Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Cell softness prevents cytolytic T cell killing of tumor-repopulating cells

2 Yuying Liu1,2,10*, Tianzhen Zhang1,10, Haizeng Zhang3,10, Jiping Li4, Nannan Zhou1,

3 Roland Fiskesund1,5, Junwei Chen6, Jiadi Lv1, Jingwei Ma7, Huafeng Zhang7, Ke

4 Tang7, Feiran Cheng1, Yabo Zhou1, Xiao-hui Zhang8, Ning Wang6,9, Bo Huang1,2,7*

5 1Department of Immunology & National Key Laboratory of Medical Molecular

6 Biology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences

7 (CAMS) & Peking Union Medical College, Beijing 100005, China

8 2Clinical Immunology Center, CAMS, Beijing 100005, China

9 3National Cancer Center/Cancer Hospital, CAMS, Beijing 100005, China

10 4Beijing Smartchip Microelectronics Technology Company Limited, Beijing 100192,

11 China

12 5Karolinska Institutet Medical School, S-171 77, Stockholm, Sweden.

13 6Laboratory for Cellular Biomechanics and Regenerative Medicine, College of Life

14 Science and Technology, Huazhong University of Science and Technology, Wuhan,

15 Hubei 430074, China.

16 7Department of Biochemistry & Molecular Biology, Tongji Medical College,

17 Huazhong University of Science & Technology, Wuhan 430030, China

18 8Peking University People's Hospital, Peking University Institute of Hematology,

19 Beijing 100044, China

20 9Department of Mechanical Science and Engineering, College of Engineering,

21 University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA

22 10These authors contributed equally.

23 Running title: Cell softness prevents cytolytic T cell killing of TRC

24 Conflict of interest: The authors declare no competing financial interests.

25 Author contributions: B.H. conceived the project. Y.L., T.Z., H.Z., N.Z., J.L., R.F., X.L.,

1

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 J.C., J.L., J.M., H.Z., K.T., S.M., F.C., Y.Z. and Y.F. performed the experiments. B.H.,

2 Y.L., X.Z., H.Z. and N.W. developed methodology. B.H., Y.L., H.Z. and T.Z. performed

3 data analysis. B.H., Y.L. and N.W. wrote the manuscript.

4 *Corresponding author: Bo Huang, Chinese Academy of Medical Sciences (CAMS)

5 & Peking Union Medical College, NO. 5, Dongdan Santiao, Dongcheng District,

6 Beijing, P. R. China; Phone: +86 01069156464; E-mail:[email protected];

7 Yuying Liu, Chinese Academy of Medical Sciences (CAMS) & Peking Union

8 Medical College, NO. 5, Dongdan Santiao, Dongcheng District, Beijing, P. R. China;

9 Phone: +86 01069156464; E-mail:[email protected]

10

2

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Abstract

2 Biomechanics is a fundamental feature of a cell. However, the manner by which

3 actomysin tension affects tumor immune evasion remains unclear. Here we show that

4 although cytotoxic T lymphocytes (CTL) can effectively destroy stiff differentiated

5 tumor cells, they fail to kill soft tumor-repopulating cells (TRC). TRC softness

6 prevented membrane pore formation caused by CTL-released perforin. Perforin

7 interacting with nonmuscle heavy chain 9 transmitted forces to less F- in

8 soft TRC, thus generating an inadequate contractile force for perforin pore formation.

9 Stiffening TRC allowed perforin the ability to drill through the membrane, leading to

10 CTL-mediated killing of TRC. Importantly, overcoming mechanical softness in

11 human TRC also enhanced TRC cell death caused by human CTL, potentiating a

12 mechanics-based immunotherapeutic strategy. These findings reveal a

13 mechanics-mediated tumor immune evasion, thus potentially providing an alternative

14 approach for tumor immunotherapy.

15 Significance

16 Tumor-repopulating cells evade CD8+ cytolytic T cell killing through a mechanical

17 softness mechanism, underlying the impediment of perforin pore formation at the

18 immune synapse site.

