screening with synthetic multifunctional pores: Focus on biopolymers

Nathalie Sorde´ , Gopal Das*, and Stefan Matile†

Department of Organic Chemistry, University of Geneva, CH-1211 Geneva, Switzerland

Edited by Daniel Branton, Harvard University, Cambridge, MA, and approved August 19, 2003 (received for review May 14, 2003)

This report demonstrates that a single set of identical synthetic multifunctional pores can detect the activity of many different . Enzymes catalyzing either synthesis or degradation of DNA ( III or polymerase I), RNA (RNase A), polysaccha- rides (heparinase I, hyaluronidase, and galactosyltransferase), and proteins (papain, ficin, elastase, subtilisin, and pronase) are se- lected to exemplify this key characteristic of synthetic multifunc- tional pore sensors. Because anionic, cationic, and neutral sub- strates can gain access to the interior of complementarily functionalized pores, such pores can be the basis for very user- friendly screening of a broad range of enzymes.

here are compelling reasons to believe that the ‘‘universal Tenzyme sensor,’’ a user-friendly, noninvasive device that can detect all existing enzyme activities, belongs to the world of fiction and, despite functional proteomics, will never become reality (1, 2, ‡). However, it would be erroneous to conclude that Fig. 1. Enzymes detected with a single set of identical SMPs. a, Reported in efforts to maximize adaptability of noninvasive enzyme sensors ref. 1. to as many enzymes as possible are a simple waste of time. In this study, selected examples from biopolymer enzymology are used to demonstrate that a set of identical synthetic multifunctional The external LWV triads in SMP 1 further serve to destabilize pores (SMPs) can be used for noninvasive detection of the external surfaces formed by L-leucines (Ls) only as in SMP 2 (W, activity of many different enzymes (Fig. 1). L-tryptophan; V, L-valine). The advantageous characteristics of In brief, if substrates bind and block the pore better than unstable (rather than stable) pores for many practical applica- products, enzyme activity gradually removes the blocking agent tions of SMPs hint to the poorly defined importance of supramo- and the pore can open. On the other hand, if products bind and lecular dynamics for function. block the pores better than the substrates, enzyme activity Among many methods available to detect ‘‘enzyme-gated’’ gradually produces blocking agents that can block the pores (Fig. pore opening and closing, we used the emission of 5(6)- carboxyfluorescein (CF) as a convenient ‘‘naked-eye’’ read-out 2). Thus, if there is substantial molecular recognition of either (1). This method is based on the encapsulation of CF within substrate or product by the same pore, the only prerequisite for vesicles at self-quenching concentrations. This self-quenching detecting enzyme activities with SMP sensors is a simple method vanishes as soon as CF can move across an open pore and dilute to distinguish between blocked and unblocked pores. into the external medium, reporting unblocked pores by an As SMPs we introduced rigid-rod ␤-barrels 1 and 2 (Fig. 2C) ␤ increase in emission intensity. CF self-quenching therefore (2). Whereas various other rigid-rod -barrel SMPs are available remains unchanged as long as the pore is blocked by a substrate. to cope with particular sensing requirements, SMPs other than ␤ If substrates bind better than enzyme products, enzyme activity rigid-rod -barrels remain underexplored (3–9), particularly is reported by a gradual increase in CF emission (Figs. 2A and when compared with the remarkable progress made with bio- 3); conversely, if products bind better than substrates, enzyme engineered multifunctional pores as stochastic sensors of single activity is reported by a gradual decrease in CF emission analytes (10–13). (Fig. 2B). The syntheses of barrel-stave supramolecules 1 (14) and 2 (15) from commercial biphenyl and amino acid derivatives in 19 steps overall each have been described. The characteristics of pores This paper was submitted directly (Track II) to the PNAS office. ␤ formed by rigid-rod -barrels 1 and 2 in spherical and planar Abbreviations: SMP, synthetic multifunctional pore; [poly(dA,dT)]2, (deoxyadenylic acid, lipid bilayer membranes have also been reported (2, 14–16). thymidylic acid)2 copolymer duplex; CF, 5(6)-carboxyfluorescein; EYPC, egg yolk phospha- Even without optimization, the extraordinary permeabilizing tidylcholine; LUV, large unilamellar vesicle; ANTS, 8-amino-1,3,6-trisulfonate; DPX, p- xylenebis(pyrimidinium)bromide; poly(C), polycytidylic acid; PLE, poly-L-glutamate; PDE, activity of these p-octiphenyl ␤-barrels makes it possible to Ͼ poly-D-glutamate; PLR, poly-L-arginine. perform 300,000 enzyme assays per mg of polypeptide (1). *Present address: Department of Chemistry, Birla Institute of Technology and Science, L-histidine (H) and L-arginine (R) residues at the inner barrel Pilani 33303, India. surface account for the multifunctionality of pore 1. Lining the †To whom correspondence should be addressed. E-mail: [email protected]. ion-conducting pathway of SMP 1, these cationic residues rec- ‡The terms ‘‘sensor’’ and ‘‘detector’’ are poorly distinguishable with enzymes because ognize anionic substrates and products Ͼ10,000 times better substrates serve as ‘‘cosensors’’ to sense enzymes in mixed analytes, whereas enzymes than biological pores such as melittin (N.S. and S.M., unpub- serve as cosensors to sense substrates in mixed analytes. For example, ATP can be sensed lished work). In SMP 2, the ion-conducting pathway is function- in mixed analytes containing other by using SMP sensors with hexokinase and glucose as cosensors; glucose can be sensed in mixed analytes containing other carbohy- alized with anionic L-aspartates (Ds) instead. drates by using SMP sensors with hexokinase and ATP as cosensors; and hexokinase can be Both barrel-stave supramolecules comprise hydrophobic sensed in mixed analytes containing other enzymes by using SMP sensors with glucose and amino acid residues at the outer barrel surface to interact with ATP as cosensors (N.S. and S.M., unpublished work). the hydrophobic core of bilayer membranes, i.e., to form pores. © 2003 by The National Academy of Sciences of the USA

