fresh PBMC of Bispecific Antibodies Directed at Membrane-proximal are isolated from peripheral blood (30-50 ml human blood) 40 Target Epitopes of Human IGF-1R by Ficoll gradient centrifugation according to standard pro In order to test the cross-species specificity of bispecific tocols. 3-5x107 PBMC are added to the precoated petridish in antibodies directed at membrane-proximal target epitopes of 120 ml of RPMI 1640 with stabilized glutamine/10% FCS/ human IGF-1R the capability of the constructs to bind to IL-2 20 U/ml (Proleukin, Chiron) and stimulated for 2 days. macaque IGF-1R and macaque CD3, respectively, is investi On the third day the cells are collected and washed once with 45 gated by FACS analysis. For this purpose the macaque IGF RPMI 1640. IL2 is added to a final concentration of 20U/ml 1R transfected CHO cells as described in example 9.15 and and the cells are cultivated again for one day in the same cell the macaque T cell line 4119LnPx (kindly provided by Prof culture medium as above. Fickenscher, Hygiene Institute, Virology, Erlangen-Nuern By depletion of CD4+ T cells and CD56+ NK cells accord berg; published in Knappe A, et al., and Fickenscher H., ing to standard protocols CD8+ cytotoxic T lymphocytes 50 Blood 2000, 95, 3256-61) are used. 200,000 cells of the (CTLs) are enriched. respective cell lines are incubated for 30 minonice with 50 ul Target cells are washed twice with PBS and labeled with of cell culture Supernatant of transfected cells expressing the 11.1 MBq 'Cr in a final volume of 100 ul RPMI with 50% cross-species specific bispecific antibody constructs. The FCS for 60 minutes at 37° C. Subsequently the labeled target cells are washed twice in PBS with 2% FCS and binding of cells are washed 3 times with 5 ml RPMI and then used in the 55 the construct is detected with a murine Penta His antibody cytotoxicity assay. The assay is performed in a 96 well plate (Qiagen; diluted 1:20 in 50 ul PBS with 2% FCS). After in a total volume of 250 ul supplemented RPMI (as above) washing, bound anti His antibodies are detected with an Fc with an E:T ratio of 10:1. Supernatant of cells expressing the gamma-specific antibody (Dianova) conjugated to phyco bispecific single chain antibody molecules in a final concen erythrin, diluted 1:100 in PBS with 2% FCS. Supernatant of tration of 50% and 20 twofold dilutions thereof are applied. 60 untransfected cells is used as a negative control. Flow cytom The assay time is 18 hours. Cytotoxicity is measured as etry is performed on a FACS-Calibur apparatus, the relative values of released chromium in the Supernatant CellOuest software is used to acquire and analyze the data related to the difference of maximum lysis (addition of Tri (Becton Dickinson biosciences, Heidelberg). FACS staining ton-X) and spontaneous lysis (without effector cells). All and measuring of the fluorescence intensity are performed as measurements are done in quadruplicates. Measurement of 65 described in Current Protocols in Immunology (Coligan, chromium activity in the Supernatants is performed with a Kruisbeek, Margulies. Shevach and Strober, Wiley-Inter Wizard R. 3" gammacounter (Perkin Elmer Life Sciences science, 2002). US 9,260,522 B2 121 122 The cross-species specific binding of the single chain mol 10.2.1 Generation of CHO Cells Expressing Human/Rat ecules listed above was clearly detectable as shown in FIG. PSMA Chimeras 17. In the FACS analysis all constructs showed binding to PSMA of rattus norvegicus, which is not bound by PSMA macaque CD3 and macaque IGF-1R compared to the nega bispecific single chain antibody PSMA-P7 HLXI2CHL, was tive control. Human/humanized equivalents of scEvs Specific used for making chimeras with human PSMA. Thus, creating for IGF-1R contained in the bispecific single chain molecules a chimera in the region containing the binding epitope of a are generated as described herein. Cloning of binding mol PSMA bispecific single chain antibody leads to loss of bind ecules based on these human/humanized scEvs and expres ing of said bispecific single chain antibody