19

20 Key words: perforin, pore formation, softness, nonmuscle myosin heavy chain 9,

21 mechanical force

3

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Introduction

2 Despite the unprecedented success of PD-1 blockade and CAR-T

3 therapy in the clinic (1, 2), relapse is still common in cancer patients that initially

4 responded with extraordinary treatment outcomes (3-5). This indicates that a

5 subpopulation of tumor cells may escape from, or be intrinsically resistant to, T cell

6 killing. However, to date, the underlying mechanism remains a mystery. Some of the

7 proposed mechanisms, such as the downregulation of MHC class I, upregulation of

8 inhibitory receptors and/or the secretion of soluble immunosuppressive molecules

9 (6-8), may explain how tumor cells can avoid T cell recognition and a subsequent

10 attack, but none of these theories directly address the possibility that a certain subset

11 of tumor cells can resist T cell-mediated killing using different mechanisms. For

12 instance, our previous studies have demonstrated that the same CTLs can effectively

13 kill bulk tumor cells but not a small subset of undifferentiated tumor cells which we

14 termed tumor-repopulating cells (TRC) due to their ability to rapidly form tumors (9,

15 10). Interestingly, both bulk tumor cells and TRCs can be recognized by CTLs (9),

16 suggesting that the resistance to T cell killing is mediated by mechanisms beyond the

17 impairment of T cell function. IFN-γ is an important cytokine in T cell-mediated

18 antitumor immunity (11, 12). Besides, perforin and granzymes are another important

19 approach used by CTLs to kill tumor cells. Following recognition, T cells release

20 perforin which forms pores in the target cell membrane allowing granzymes to enter

21 the cell, thus initiating apoptosis (13, 14). Whether tumor cells can evade T cell

22 killing at this stage has not been well documented.

23 Stiffness is an inherent feature of a cell, which is mainly provided by F-

24 filaments (15). Different cell types display varying levels of stiffness, which matches

25 the stiffness of the local extracellular matrix, allowing the cells to properly sense and

4

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 respond to the surrounding mechanical microenvironments (16, 17). Based on this

2 understanding, we have developed a mechanics-based culture system using 3D soft

3 fibrin gels that can selectively amplify TRCs from primary tumor tissues and cancer

4 cell lines in both murine and human settings (9, 10). These 3D gel-amplified TRCs

5 display typical characteristics of stem cell-like cancer cells (CD133+ or ALDH+ and

6 forming a tumor with as few as 5 cells) and are much softer than their differentiated

7 counterparts (9, 18, 19). This latter observation led us to investigate whether the

8 differential killing of TRCs and differentiated tumor cells by CTLs was related to

9 their stiffness. It is likely that in addition to a series of chemical reactions,

10 perforin-mediated pore formation also involves a more physical, mechanical process.

11 We hypothesized that a certain level of cellular stiffness is required to support pore

12 formation by the perforin molecules and that TRCs may use their mechanical softness

13 (which is the inverse of stiffness) to impede perforin drilling a pore, thus escaping T

14 cell-mediated death. In this study, we demonstrate that cell softness is a fundamental

15 mechanism TRCs use to evade T cell killing, and that downregulation of cell softness

16 (i.e., upregulation of cell stiffness) enhances CTL killing of the TRCs by allowing

17 perforin-mediated pore formation.

5

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Materials and methods

2 Animals

3 Female C57BL/6 mice or NSG mice, 6-8 weeks old, were purchased from the

4 Center of Medical Experimental Animals of Chinese Academy of Medical Science

5 (Beijing, China). OT-I transgenic mice were gifted by Dr. Hui Zhang (Sun Yat-Sen

6 University). Pmel-1 transgenic mice, which carry a rearranged T cell receptor specific

7 for the mouse homologue (pmel-17) of human SILV (gp100) (20), were presented by

8 Dr. Ying Wan (Third Military Medical University). These animals were maintained in

9 the Animal Facilities of Chinese Academy of Medical Science under pathogen-free

10 conditions. All studies involving mice were approved by the Animal Care and Use

11 Committee of Chinese Academy of Medical Science.

12 Cell lines

13 Mouse tumor cell lines B16, OVA-B16 (melanoma), MC38 (colon cancer) and 4T1

14 (breast cancer) and human tumor cell lines A375 (melanoma), HepG2 (hepatocellular

15 cancer) and MCF-7 (breast cancer) were purchased from China Center for Type

16 Culture Collection (Beijing, China) and cultured in RPMI 1640 (Thermo Scientific)

17 with 10% fetal bovine serum (FBS) (Gibco, USA). Cells were tested

18 for mycoplasma detection, inter-species cross contamination and authenticated by

19 isoenzyme and short-tandem repeat (STR) analyses in Cell Resource Centre of Peking

20 Union Medical College before the study. Cell lines used in the experiments were

21 within 20 passages.

22 Human samples

23 Resected human melanoma or colon cancer tissues were obtained from patients at

24 the Peking Union Medical College Hospital or National Cancer Center/Cancer

25 Hospital. Ethical permission was granted by the Clinical Trial Ethics Committee of

6

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Peking Union Medical College Hospital or National Cancer Center/Cancer Hospital.