11964–11969 ͉ PNAS ͉ October 14, 2003 ͉ vol. 100 ͉ no. 21 www.pnas.org͞cgi͞doi͞10.1073͞pnas.2132894100 Downloaded by guest on September 28, 2021 Fig. 2. Enzyme sensing with SMPs shown as schematic oversimplification to outline the concept. (A) If substrates block the pore better than products, enzyme action produces reaction mixtures with gradually reduced ability to block the pore. (B) Enzyme-gated pore closing occurs if products block the pore better than substrates. (A and B) Rigid-rod ␤-barrel SMPs are depicted in axial view with ␤-strands as solid (backbone) and dotted (hydrogen bonds) lines. (C) Rigid-rod ␤-barrel SMPs 1 and 2 depicted as schematic cutaway suprastructures with peptide sequences B–F depicted dark on white for external and white on dark for internal amino acid residues (single-letter abbreviations). We reiterate that all reported suprastructures may be viewed as, at worst, simplifying but productive working hypotheses compatible with experimental data.

In the following, we first use the example of DNA exonuclease adaptability of the same enzyme sensor (i.e., SMP 1) to watch III to describe the methodology of noninvasive, fluorometric enzymes catalyzing either synthesis or degradation of DNA, enzyme sensing with SMPs in detail. Then, we demonstrate RNA, polysaccharides, and proteins work. Proteases are finally CHEMISTRY used to demonstrate access to cationic, anionic, and neutral substrates by using SMPs with complementary internal functionalization. Materials and Methods Materials. Egg yolk phosphatidylcholine (EYPC) was from Avanti Polar Lipids, and CF, buffers, salts, substrates, products, and enzymes were from Sigma or Fluka–Aldrich. Average molecular weights of polymers were calculated from the number ϭ ͞ n, of monomers per polymers in Table 1 {n KD(MONO) KD; (deoxyadenylic acid, thymidylic acid) copolymer duplex ϭ ͞ 3 3 2 3 2 3 2 3 [poly(dA,dT)]2, n KD(MONO) 2KD}. 1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 - Octakis(-OCH2CO-Leu-Arg-Trp-His-Val-NH2)-p-octiphenyl (1m) was synthesized as reported (14). Concentration of mono- mer stock solutions in DMSO were corroborated by UV-visible in diluted MeOH [␧(p-octiphenyl) ϭ 46.1 mMϪ1⅐cmϪ1 (320 nm); Varian Cary 1 Bio spectrophotometer]. Concentrations of pore Fig. 3. Fluorometric detection of enzyme activity with SMPs for 1 refer to tetrameric self-assemblies of 1m. Stock solutions of min(SUBSTRATE)͞ min(PRODUCT) Ͻ KD KD 1. The addition of an aliquot taken from the large unilamellar vesicles (LUVs) composed of EYPC and ʛ reaction mixture at the beginning of the reaction to EYPC-LUVs CF followed loaded with CF (i.e., EYPC-LUVs ʛ CF) were prepared by by SMP addition does not result in an increase in CF emission because CF efflux is blocked by substrates bound to the SMP. Performance of the same experi- extrusion (see Supporting Text, which is published as supporting ment with an aliquot taken at the end of the reaction gives an increase in CF information on the PNAS web site, www.pnas.org). The char- emission because the good substrate blocker is now converted into a poor acteristics are: 1.8 mM EYPC [phosphate analysis, Ն22 nM product blocker. LUVs (assuming diameter Յ 100 nm)]; inside, 50 mM CF͞10

Sorde´ et al. PNAS ͉ October 14, 2003 ͉ vol. 100 ͉ no. 21 ͉ 11965 Downloaded by guest on September 28, 2021 Table 1. Enzyme screening with SMPs

† ‡§ ¶ †࿣ ‡§ †† Enzyme Pore* Substrates (S) KD,M KD(MONO),M Products (P) KD,M S, M** SR