to the respective sion and purification of these bispecific single chain mol PSMA construct. ecules is performed as described above. Flow cytometric 10 The coding sequence of human PSMA as published in analysis of bispecific binding and analysis of bioactivity by GenBank (Accession number NM 004476) and the coding chromium 51 (51Cr) release in vitro cytotoxicity assays is sequence of rat PSMA (NM 057185, Rattus norvegicus performed as described above. Based on demonstrated bispe folate hydrolase (Folh1), mRNA, National Center for Bio cific binding and recruited cytotoxicity binding molecules are 15 technology Information, http://www.ncbi.nlm.nih.gov/en selected for further use. trez) were used for generation of the chimeric constructs. A set of 6 chimeric cDNA constructs was designed and Example 10 generated by gene synthesis according to standard protocols. In the constructs segments of the coding sequences for the 10.1. Generation of PSMA- and CD3-directed Bispecific amino acids 140 to 169,281 to 284,300 to 344, 589 to 617, Single Chain Antibodies 683 to 690 and 716 to 750, respectively, were exchanged for Bispecific single chain antibodies comprising either Sclv binding domain P7 against a PSMA-epitope with <60 A the homologous sequences of rat PSMA. membrane-distance or scv binding domain D4 against a Chimeric PSMA constructs were generated as described PSMA-epitope with >60 A membrane-distance and the scFv above and designated as set out in the following Table 10: binding domain I2C directed at CD3epsilon on human T cells 25 were obained by gene synthesis. The gene synthesis frag TABLE 10 ments were designed as to contain first a Kozak site for Designation of chimeric PSMA constructs eukaryotic expression of the construct, followed by a 19 amino acid immunoglobulin leader peptide, followed in SEQ ID frame by the coding sequence of the bispecific single chain 30 (nucliprot) Designation antibody, followed in frame by the coding sequence of a 6 461f462 huPSMArat140-169 463,464 huPSMArat?81-284 histidine tag and a stop codon. The variable region arrange 465,466 huPSMArat00-344 ments as well as the SEQ ID Nos of the cDNA- and amino 467.468 huPSMAlrats'98-617 acid sequences are listed in the table 9 below. 35 469,470 huPSMArató83-690 471,472 PSMAratf1 6-7SO

SEQ ID Formats of protein constructs (nucliprot) (N -> C) The gene synthesis fragments were designed as to contain first a Kozak site for eukaryotic expression of the construct 488,487 PSMA-P7 HL XI2CHL 40 followed by the coding sequence of the chimeric PSMA 474,473 PSMA-D4 HL XI2CHL proteins, followed in frame by the coding sequence of a FLAG-tag and a stop codon. The gene synthesis fragments The gene synthesis fragments were also designed as to were also designed as to introduce restriction sites at the introduce suitable restriction sites at the beginning (EcoRI) beginning and at the end of the fragments. The introduced and at the end of the fragment (Sal I) for cloning of the gene 45 restriction sites, EcoRI at the 5' end and Salat the 3' end, were synthesis fragment into the mammalian cell expression vec utilized in the following cloning procedures. Undesirable torpEF-DHFR (pEF-DHFR is described in Raum et al. Can internal restriction sites were removed by silent mutation of cer Immunol Immunother 50 (2001) 141-150). The afore the coding sequence in the gene synthesis fragments. The mentioned procedures were carried out according to standard gene synthesis fragments were cloned via EcoRI and Sal into protocols (Sambrook, Molecular Cloning: A Laboratory 50 a plasmid designated pBF-DHFR (pEF-DHFR is described in Manual, 3rd edition, Cold Spring Harbour Laboratory Press, Raum et al. Cancer Immunol Immunother 50 (2001) 141 Cold Spring Harbour, N.Y. (2001)). A clone with sequence 150) following standard protocols. The aforementioned pro verified nucleotide sequence was transfected into DHFR defi cedures were carried out according to standard protocols cient CHO cells for eukaryotic expression of the construct. (Sambrook, Molecular Cloning: A Laboratory Manual, 3rd Eukaryotic protein expression in DHFR deficient CHO cells 55 edition, Cold Spring Harbour Laboratory Press, Cold Spring was performed as described by Kaufmann R. J. (1990) Meth Harbour, N.Y. (2001)). A clone with sequence-verified nucle ods Enzymol. 