2 All patients provided written informed consent to participate in the study.

3 Tumor cells cultured on 2D rigid dish or in 3D fibrin gels

4 For conventional 2D cell culture, tumor cells were maintained in a rigid dish with

5 complete culture medium. TRC culture was performed according to our previously

6 published protocol (9, 10). In brief, fibrinogen (Searun Holdings Company, Freeport,

7 ME) was diluted into 2 mg/ml with T7 buffer (pH 7.4, 50 mM Tris, 150 mM NaCl).

8 Then, a 1:1 (volume) mixture of fibrinogen and cell solution was made. 250 µl

9 cell/fibrinogen mixtures were seeded into 24 well-plate and mixed well with

10 pre-added 5 µl thrombin (0.1 U/µl, Searun Holdings Company). After 30 min

11 incubation at 37°C, these cells were supplemented with 1 ml completed culture

12 medium. After 5 days culture, dispase II was added to and digested the cultured gels

13 for 10 min at 37°C. The spheroids were harvested and further digested with 0.25%

14 trypsin for 3 min to obtain a single cell suspension for the following experiments.

15 Recombinant perforin

16 PFR and 6xHis tag at the C-terminus was subcloned into pFastBac1 (presented by

17 Dr. Feng Shao) to construct the recombinant plasmid of pFastBac1-PRF1. Then,

18 DH10Bac E.coli cells (provided by Dr. Feng Shao) were transformed with this

19 plasmid to obtain a recombinant bacmid. Sf21 insect cells were transfected with this

20 recombinant bacmid using Cellfectin II (Gibco,USA) according to the manufacturer's

21 instructions. rPFR with C-terminal 6xHis tag was purified from the supernatants of

22 baculovirus infected Sf21 cells. The purity of the purified rPFR was assessed by

23 electrophoresis on a 10% polyacrylamide gel and a 69 kDa band was visualized by

24 silver staining and an immunoblot.

25 Coupling the anti-SLO or PFR to the AFM-tip

7

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 The AFM tips (NPG-10) were functionalized with anti-PFR or anti-SLO antibody

2 as described before (21). In brief, NPG-10 tips were coated with the crosslinker of

3 NHS-PEG-maleimide for 1h at room temperature. Meanwhile, anti-perforin mAbs

4 Pf-80/164 (1 μg/μl, Mabtech AB, Sweden) or anti-SLO (1 μg/μl) antibody was

5 incubated with SATA solution for 30 min at room temperature, followed by 2 hours

6 incubation with hydroxylamine solution at room temperature. Then, the

7 PEG-maleimide-covered NPG-10 tips were mixed with the thiol-functionalized

8 antibody, leading to the binding of NPG-10 tips with anti-PFR or anti-SLO antibody

9 via the reaction of maleimide and thiol.

10 Ultra-High-Resolution SIM

11 Cells were treated with 100 U rPFR for 20 min and fixed in 4 % paraformaldehyde

12 and permeabilized with 0.5 % Triton X-100 at 4°C for 10 min. Then, these cells were

13 blocked with 5% BSA for 20 min at room temperature. After incubation with

14 anti-Flag (Cat: F1804; RRID:AB_262044; Sigma, USA); anti-MYH9 (Cat: 3403;

15 RRID:AB_2147297; CST, USA); anti-F-actin (1:1000, Cat: ab205;

16 RRID:AB_302794; Abcam, UK) at 4°C overnight. Then the cells were incubated with

17 Alexa fluor 488-conjugated secondary antibody (Invitrogen, 1:1000), Alexa fluor

18 594-conjugated secondary antibody (Invitrogen,1:1000) and Alexa fluor

19 647-conjugated secondary antibody (Invitrogen,1:1000). At last, the slides were

20 counterstained with DAPI and mounted for Structured illumination microscopy (SIM,

21 GE Deltavision SR) analysis.

22 Bio-layer interferometry (BLI)

23 The recombinant perforin (rPFR) and mutated perforin (rPFRm) were purified

24 according to the method described in the former context, while the SLO was

25 purchased from Sigma and the perforin (PFR-mix) from T cell cytotoxic vesicles were

8

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 extracted according to the method described above. Cyto- from

2 different cell lines was extracted with a cyto-membrane protein extraction kit

3 (Biovision, cat: K268-50), and the protein was quantified and adjusted to the same

4 concentration. Protein interactions were measured and analyzed by an Octet Red

5 instrument (PALL, NY). The Octet Amine Reactive 2nd Generation (AR2G)