Ϫ10 Ϫ6 Ϫ3 Ϫ10 Ϫ3 1 DNA exonuclease III 1 [poly(dA,dT)]2 1.9 ϫ 10 2.3 ϫ 10 5Ј-dAMP 1.2 ϫ 10 1.9 ϫ 10 2.0 ϫ 10 5Ј-dTMP 2.5 ϫ 10Ϫ3 Ϫ5 Ϫ10 Ϫ10 1 2 DNA polymerase I 1 dATP 2.9 ϫ 10 [poly(dA,dT)]2 1.9 ϫ 10 1.9 ϫ 10 1.3 ϫ 10 Klenow fragment dTTP n.d. Pyrophosphate 1.0 ϫ 10Ϫ3 3 RNase A 1 poly(C) 1.9 ϫ 10Ϫ9 9.3 ϫ 10Ϫ7 3Ј-CMP 4.5 ϫ 10Ϫ3 1.9 ϫ 10Ϫ9 2.1 ϫ 10Ϫ4 4 RNase A 1 Polyuridylic acid 2.1 ϫ 10Ϫ10 5.8 ϫ 10Ϫ7 3Ј-UMP n.d. 2.1 ϫ 10Ϫ10 n.d. 5 RNase A 1 Polyaderylic acid 7.2 ϫ 10Ϫ10 9.7 ϫ 10Ϫ7 3Ј-AMP n.d. 7.2 ϫ 10Ϫ10 n.d. 6 Heparinase I 1 Heparin 4.7 ϫ 10Ϫ8 n.d. ⌬DiHs n.d. 4.7 ϫ 10Ϫ8 n.d. 7 Hyaluronidase 1 Hyaluronan 1.6 ϫ 10Ϫ10 1.0 ϫ 10Ϫ6 ⌬DiHAs n.d. 1.9 ϫ 10Ϫ10 n.d. 8‡‡ Galactosyl- 1 UDPGal Ͼ7 ϫ 10Ϫ2 5Ј-UDP 1.2 ϫ 10Ϫ3 1.2 ϫ 10Ϫ3 Ͼ4 ϫ 101 GlcNAc Ͼ5 ϫ 10Ϫ2 Gal␤1 3 GlcNAc Ͼ5 ϫ 10Ϫ2 Ϫ Ϫ Ϫ Ϫ Ϫ 9 Pronase 1 PLE 1.4 ϫ 10 8 1.3 ϫ 10 6 L-glutamate (E) Ͼ5 ϫ 10 4 1.4 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 10 Papain 1 PLE 1.4 ϫ 10 8 1.3 ϫ 10 6 L-glutamate (E) Ͼ5 ϫ 10 4 1.4 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 11 Ficin 1 PLE 1.4 ϫ 10 8 1.3 ϫ 10 6 L-glutamate (E) Ͼ5 ϫ 10 4 1.4 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 12 Elastase 1 PLE 1.4 ϫ 10 8 1.3 ϫ 10 6 L-glutamate (E) Ͼ5 ϫ 10 4 1.4 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 13 Subtilisin 1 PLE 1.4 ϫ 10 8 1.3 ϫ 10 6 L-glutamate (E) Ͼ5 ϫ 10 4 1.4 ϫ 10 8 Ͻ3 ϫ 10 8 Ϫ Ϫ Ϫ Ϫ Ϫ 14 Subtilisin 1 PDE 1.9 ϫ 10 8 1.6 ϫ 10 6 D-glutamate Ͼ5 ϫ 10 4 1.9 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 15 Papain 2 PLR 1.9 ϫ 10 8 1.3 ϫ 10 6 L-arginine (R) Ͼ5 ϫ 10 4 1.9 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 16 Papain 2 PLK 1.1 ϫ 10 8 1.5 ϫ 10 6 L-lysine (K) Ͼ5 ϫ 10 4 1.1 ϫ 10 8 Ͻ3 ϫ 10 3 Ϫ Ϫ Ϫ Ϫ Ϫ 17 n.d. 1 PLN 3.7 ϫ 10 8 3.5 ϫ 10 6 L-asparagine (N) Ͼ1 ϫ 10 2 3.7 ϫ 10 8 Ͻ4 ϫ 10 4

n.d., not determined. *See Fig. 2. †⌬DiH, heparin disaccharide product; ⌬DiHA, hyaluronan disaccharides; PLK, poly-L-lysine; PLN, poly-L-asparagine. ‡ KD ϭ global dissociation constant of [(supra)macro] molecular pore blockers. § All reported KD values are conditional to the employed method of detection (see text) and can vary substantially with experimental conditions. ¶ KD(MONO) ϭ average KD per monomer in (supra)macromolecular guests. ࿣Eventual oligomer products are not indicated for clarity only. **S ϭ sensitivity of pore sensor. †† SR ϭ selectivity requirement for pore sensor. ‡‡Data are from ref. 1. [Reproduced with permission from ref. 1 (Copyright 2002, AAAS, www.sciencemag.org).]