185, 537-566. Gene amplification of the con otide sequence was transfected into DHFR deficient CHO struct was induced by increasing concentrations of methotr cells for eukaryotic expression of the construct. Eukaryotic exate (MTX) to a final concentration of up to 20 nM MTX. protein expression in DHFR deficient CHO cells was per After two passages of stationary culture cell culture Superna 60 formed as described by Kaufmann R. J. (1990) Methods tant was collected and used in the Subsequent experiments. Enzymol. 185, 537-566. Gene amplification of the construct 10.2 Epitope Mapping of the PSMA- and CD3-Reactive was induced by increasing concentrations of methotrexate Bispecific Single Chain Antibody Molecule PSMA-P7 (MTX) to a final concentration of up to 20 nM MTX. HLXI2CHL 10.2.2 Flow Cytometric Binding Analysis for Epitope Map APSMA-epitope with <60 A membrane-distance of bispe 65 ping of the PSMA- and CD3-Reactive Bispecific Single cific single chain antibody PSMA-P7 HLXI2CHL was con Chain Antibody Molecule PSMA-P7 HLXI2CHL Using Chi firmed by epitope mapping using chimeric PSMA constructs. meric PSMA Proteins US 9,260,522 B2 123 124 In order to determine the binding epitope of the PSMA PSMA-D4 HLXI2CHL recognizes a membrane-distal target bispecific single chain antibody PSMA-P7 HLXI2C HL a epitope of human PSMA with 60 A membrane-distance as FACS analysis was performed. For this purpose CHO cells defined herein. transfected with human/rat chimeric PSMA molecules as 10.4 Comparative Analysis of Cytotoxic Activity of Bispe described in Example 10.2.1 above were used. FACS analysis cific Antibodies Single-chain Antibodies Directed at Mem with supernatant of CHOcells expressing PSMA-P7 HLXI2C brane-proximal and Membrane-distal Target Epitopes of HL was performed as described herein. Detection of binding Human PSMA of PSMA-P7 HLXI2C HL was performed using a murine Bioactivity of bispecific single chain antibodies was ana Penta His antibody and as second step reagent an Fc gamma lyzed by a CytoTox-GloTM cytotoxicity assay with unstimu specific antibody conjugated to phycoerythrin. Supernatant 10 lated human PBMC using the CHO cells transfected with of untransfected cells was used as a negative control. Super human PSMA described in Example 2.1. As effector cells natant of CHO cells expressing the bispecific single chain unstimulated human PBMC were used. antibody construct PSMA-D4 HLXI2C HL cross-reactive Unstimulated human PBMC were obtained as follows: with rat PSMA was used as control for expression of the 15 Human peripheral blood mononuclear cells (PBMC) were chimeric PSMA constructs. prepared by Ficoll density gradient centrifugation from As shown in FIG. 20 both PSMA bispecific single chain enriched lymphocyte preparations (buffy coats), a side prod antibodies, PSMA-P7 HLXI2C HL and PSMA-D4 HLXI2C uct of blood banks collecting blood for transfusions. Buffy HL, showed binding to the chimeric constructs huPSMA coats were supplied by a local blood bank and PBMC were rat 140-169, huPSMArat281-284, huPSMArat300-344, prepared on the same day of blood collection. After Ficoll huPSMArat683-690 and huPSMArat71 6-750. AS further density centrifugation and extensive washes with Dulbecco's more shown in FIG. 20 there is a lack of binding for PSMA PBS (Gibco), remaining erythrocytes were removed from P7 HLXI2CHL to the construct huPSMArats 98-617, which PBMC via incubation with erythrocyte lysis buffer (155 mM demonstrates the presence of its binding epitope in the region NH4C1, 10 mM. KHCO, 100 M EDTA). Platelets were of amino acids 598 to 617 of human PSMA. 25 removed via the supernatant upon centrifugation of PBMC at As shown in Table 1 the amino acids 598 to 617 constitute 100x.g. Remaining lymphocytes mainly encompass B and T a membrane proximal epitope as defined herein. In conclu