6 Biosensors were dipped into solution containing rPFR, rPFRm, PFR-mix, negative

7 PFR or SLO (all solution was kept in 1ug/ml) and subsequently loaded with various

8 cell membrane protein solution. The protein association and disassociation process

9 was monitored and analyzed by Octet software and processed and graphed with

10 Graphpad software.

11 Animal experiments and treatment protocol

12 C57BL/6 mice or NSG mice were injected with 1 × 105 OVA-B16 cells,

13 SGCTRL-OVA-B16 cells, Myh9-SGs-OVA-B16 cells, SGCTRL-B16 cells,

14 Myh9-SGs-B16 cells, MC-38 cells, SGCTRL-MC-38 cells or Myh9-SGs-MC-38 cells

15 (s.c.). When the tumor grew to 5 × 5 mm, mice were randomized into different groups

16 (sample size was n = 5-10) based on similar tumor size and body weight. And then,

17 these mice were transferred with or without tumor specific CD8+ T cells (4 ×

18 106/mouse, every three days), Ble (182 μg/kg, intratumorly, once every two days), Jas

19 (175 μg/kg, intratumorly, once every two days), anti-PD-1 neutralizing antibody (250

20 μg/mouse, once every two days) or anti-PD-1 + Jas for the indicated time. The mice in

21 the control groups received an equal volume of saline. The tumor growth was

22 measured and the tumor formation and long term survival was recorded.

23 Quantification and Statistical Analysis

24 All experiments were performed at least three times. Results are analyzed by

25 Student's t-test or 1-way ANOVA followed by Boferroni’s test. The long-term survival

9

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 was analyzed by the Log-rank Test. The p value < 0.05 was considered statistically

2 significant. The analysis was conducted using the Graphpad 6.0 software. Sample

3 exclusion was never carried out.

10

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Results

2 Soft TRCs are resistant to cytolytic T cells

3 To explore whether the stiffness of tumor cells influences T cell killing, we set up a

4 B16 melanoma tumor cell line expressing the model tumor antigen ovalbumin, which

5 is recognized by OVA-specific CTLs. Effector (activated OT-I CTLs) and target

6 (OVA-B16) cells were then co-cultured at different ratios or for different times for the

7 cytotoxicity assay. Despite the killing efficiency, some tumor cells evaded death, even

8 in the presence of exceedingly high numbers of CTLs with sufficient killing time (Fig.

9 1A; Supplementary Fig. S1A and S1B). This evasion was not due to the loss of OVA

10 expression and presentation, because both the surviving and the bulk tumor cells

11 displayed equal amounts of MHC class I-OVA peptide complexes on their surface

12 (Supplementary Fig. S1C). Moreover, the surviving tumor cells did not have an effect

13 on CTL degranulation, as indicated by the analysis of CD107a, a marker for CTL

14 degranulation (Supplementary Fig. S1D). These results led us to hypothesize that the

15 surviving tumor cells probably mobilize their biomechanics to avoid T cell killing.

16 Based on the measurements of a few hundred bulk cells, cellular stiffness was found

17 to be widely distributed (Supplementary Fig. S1E). A small population of tumor cells

18 was found with a stiffness of less than 0.3 KPa, while most cells had more than 0.5

19 KPa stiffness. Thus, we subjectively set up a cut-off for cellular stiffness with <0.3

20 KPa being defined as “soft”, ˃ 0.5 KPa as “stiff”, and between 0.3-0.5 KPa as

21 “sub-stiff”. Intriguingly, following T cell killing the average stiffness of tumor cells

22 that had survived was strikingly decreased from 0.64 to 0.17 KPa (Fig. 1B),

23 suggesting that a subpopulation of soft tumor cells evade T cell killing. In addition to

24 OT-I T cells, gp100-specific CD8+ T cells were not able to kill all the cognate B16

25 cells (Supplementary Fig. S1F), even though gp100 was highly expressed

11

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 (Supplementary Fig. S1G). In this context, surviving B16 cells were also softer than

2 bulk B16 cells (Supplementary Fig. S1H). Previous findings have demonstrated that

3 TRCs are very soft and can grow colonies in soft 3D fibrin gels (9, 10). Notably, 93%

4 of the above surviving tumor cells from the 100:1 or 200:1 ratio group were able to

5 form colonies in the soft fibrin gels (Supplementary Fig. S1I), which had the same

6 stiffness as the gel-derived TRCs and were much lower than that of the control cells

7 (Fig. 1C), suggesting that the surviving tumor cells are TRCs. Indeed, we observed

8 the selective killing of bulk OVA-B16 or B16 cells rather than corresponding TRCs