(t) mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid I values were converted into fractional pore blockage YE (YE (Hepes)͞10 mM NaCl, pH 7.4; outside, 10 mM Hepes͞107 mM ϭ 1 Ϫ {[I(tϭ0) Ϫ I(t)]͞[I(tϭ0) Ϫ I(tϭϱ)]}) and plotted as a function NaCl, pH 7.1. For the general procedure below, 0.1 ml of of reaction time t (Fig. 4D, filled circles). Control experiments EYPC-LUVs ʛ CF stock solution was added to 1.9 ml of buffer included repetition of the above-described set of experiments (10 mM Hepes͞107 mM NaCl, pH 6.5). with less (25 units͞ml) and no (Fig. 4D, open circles and X) heparinase (see Special Cases). Enzyme Sensing with SMP 1 (General Procedure). Heparin (20 ␮M) ͞ was incubated with heparinase I (50 units ml; EC 4.2.2.7, Dissociation Constants (KD Values) of Substrate͞Product Blockers. Flavobacterium heparinum) in buffer (10 mM Hepes͞107 mM Stock solutions of blockers (substrates and products) were NaCl, pH 7.5) at room temperature. After 30 min, 20 ␮lofthe prepared in buffer (10 mM Hepes͞107 mM NaCl) with pH reaction mixture was taken and added to 2 ml of gently stirred adjusted to 6.5 following the previously described fluoromet- EYPC-LUVs ʛ CF suspension in a thermostated fluorescence ric method without change (16). Then, x ␮l of blocker stock ␮ ͞ ͞Յ cuvette [90 M EYPC 50 mM CFINSIDE 200 nM heparin solution, (1,900 Ϫ x) ␮l of buffer (10 mM Hepes͞107 mM NaCl, ϭ (KD 47 nM) [see Dissociation Constants (KD Values) of pH 6.5), and 100 ␮l of EYPC-LUVs ʛ CF were mixed in a ͞ ͞ Substrate/Product Blockers] 10 mM Hepes 107 mM NaCl, pH fluorescence cuvette, pore 1 was added, and pore blockage was ␮ ␮ 6.5]. Then, pore 1 was added (20 lof25 M monomer in determined following the above-described general procedure ͞ ϭ ϫ 3 DMSO, 63 nM final; lipid pore 1.4 10 ) (see Calibration of (e.g., Fig. 5A,a–g). I(c) values were taken [I(c) ϭ I at maximal SMP 1), and the change in CF emission intensity It with time was emission intensity before lysis for pores in presence of blockers ␭ ϭ ␭ ϭ monitored ( em 510 nm, ex 495 nm; FluoroMax-2, Jobin- at concentration c], converted into fractional pore blockage Yvon, Longjumeau, France). Approximately 250 sec after pore Y {Y ϭ [I(cϭ0) Ϫ I(c)]͞[I(cϭ0) Ϫ I(cϭϱ)]}, and plotted as a function ␮ addition, 40 l of 1.2% aqueous Triton X-100 was added for lysis. of blocker concentration c to determine dissociation constants Fluorescence kinetics were normalized to fractional emission KD by using the Hill equation, intensity I by using ͓ ͑͞ Ϫ ͔͒ ϭ ϫ Ϫ ϫ ϭ ͓͑ Ϫ ͒͑͞ Ϫ ͔͓͒͑͞ Ϫ ͒͑͞ Ϫ ͔͒ log Y 1 Y nHill log[blocker] nHill log KD I It I0 Iϱ I0 IMAX I0 Iϱ I0 , [1] [2] ϭ ϭ where I0 It at pore addition, Iϱ It at saturation after lysis, and ϭ ϭ Ͼ ϭ IMAX It at maximal emission intensity with blocker-free pore (nHill Hill coefficient; e.g., Fig. 5B). For macro- (n 1, m before lysis (see Calibration of SMP 1). From the normalized 1), supra- (n ϭ 1, m Ͼ 1), and supramacromolecules (n Ͼ 1, m Ͼ (tϭ0.5) (tϭ0.5) ϭ curves, I was recorded [I ϭ maximal I before lysis for 1), the KD per monomer was calculated by using KD(MONO) ϫ ϫ pores in presence of an aliquot reaction mixture taken 0.5 h after KD n m (e.g., Fig. 5B, open circles). The sensitivity S of SMP ϭ min min the beginning of the reaction]. This procedure was repeated with sensors was approximated as S KD , where KD is the lowest 20 ␮l of reaction mixture taken after 2, 4, 6, 8, and 11 h of KD among all involved compounds for a given enzyme. Initial (tϭ2) (tϭ4) (tϭ6) (tϭ8) (tϭ11) Ϸ ϫ min reaction time to obtain I , I , I , I , and I .All substrate concentrations, 100 to 1,000 KD , were ideal to

11966 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.2132894100 Sorde´ et al. Downloaded by guest on September 28, 2021 Fig. 4. Real-time enzyme screening monitored as change in fractional blockage YE of pores 1 (A–K) and 2 (L) with reaction time t exemplified for DNA polymerase I Klenow fragment (A), DNA exonuclease III (B), RNase A (C), heparinase I (D), hyaluronidase (E), galactosyltransferase (F) [reproduced with permission from ref. 1 (Copyright 2002, AAAS, www.sciencemag.org)], pronase (substrate: PLE) (G), papain (substrate: PLE) (H), ficin (substrate: PLE) (I), elastase (substrate: PLE) (J), subtilisin {substrate candidates: x ϭ [PLE]͞([PLE] ϩ [PDE]) ϭ 1.00 (filled circles), 0.75 (open circles), 0.50 (open squares), and 0.00 (filled squares) at [PLE] ϩ [PDE] ϭ constant} (K), and papain (substrate: PLR) (L) at low (open circles) and high (filled circles) enzyme concentrations against controls (X) except for K. Substrates (S) are as in Table 1.

min Յ Յ min ͞ reach KD c 10KD after the 100-fold dilution before [8-amino-1,3,6-trisulfonate (ANTS) p-xylenebis(pyrimidinium)- detection. The selectivity requirement SR was approximated as bromide (DPX)] for cationic substrates (Fig. 4L). Sensing with SMP 2 and specific conditions for each enzyme detected with ϭ min(SUBSTRATE)͞ min(PRODUCT) SR KD(MONO) KD(MONO) , [3] SMP 1 are reported in Supporting Text.