lymphocytes, NK cells and monocytes. PBMC were kept in sion the results of the mapping based on chimeric PSMA culture at 37° C. and 5% CO, in RPMI medium (Gibco) with constructs demonstrate that bispecific single-chain antibody 10% FBS (Gibco). All procedures were performed according PSMA-P7 HLXI2CHL recognizes a membrane-proximaltar 30 to standard protocols (Current Protocols in Immunology; get epitope of human PSMA with <60 A membrane-distance Coligan, Kruisbeek, Margulies, Shevach and Strober; Wiley as defined herein. Interscience, 2002) 10.3 Epitope Mapping of the PSMA- and CD3-reactive The CytoTox-GloTM cytotoxicity assay (Kit from Bispecific Single Chain Antibody Molecule PSMA-D4 35 Promega) was used according to the instructions provided by HLXI2CHL the manufacturer. A PSMA-epitope with 260 A membrane-distance of bispe Each measurement was performed in triplicates with cific single chain antibody PSMA-D4 HLXI2CHL was con defined dilution series of purified PSMA specific bispecific firmed by epitope mapping using a peptide scanning antibodies (0.001 ng/ml to 250 ng/ml) and appropriate con approach. Peptide scanning uses overlapping peptides of a 40 trols to define spontaneous lysis (effector and target cells given protein and analyses antibody binding to immobilized without bispecific antibodies) and maximum lysis (addition peptides by enzyme-linked immunosorbent assays (ELISAS). of detergent digitonin to cells). 10000/well target and The epitope mapping experiments with the PSMA bispecific 100000/well effector cells were mixed in a defined ratio with single chain antibody PSMA-D4 HLXI2C HL were per effector cells in excess (E:T ratio of 10:1). After incubation at formed as described in detail in Bernard et al. 2004, J. Biol. 45 37° C. for 20-24 hours CytoTox-Glo cytotoxicity assay Chem..., 279: 24313-22 and Teeling et al. 2006, J. Immunol. reagent (AAF-GloTM; part of the CytoTox-Glo Kit from 177: 362-71. Promega) was added to all wells. The cells and the reagent In brief, 693 different 15-mer peptides were synthesized were mixed by orbital shaking and incubated at room tem perature for 1 hour. Determination of the number of dead cells that span the entire extracellular amino acid sequence of was performed Subsequently by measuring luminescence human PSMA and overlap with each neighboring 15-mer 50 peptide by 14 amino acids. These peptides were coated to with a plate reader (TECAN Spectrafluorometer). The mea ELISA wells in a 384-well plate format. For this series of sured signal correlates directly with the amount of lysed cells. experiments, the anti-PSMA schv fragment of bispecific On the basis of those measured values the cytotoxicity single chain antibody PSMA-D4 HLXI2CHL was produced values of every individual sample were calculated according in E. coli and used for ELISA as crude periplasmic extracts 55 to following formula: prepared as described herein. The schv antibody was incu bated with the peptides and specific binding detected using an anti-His antibody. Binding signals were measured in a 384 Wherein S is specific toxicity, V is the measured value, Bis well ELISA reader. As shown in the FIG.21 a clear maximum the average of blank values and M is the average of maximum signal was obtained for a peptide spanning over the amino 60 lysis. acids threonine 334 to threonine 339, which demonstrates the As shown in FIG. 22 the bispecific antibody PSMA-P7 presence of binding epitope of PSMA-D4 HLXI2CHL in the HLXI2CHL directed at a membrane-proximal target epitope region of amino acids 334 to 339 of human PSMA. As shown of human PSMA (as shown in Example 10.2) demonstrated in Table 1 threonine 334 and threonine 339 constitute a mem superior cytotoxic activity against human PSMA positive brane-distal epitope as defined herein. 65 target cells as compared to the bispecific single-chain anti In conclusion the results of the mapping based on peptide body PSMA-D4HLXI2CHL directed at membrane-distal tar scanning demonstrate that bispecific single-chain antibody get epitope of human PSMA (as shown in Example 10.3).