9 by OT-I or pmel-1 CTLs (Fig. 1D; Supplementary Fig. S1J-S1L). In addition, TRCs

10 also had no effect on CTL degranulation (Supplementary Fig. S1M). We also

11 excluded the possibility that decreased cellular stiffness following CTL-mediated

12 killing was due to the CTL co-culture, an effect of cell density or cell cycle stage

13 (Supplementary Fig. S1N-S1P). Together, these results suggest that tumor-specific

14 CD8+ T cells are incapable of killing soft TRCs.

15 Perforin pore formation is impaired in soft tumor-repopulating cells

16 Next, we wondered whether TRCs evade CTL killing at the killing stage,

17 considering the expression of MHC class I by B16 TRCs (9). To test this, we used

18 isolated perforin to treat TRCs and control tumor cells, and monitored the cellular

19 influx of propidium iodide (PI), a fluorescent agent used to reflect the effect of

20 perforin pore formation (21). We found that perforin, purified either from

21 recombinant baculovirus-expressing system (rPFR) or CD8+ T cells (PFR-mix),

22 displayed much higher pore forming potency against differentiated OVA-B16 tumor

23 cells (Fig. 1E). Analogous pore formation could not be induced by perforin mutant

24 without activity (rPFRm) or membranous organelles (isolated from HL-60 cells which

25 don’t express PFR) as a negative PFR (Supplementary Fig. S1Q), indicating that

12

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 perforin does indeed induce pore formation in differentiated tumor cells. Much more

2 PI entered bulk cells than the TRCs when co-cultured with OT-1 T cells in vitro (Fig.

3 1F). The resistance of TRCs to perforin was also confirmed in other tested murine and

4 human tumor cell lines (A375, HepG2, MCF-7 and 4T1) (Supplementary Fig. S1R

5 and S1S). Streplysin O (SLO), a bacterial product, is an analog of perforin. Using

6 recombinant SLO, we also confirmed the above results (Supplementary Fig. S1T). In

7 line with PI influx, we also demonstrated a much higher uptake of apoptosis-inducing

8 Granzyme B (GrB) in bulk tumor cells, which was scarcely present in TRCs

9 following perforin treatment (Supplementary Fig. S1U). Such GrB entry was also

10 observed in OVA-B16 cells but not in OVA-TRCs when co-cultured with OT-1 T

11 cells (Fig. 1G), suggesting that perforin has a weak effect to form a pore in the

12 membrane of TRCs. Moreover, we used atomic force microscopy (AFM) to directly

13 visualize the formed pores in the cellular membrane (21). Fewer pores were

14 visualized in either perforin-treated or CTL-co-cultured OVA-TRCs versus bulk

15 OVA-B16 cells (Fig. 1H, I; Supplementary Fig. S1V). It has been reported that

16 lysosomes release enzymes, such as cathepsins, to degrade perforin (22), thus

17 blocking granzyme entry. However, using a lysosomal exocytosis inhibitor vaculin-1

18 did not improve the entry of exogenous PI into perforin-treated TRCs (Supplementary

19 Fig. S1W). In addition, the inability of CTLs to kill soft TRCs was not attributed to

20 the released perforin that attacked CTLs themselves, as evidenced by CTL-TRC

21 interactions not inducing CTL death (Supplementary Fig. S1X). Together, these data

22 suggest that soft TRCs use certain mechanism(s) to prevent perforin from drilling a

23 pore, thus evading T cell killing.

24 Cellular softness interferes with perforin pore formation in TRCs

13

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Next, we investigated whether the mechanical softness of TRCs prevented perforin

2 pore formation. In line with our previous reports (10, 18), here we further confirmed

3 that cellular softness is a common feature for TRCs, tested in OVA-B16, B16, A375,

4 MCF-7, 4T1 and HepG2 TRCs by AFM (Fig. 1C; Supplementary Fig. S2A). F-actin

5 is an essential element that contributes to cellular stiffness (15). When we used

6 Cytochalasin D (Cyto), an inhibitor of actin polymerization, to treat the stiff bulk

7 tumor cells, the cells became soft (Fig. 2A), accompanied with simultaneous

8 resistance to perforin pore formation (Fig. 2B; Supplementary Fig. S2B and S2C) and

9 decreased entry of PI into the cytoplasm (Fig. 2C and 2D; Supplementary Fig. S2D).