min(SUBSTRATE) where KD(MONO) is the lowest KD(MONO) among all Results and Discussion min(PRODUCT) substrates, and KD(MONO) is the lowest KD(MONO) among To detect enzyme activity with SMPs by using fluorometric all products of a specific enzyme. real-time detection, enzymes and substrates are incubated under appropriate conditions. After a given period, an aliquot of the Calibration of SMP 1. The above-described general procedure was reaction mixture is removed and added to vesicles containing performed in the absence of reaction mixtures [or individual entrapped CF. Then, the pore sensor is added, and CF release blockers as described in Dissociation Constants (KD Values) of is determined and compared with CF release at the beginning of Substrate͞Product Blockers]. The concentration of blocker-free the reaction (e.g., Fig. 5C, b versus a). This procedure (i.e., SMP 1 was adjusted to cause nearly complete CF efflux (com- removal of an aliquot of reaction mixture, addition to CF-loaded pared with lysis by Triton X-100) within Ϸ150 sec (Fig. 5A, a). vesicles followed by addition of the SMP, and determination of Calibration using Eq. 1 gave the IMAX required for data analysis. CF release) is repeated in meaningful intervals as the reaction proceeds (e.g., Fig. 5C,b–f). The individual experiments are then Special Cases. All enzymes listed in Table 1 and Fig. 4 were summarized and plotted as fractional pore blockage YE as a Ϫ studied following this general procedure by using initial sub- function of reaction time t. The obtained YE t profile reveals Ϸ ϫ min strate concentrations c of 100 to 1,000 KD and the following enzyme activity as increasing (for product blockers) or decreas- ϭ (tϭ0) Ϫ (t) ͞ (tϭ0) Ϫ (tϭϱ) exceptions: (i) YE [I I ] [I I ] for product ing (for substrate blockers) pore blockage with time (e.g., Fig. blockage (Fig. 4 A and F) and (ii) SMP 2 and EYPC-LUVs ʛ 4B). To apply this method to a given enzyme, the ability of all involved compounds to block the pore sensor must be deter- mined first. Then, concentrations must be adjusted to assure CHEMISTRY detectability of consumption͞production of the best substrate͞ product blocker involved. In the following, this calibration process is described in detail by using DNA exonuclease III as an example. DNA exonuclease III catalyzes the sequential hydrolysis from the 3Ј terminus of duplex DNA into 5Ј-monophos- phates (Table 1) (17–19). The [poly(dA,dT)]2 is widely used as a model substrate of exonuclease III. Because of its alternating, repeating sequence, duplex [poly(dA,dT)]2 remains double- stranded even after extensive degradation and hydrolysis by exonuclease III continues almost to completion (17). To detect DNA exonuclease III with SMP 1, the same pore must recognize either substrate [poly(dA,dT)] or the products Fig. 5. Original data for detection of the hydrolysis of [poly(dA,dT)] by DNA 2 2 2Ј-deoxyadenosine 5Ј-monophosphate (5Ј-dAMP) and 2Ј- exonuclease III with SMP 1.(A) Characterization of substrate [poly(dA,dT)]2 as Ј Ј pore blocker. Fractional CF emission I (␭em, 510 nm, ␭ex, 495 nm) as a function deoxythymidine 5 -monophosphate (5 -dTMP). To characterize of time after addition of 0 (a), 0.09 (b), 0.17 (c), 0.29 (d), 0.73 (e), 1.46 (f), and this supramolecular host–guest chemistry quantitatively, LUVs 2.92 (g) nM [poly(dA,dT)]2 and pore 1 (63 nM, arrow) to EYPC-LUVs ʛ CF. (B) composed of EYPC were loaded with CF (Fig. 3). CF efflux from Dependence of fractional pore blockage Y on concentration c of substrate and EYPC-LUVs ʛ CF through SMP 1 was easily detectable as an product blockers {[poly(dA,dT)]2: filled circles, per duplex; open circles, per increase in CF emission after pore addition (Fig. 5A, a). Neg- nucleotide; data are from A; filled squares, dAMP; open squares, dTMP}. (C) ligible CF efflux was observed without pore. The presence of Detection of exonuclease activity. Twenty microliters of reaction mixture [0.83 units/ml exonuclease͞37 nM [poly(dA,dT)] were taken after a reaction time of increasing concentrations of [poly(dA,dT)]2 reduced CF efflux 2 mediated by pore 1 (Fig. 5A,b–g). A plot of blocker concen- 0 min and added to 2 ml of EYPC-LUVs ʛ CF {Ϸ0.37 nM [poly(dA,dT)]2 final}; then, the change in CF emission after pore addition was measured as described tration as a function of the fractional blockage Y of tetramer 1 for A (a). The same experiment was performed with 20-␮l reaction mixtures gave the dose–response isotherm for [poly(dA,dT)]2 (Fig. 5B, taken after reaction times of 5 (b), 10 (c), 20 (d), 30 (e), and 130 (f) min. filled circles). Hill analysis of this isotherm yielded a global

Sorde´ et al. PNAS ͉ October 14, 2003 ͉ vol. 100 ͉ no. 21 ͉ 11967 Downloaded by guest on September 28, 2021 ϭ Ϸ Ϸ dissociation constant KD 185 pM (Table 1, entry 1). Recog- SR 13 together with a superb sensitivity S 185 pM implied nition of picomolar concentrations of substrate [poly(dA,dT)]2 unproblematic detectability of polymerase I (Klenow fragment) implied superb sensitivity S for the detection of DNA exonu- as enzyme-gated pore closing. This was corroborated experi- clease III with sensor 1 (Table 1). In an unoptimized multiwell mentally (Fig. 4A). end-point assay with a sample volume of 100 ␮l (1), for example, RNase A, a pyrimidine specific RNase from bovine pancreas, Ϸ O 5Ј O 2Ј the detection limit was calculated to 19 fmol of [poly(dA,dT)]2. catalyzes cleavage of the P O bond and the P O bond in Application of the same protocol to exonuclease products cyclic intermediates (24). Although cleavage of RNA homopoly- revealed 107 times poorer affinity of dAMP and dTMP to pore mers by wild-type RNase A is distributive rather than processive 1 (Fig. 5B, squares). To determine the selectivity requirement SR as with the above-mentioned DNA exonuclease III, nucleotide ϭ Ј for SMP 1 to detect [poly(dA,dT)]2 hydrolysis, a KD(MONO) 2.3 3 -monophosphates can be assumed as ultimate products of ␮M per monomer was calculated for supramacromolecule complete hydrolysis. Not surprisingly, substrate polycytidylic Ͼ [poly(dA,dT)]2 (Fig. 5B, open circles). Comparison of this value acid [poly(C)] turned out to block SMP 1 5 orders of magnitude Ϸ Ј Ј with the KD values of the products revealed SR 0.002. This better than cytidine 3 -monophosphate (3 -CMP) (Table 1, entry ϽϽ Ϸ Ϸ selectivity requirement SR 1 corroborated that differentia- 3). Predictable from SR 0.0002 and S 2 nM, unproblematic tion of substrate and products of DNA exonuclease III by SMP and sensitive detection of poly(C) hydrolysis by RNase A as 1 would be unproblematic and that enzymatic activity would be enzyme-gated pore opening was confirmed experimentally (Fig. detectable as enzyme-gated pore opening (Figs. 2A and 3). 4C). The ability of homopolymers poly(C), polyuridylic acid, and Fluorometric real-time detection by SMPs was used to mon- polyadenylic acid to block SMP 1 was quite similar (Table 1, itor DNA exonuclease III. Alternative modes of detection have entries 3–5). Consistent with the literature (24), RNase A sensing been described elsewhere [e.g., multiwell screening and end- with SMP 1 indicated that hydrolysis of polyuridylic acid is point detection (1)] or remain to be explored (e.g., continuous almost as fast as poly(C), whereas that of polyadenylic acid was and electric detection in black or supported bilayers). For much slower (data not shown). real-time detection of exonuclease III, 37 nM [poly(dA,dT)]2 duplex and 0.833 units͞ml of the enzyme were incubated at pH Polysaccharides. Heparinase I catalyzes the eliminative depoly- 8.0 and 30°C. At different points in time, 20 ␮l of the reaction merization of heparin, a biologically important and chemically mixture were taken, added to 2 ml of EYPC-LUVs ʛ CF, and unique polysaccharide that is widely used as an anticoagulant tested for blockage of the subsequently added pore 1 (Fig. 5C). drug (25). Spectroscopic detection of either heparin or hepari- Consistent with conversion of the good blocker [poly(dA,dT)]2 nase is not straightforward, because the substrate contains no (Fig. 5B, circles) into the poor blockers dAMP and dTMP (Fig. chromophore (26, 27). The invention of assays for noninvasive, 5B, squares), the ability of the reaction mixture to block SMP 1 high-throughput detection of heparinase, however, is desirable decreased with increasing reaction time (Fig. 4B). to contribute to the search for competitive inhibitors as potential heparin mimics for use in cancer therapy (28, 29). Sensitive Polynucleotides. DNA exonuclease III was selected as an example detection of heparinase I activity as enzyme-gated pore opening for fluorometric enzyme sensing with SMPs because of its broad (Fig. 4D) was as unproblematic as expected from superb mo- scientific significance (Table 1, entry 1) (17–19). Among diverse lecular recognition of the polyanionic heparin substrate by applications in biotechnology, the use of exonuclease III in gene polycation 1 (Table 1, entry 6). Pore blockage by the ultimate sequencing (18, 19) may be particularly noteworthy because mixture of heparin disaccharide products (⌬DiHs) was not combination with exonuclease III may ultimately be necessary to determined. overcome the intrinsic resolution limits of single-gene sequenc- Hyaluronan, discovered 70 years ago by Meyer and Palmer ing with pores (10, 12, 13). Experimental evidence for detect- (38), is a viscous, acidic glycosaminoglycan with structural and ability of exonuclease III with SMP 1 is detailed above (Figs. 4B signaling functions mainly in the extracellular matrix (30). Hy- and 5). aluronidase converts this polysaccharide into hyaluronan disac- DNA polymerases catalyze the template-directed incorpora- charides (⌬DiHAs). The ability of hyaluronan to block SMP 1 tion of deoxyribonucleotide monophosphates from 5Ј- was substantial (Table 1, entry 7). Unproblematic detectability deoxyribonucleotide triphosphate substrates to the 3Ј-hydroxyl of hyaluronidase activity as enzyme-gated pore opening there- terminus of a growing DNA strand (19–23). Because of their fore was not surprising (Fig. 4E). central role in biology, biochemistry, and biotechnology (19, 20), Detectability of enzymes involved in polysaccharide synthesis polymerases have been the subject of extensive structural and with SMP 1 has been reported (Table 1, entry 8, and Fig. 4F) (1). mechanistic studies. DNA polymerase I enzymes, characterized The invention of adaptable assays compatible with glycosyltrans- by multidomain architecture including 5Ј 3 3Ј , are ferase screening are of high interest in fields of application as involved in the repair of DNA lesions. The Klenow fragment of diverse as glycobiology, drug discovery, and the invention of polymerase I, shaped like a right hand that can ‘‘hold’’ the automated polysaccharide syntheses (31–33). growing DNA duplex with palm, fingers, and thumb, contains the polymerase domain. Polypeptides. Initial use of poly-L-glutamate (PLE) as the model Similar to that for the complementary DNA , duplex polypeptide to explore adaptability of pore 1 to detect the [poly(dA,dT)]2 provides a convenient model for studying DNA activity of proteases was an obvious choice for several reasons. polymerases because poly(dA,dT) strands can serve both as PLE has been used to explore the usefulness of dipole–potential templates and primers (23). Different than DNA nucleases, interaction for ␣-helix recognition by SMPs in polarized bilayers however, SMP sensors have to differentiate between nucleotide (34). PLE has been used further to demonstrate that molecular triphosphate substrates rather than monophosphate products to recognition by SMP 1 exceeds that by the biological pore melittin ϭ ␮ Ͼ detect DNA polymerases. A KD 29 M was determined for 10,000 times (N.S. and S.M., unpublished work). 2Ј-deoxyadenosine 5Ј-triphosphate (5Ј-dATP) as a blocker of Unproblematic detectability of monitor hydrolysis of the SMP 1 (Table 1, entry 2). The ability of substrate 5Ј-dATP to excellent blocker PLE by proteases was indicated by the fact that block pore 1 was Ϸ5 orders of magnitude weaker than that of the L-glutamate did not block SMP 1 under meaningful experimen- supramacromolecular product; that of 2Ј-deoxythymidine 5Ј- tal conditions (Table 1, entry 9). Decreasing ability of reaction triphosphate (5Ј-dTTP) was assumed to be in the range of mixtures to block pore 1 with increasing reaction time was as 5Ј-dATP. Contributions from coproduct pyrophosphate to pore expected for PLE hydrolysis by protases (35–37) such as pronase, blockage were irrelevant. The resulting selectivity requirement papain, ficin, elastase, and subtilisin (Fig. 4 G–K).