10 A similar result was also obtained with another F-actin inhibitor, Latrunculin A

11 (Lat-A) (Fig. 2B; Supplementary Fig. S2B, S2C, S2E). On the other hand, when we

12 used jasplakinolide (Jas), a natural cyclodepsipeptide that is a potent inducer of actin

13 polymerization, to increase TRC stiffness (Fig. 2E), we observed increased perforin

14 pore formation (Fig. 2F; Supplementary Fig. S2F) and more PI entered the cytoplasm

15 (Fig. 2G and 2H; Supplementary Fig. S2G), suggesting that perforin pore formation is

16 easier in the membrane of stiff cells but is more difficult in a more deformable plasma

17 membrane. In addition, Cyto, Lat-A or Jas treatment had no direct effect on the entry

18 of PI into the cells (Supplementary Fig. S2H). In line with these results, both the

19 activity and expression of Cdc42, a member of the Rho family that is critical for

20 F-actin biogenesis (23, 24), was found to be downregulated in TRCs relative to

21 differentiated tumor cells (Supplementary Fig. S2I), consistent with published results

22 (18). Forced overexpression of Cdc42 resulted in increased stiffness of TRCs

23 (Supplementary Fig. S2J), concurrent with more PI entering the cytoplasm and

24 increased cytolysis of TRCs by CTLs (Supplementary Fig. S2K and S2L). In addition,

25 the surviving Cdc42-OVA-B16 TRCs following CTLs mediated killing were much

14

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 softer than bulk cells (Supplementary Fig. S2M). On the other hand, Cdc42

2 knockdown resulted in the decreased stiffness of bulk tumor cells (Supplementary Fig.

3 S2N and S2O), parallel with less PI entering the cytoplasm (Supplementary Fig. S2P

4 and S2Q), and decreased cytolysis of bulk tumor cells by CTLs (Supplementary Fig.

5 S2R). Together, these data suggest that TRCs use their inherent softness to interfere

6 with perforin pore formation in order to escape CTLs’ killing.

7 MYH9 is identified as a molecule required for perforin pore formation

8 To investigate how the mechanical force was involved in the process of perforin

9 pore formation, we speculated that certain membrane molecule(s) were able to

10 interact with perforin, leading to a “drill” force generation. Bio-layer interferometry

11 (BLI) is a label-free technology for measuring biomolecular interactions. Intriguingly,

12 BLI analysis showed that both OVA-B16 and MCF-7 cellular lysates were able to

13 bind perforin (rPFR, PFR-mix) or SLO, but not rPFRm or negative-PFR, encouraging

14 us to identify the potential interacting protein(s) (Fig. 3A; Supplementary Fig. S3A).

15 Using anti-perforin to pull down the binding in the perforin-treated cells, a

16 panel of candidates including beta chain, brain 1 (SPTBN1), flotillin-1

17 (FLOT1), myosin heavy chain 9 (MYH9) and myosin-Ic (MYO1C) were identified by

18 mass spectrometry (Supplementary Fig. S3B). Among them, MYH9 and MYO1C

19 especially caught our attention. MYH9 is a motor protein, participating in a variety of

20 processes requiring contractile force (25); and MYO1C is a type-I myosin that links

21 cell membranes to the (26). The results from immunoprecipitation and

22 western blot showed that perforin indeed interacted with MYH9 and MYO1C (Fig.

23 3B; Supplementary Fig S3C). Similar results were obtained by using anti-SLO

15

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 antibody (Supplementary Fig. S3D-S3F). Moreover, cell lysates from SGCTRL and

2 MYH9-SGs OVA-B16 or MCF-7 cells were analyzed by BLI, which showed that

3 rPFR and PFR-mix could interact with SGCTRL cell lysates but not MYH9-SGs cell

4 lysates and rPFRm did not interact with cell lysates (Fig. 3C; Supplementary Fig.

5 S3G). Notably, when we used purified perforin and MYH9 protein to repeat the BLI

6 assay, we did not find the direct interaction between these two proteins

7 (Supplementary Fig. S3H), indicating that perforin indirectly interacts with MYH9.

8 In addition, we found that there was barely an interaction between perforin and

9 MYO1C (Supplementary Fig. S3I). In line with this finding, only the Myh9 knockout

10 but not Myo1c knockout resulted in the disruption of perforin pore formation in

11 OVA-B16 cells (Fig. 3D-3F). This result was also confirmed by AFM analysis, which

12 directly showed an absence of pores formed in cells with a MYH9 deficiency (Fig.

13 3G). Moreover, we used fluorescence-labeled rPFR to treat OVA-B16 or

14 Myh9-deficient OVA-B16 cells and imaged the same cells with fluorescence

15 microscopy and AFM, respectively. After merging the two images, we could observe

16 the pores surrounded by rPFR, while Myh9 knockout abrogated this phenomenon

17 (Supplementary Fig. S3J). A very close location of PFR with MYH9 was observed in

18 the rPFR-treated tumor cells under ultra-high-resolution structured illumination

19 microscope (Supplementary Fig. S3K). In addition, we also knocked out MYH9 in

20 MCF-7 cells, which showed a similar result of impaired pore formation induced by

21 perforin, as shown by less PI entering, decreased pore formation, and smaller pore

22 size in these cells (Supplementary Fig. S3L-S3N). Taken together, these results

16

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 identified that MYH9 acts as a perforin-interacting molecule that is required for

2 perforin pore formation.

3 A low perforin-generated MYH9-mediated force in TRCs is explained by

4 softness

5 Next, we investigated how MYH9 regulates perforin pore formation. Generation of

6 a mechanical force requires myosin II, especially the heavy chain, which crosslinks

7 and contracts actin (27, 28). We thus inquired whether the force generation by

8 perforin-MYH9 interaction was required for perforin pore formation. To test this

9 possibility, we used AFM to detect the generated force between two molecules on the

10 cellular membrane (21). We linked pf-80, an anti-perforin mAb that recognizes a

11 perforin epitope but does not interfere with the binding of perforin to the plasma

12 membrane (21, 29), to the tip of AFM and used it to detect the adhesive force. Indeed,

13 the adhesive force between pf-80 and perforin was found to be very high (about 202

14 pN from OVA-B16; 191 pN from MCF-7) in bulk tumor cells, but very low (about 83

15 pN from OVA-B16; 100 pN from MCF-7) in TRCs (Fig. 4A). In addition, we also

16 used a covalently functionalized AFM tip with anti-SLO antibodies and performed the

17 force spectroscopy on cells treated with SLO. Consistently, a much lower force was

18 detected in TRCs relative to their bulk counterparts (Supplementary Fig. S4A). Then,

19 we clarified whether such detected force was mediated through MYH9 or other

20 perforin-interacting proteins. We conducted the AFM assay with MYH9-/- cells. In the

21 setting of MYH9 deficiency, around half of perforin was still located on the cellular

22 membrane (Supplementary Fig. S4B). However, MYH9 knockout resulted in a

17

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 remarkable decrease of perforin-caused force in the bulk tumor cells, and such force

2 was also further reduced in MYH9-/- TRCs (Fig. 4B; Supplementary Fig. S4C),

3 suggesting that MYH9 mediates the force generation through perforin. A similar

4 result was obtained in SLO-treated MYH9-/- cells (Supplementary Fig. S4D). MYH9

5 deficiency also disrupted the pore formation (Fig. 3D-3G; Supplementary Fig.

6 S3L-S3N). Blebbistatin (Ble) has been reported to be a reversible inhibitor of MYH9

7 (30). Consistently, the use of Ble also abrogated the perforin-generated force in TRCs

8 and the bulk counterparts (Fig. 4C). In addition, we also tested MYO1C-/- cells. We

9 found that MYO1C deficiency had a minor effect on such force generation

10 (Supplementary Fig. S4E), highlighting that MYH9 plays a pivotal role in mediating

11 the force generation by perforin.

12 Next, we wondered how MYH9 mediated a low perforin-generated force in TRCs

13 but a high force in bulk tumor cells. Despite the requirement of myosin II for

14 mechanical force generation, the force magnitude is actually decided by the amount of

15 F-actin directly interacting with myosin II. We thus hypothesized that the

16 transmission of the perforin-generated force via MYH9 to fewer F-actins (i.e., greater

17 softness) led to a lower force; but transmission of the force to more F-actins (thus

18 higher cell stiffness) caused a stronger force. Indeed, fewer F-actins were observed in

19 TRCs, compared to their bulk counterparts (Fig. 4D; Supplementary Fig. S4F). When

20 we used Jas to treat TRCs to increase F-actin, in order to enhance the cell stiffness

21 (Supplementary Fig. S4G), the perforin-generated force was increased

22 (Supplementary Fig. S4H). In contrast, when we used Cyto to treat bulk tumor cells to

23 reduce the cell stiffness, the corresponding force was reduced (Supplementary Fig.