11968 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.2132894100 Sorde´ et al. Downloaded by guest on September 28, 2021 The binding of poly-D-glutamate (PDE) and PLE to SMP 1 detection of pore activity with CF and ANTS͞DPX assay, was nearly identical (Table 1, entries 9 and 14). In clear contrast however, was identical. According to decreasing efflux of either to PLE, however, treatment of PDE with subtilisin did not result intravesicular fluorophore ANTS or quencher DPX with in- in the opening of pore 1 (Fig. 4K, filled squares). This clear creasing polypeptide concentration, polycations poly-L-lysine difference between PLE and PDE suggested that the latter was and PLR blocked anionic SMP 2 about as efficiently as poly- not converted by subtilisin. This finding was as expected from the anion PLE blocked cationic SMP 1 (Table 1, entries 15 and 16). literature (36) and the use of this endoprotease in kinetic The opening of pore 2 during PLR hydrolysis by papain con- resolutions (35). However, treatment of a racemic mixture of firmed detectability of enzymes with cationic substrates by using PLE and PDE with subtilisin for up to 4 days did not produce anionic pore sensors (Fig. 4L). reaction mixtures with reduced ability to block pore 1 (Fig. 4K, Experimental evidence for detectability of enzymes with Ͼ open squares). Fractional pore blockage YE 0.6 was observed cationic substrates by using anionic SMP 2 and of those with for a 3:1 mixture of PLE and PDE after incubation with subtilisin anionic substrates by using cationic SMP 1 did not answer the Յ for hours, a value clearly above the YE 0.25 expected for full question about substrates without charge. We investigated poly- PLE conversion (Fig. 4K, open circles). One interpretation of L-asparagine for this purpose. This water-soluble polypeptide this apparently poor degradation of PLE in PLE͞PDE mixtures was found to hinder CF efflux through SMP 1 almost as is that the conversion of polyglutamates by subtilisin is stereo- efficiently as the anionic polypeptide PLE (Table 1, entry 17). selective, whereas their low-affinity binding (37) is not. Current Conclusion studies on the use of SMP sensors to screen for enzyme inhibitors focus on examples of medicinal relevance. The reported examples confirm that a single set of identical SMPs can be used to detect the activity of many different Conversion of cationic polypeptides such as poly-L-lysine or enzymes in a straightforward manner. Such broad adaptability of poly-L-arginine (PLR) by proteases was not detectable with pore 1 because polycationic substrates are poorly recognized by noninvasive and user-friendly enzyme sensors is ideal for prac- tical applications. polycationic pores. The cation-selective (2, 16) polyanionic pores formed by rigid-rod ␤-barrel 2 with internal aspartates, there- We thank P. Talukdar and D. Gerard for assistance in organic synthesis; fore, were evaluated as sensors for enzymes with cationic N. Sakai, D. Branton, and two anonymous referees for very helpful substrates. Because SMP 2 is impermeable for anion CF (16), the advice; and the financial support of Swiss National Science Foundation less ion-selective but also less sensitive ANTS͞DPX assay was Grant 2000-064818.01 and National Research Program ‘‘Supramolecular applied (15). Except for these differences, the fluorometric Functional Materials’’ 4047-057496.

1. Das, G., Talukdar, P. & Matile, S. (2002) Science 298, 1600–1602. 19. Stephan, J., Dorre, K., Brakmann, S., Winkler, T., Wetzel, T., Lapczyna, M., 2. Sakai, N. & Matile, S. (2003) Chem. Commun. (Advance Article), DOI: Stuke, M., Angerer, B., Ankenbauer, W., Foldes-Papp, Z., et al. (2001) 10.1039͞b303649a. J. Biotechnol. 86, 255–267. 3. Gokel, G. W. & Mukhopadhyay, A. (2001) Chem. Soc. Rev. 30, 274–286. 20. Levene, M. J., Korlach, J., Turner, S. W., Foquet, M., Craighead, H. G. & 4. Scrimin, P. & Tecilla, P. (1999) Curr. Opin. Chem. Biol. 3, 730–735. Webb, W. W. (2003) Science 299, 682–686. 5. Kirkovits, G. J. & Hall, C. D. (2000) Adv. Supramol. Chem. 7, 1–47. 21. Steitz, T. A. (1999) J. Biol. Chem. 274, 17395–17398. 6. Terrettaz, S., Ulrich, W.-P., Guerrini, R., Verdini, A. & Vogel, H. (2001) 22. Carroll, S. S. & Benkovic, S. J. (1990) Chem. Rev. (Washington, D.C.) 90, Angew. Chem. Int. Ed. Engl. 40, 1740–1743. 1291–1307. 7. Wyman, T. B., Nicol, F., Zelphati, O., Scaria, P. V., Planck, C. & Szoka, F. C., 23. Klenow, H. & Henningsen, I. (1970) Proc. Natl. Acad. Sci. USA 65, 168–175. Jr. (1997) Biochemistry 36, 3008–3017. 24. Raines, R. T. (1998) Chem. Rev. (Washington, D.C.) 98, 1045–1066. 8. Wang, D., Guo, L., Zhang, J., Jones, L. R., Chen, Z., Pritchard, C. & Roeske, 25. Capila, I. & Linhardt, R. J. (2002) Angew. Chem. Int. Ed. Engl. 41, 390–412. R. W. C. (2001) J. Pept. Res. 57, 301–306. 26. Czarnik, A. W. (1994) Acc. Chem. Res. 27, 302–308. 9. Sanchez-Quesada, J., Isler, M. P. & Ghadiri, M. R. (2002) J. Am. Chem. Soc. 27. Zhong, Z. & Anslyn, E. V. (2002) J. Am. Chem. Soc. 124, 9014–9015. 124, 10004–10005. 28. Petitou, M., He´rault, J.-P., Bernat, A., Driguez, P.-A., Duchaussoy, P.,

10. Bayley, H. & Cremer, P. S. (2001) Nature 413, 226–230. Lormeau, J.-C. & Herbert, J. M. (1999) Nature 398, 417–422. CHEMISTRY 11. Nestorovich, E. M., Danelon, C., Winterhalder, M. & Bezrukov, S. M. (2002) 29. Neuger, L., Ruge, T., Makoveichuk, E., Vlodavsky, I. & Olivecrona, G. (2001) Proc. Natl. Acad. Sci. USA 99, 9789–9794. Atherosclerosis (Shannon, Irel.) 157, 13–21. 12. Deamer, D. W. & Branton, D. (2002) Acc. Chem. Res. 35, 817–825. 30. McDonald, J. & Hascall, V. C. (2002) J. Biol. Chem. 277, 4575–4596. 13. Vercoutere, W. & Akeson, M. (2002) Curr. Opin. Chem. Biol. 6, 816–822. 31. Patenaude, S. I., Seto, N. O. N., Borisova, S. N., Szpacenko, A., Marcus, S. L., 14. Talukdar, P., Sakai, N., Sorde´, N., Gerard, D., Cardona, V. M. F. & Matile, S. Palcic, M. M. & Evans, S. V. (2002) Nat. Struct. Biol. 9, 685–690. (2003) Bioorg. Med. Chem., in press. 32. Wymer, N. & Toone, E. J. (2000) Curr. Opin. Chem. Biol. 4, 110–119. 15. Das, G. & Matile, S. (2002) Proc. Natl. Acad. Sci. USA 99, 5183–5188. 33. Hurtley, S., Service, R. & Szuromi, P. (2001) Science 291, 2337–2378. 16. Das, G., Onouchi, H., Yashima, E., Sakai, N. & Matile, S. (2002) Chembiochem 34. Sorde´, N. & Matile, S. (2002) J. Supramol. Chem. 2, 191–199. 3, 1089–1096. 35. Hedstrom, L. (2002) Chem. Rev. (Washington, D.C.) 102, 4429–4906. 17. Richardson, C. C., Lehman, I. R. & Kornberg, A. (1964) J. Biol. Chem. 239, 36. Neumann, H., Sharon, N. & Katchalski, E. (1962) Nature 195, 1002–1003. 251–258. 37. Miller, W. G. (1964) J. Am. Chem. Soc. 86, 3913–3918. 18. Guo, L.-H. & Wu, R. (1983) Methods Enzymol. 100, 60–97. 38. Meyer, K. & Palmer, J. (1934) J. Biol. Chem. 107, 629–634.

Sorde´ et al. PNAS ͉ October 14, 2003 ͉ vol. 100 ͉ no. 21 ͉ 11969 Downloaded by guest on September 28, 2021