18

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 S4I). These results were consistent with the finding that the perforin pore formation

2 was enhanced in Jas-treated TRCs but decreased in Cyto-treated bulk tumor cells (see

3 Fig. 2B-2H). In addition to F-actin, MYH9 was also able to regulate the cellular

4 stiffness. Knockout of MYH9 resulted in the decrease of the cellular stiffness from

5 0.63 kPa (OVA-B16) or 1.04 kPa (MCF-7) to around 0.25 kPa or 0.42 kPa,

6 respectively (Fig. 4E). In addition, fewer F-actins were also observed in

7 MYH9-knockout tumor cells (Supplementary Fig. S4J). Such an effect might be due

8 to the elimination of the capability of MYH9 to crosslink and to stiffen F-actin (31).

9 We found that the expression of MYH9 was decreased in TRCs, compared to the

10 corresponding bulk cells (OVA-B16 and MCF-7) (Fig. 4F). Together, these results

11 suggest that TRCs use their softness to reduce the MYH9-mediated force generation

12 linked to perforin and required for pore formation.

13 Cell softness hinders CTL-mediated perforin pore formation in vivo

14 The preceding experiments indicated that highly tumorigenic cells mobilized their

15 intrinsic softness to hinder perforin pore formation in order to evade CTL killing.

16 Here, we further validated this process in vivo. Three days after the adoptive transfer

17 of OVA-specific CD8+ T cells into C57BL/6 mice bearing a 5×5 mm OVA-B16

18 melanoma, we isolated single tumor cells for flow cytometric analysis (Fig. 5A).

19 Previously, we found that CD133 and ALDH could be a marker for TRCs in B16

20 melanoma and 4T1 breast cancer (9), respectively. Here, we found that CD8+ T cell

21 adoptive transfer resulted in the effective entry of GrB into CD133- OVA-B16 cells

22 (Fig. 5B), as well as perforin pore formation (Fig. 5C), which, however, was reversed

23 by blebbistatin treatment (Fig. 5B and 5C). By contrast, adoptive transfer of CD8+ T

24 cells resulted in poor entry of GrB into CD133+ tumor cells and reduced pore

25 formation, whereas Jas treatment promoted the entry of GrB into CD133+ tumor cells

19

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 and increased pore formation (Fig. 5D and 5E). In line with this result, CD133+ tumor

2 cells were much softer than CD133- cells, whose stiffness was enhanced by Jas, while

3 Ble treatment decreased the stiffness of CD133- cells (Supplementary Fig. S4K).

4 Since MYH9 knockout made the cells softer, we inoculated a small number of

5 SGCTRL- or Myh9-SGs-B16 cells into C57BL/6 mice or NOD-SCID mice. As

6 expected, the Myh9-deficient melanoma grew faster than the Myh9-expressing

7 melanoma in wild-type mice, while the absence of T cells blocked the Myh9

8 deficiency-triggered tumor promoting effect in the mice (Fig. 5F and 5G). In addition,

9 we found that MYH9 deficient cells had a similar proliferation rate to

10 MYH9-expressing cells in vitro (Supplementary Fig. S4L). These results suggested

11 that MYH9 regulates intrinsic CTL killing in vivo. Consistently, when we inoculated

12 Myh9-/- OVA-B16 cells in mice, followed by OT-1 T cell adoptive transfer, we found

13 that adoptive transfer-produced inhibory effect on tumor growth was some impaired

14 (Fig. 5H), simultaneous with decreased IFN-γ levels in tumor mass (Supplementary

15 Fig. S4M). We also adoptively transferred gp100-specific CD8+ T cells to B16

16 melanoma-bearing mice. Similar results were obtained (Supplementary Fig.

17 S4M-S4R). Neo-antigen (Adpgk) is capable of inducing endogenous tumor-specific

18 CD8+ T cells to destroy MC-38 tumor cells in vivo (32). Here, we also used the

19 Myh9-/- MC-38 tumor model to further validate or refute the above result. We

20 immunized C57BL/6 mice with the neo-antigen and subsequently challenged the mice

21 with SGCTRL or Myh9-SGs MC-38 tumor cells (Fig. 5I). Consistently, softening

22 MC-38 tumor cells using a Myh9 knockout or Ble treatment also attenuated the

23 endogenous antitumor T cell immunity, as evidenced by accelerated tumor growth

24 and increased tumor incidence (Fig. 5J and 5K). Together, these data suggest that

20

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 TRCs use mechanical softness to hinder CTL-mediated perforin pore formation in

2 vivo.

3 Stiffening tumor cells improves T cell immunotherapy in mouse models

4 Next, we explored whether stiffening tumor cells could improve T cell

5 immunotherapy. PD-1 blockade therapy has yielded antitumor T cell immune

6 responses in patients with a broad spectrum of cancers (33, 34). In this regard, we

7 adoptively transferred OT-1 T cells to mice bearing OVA-B16 melanoma (5×5 mm),

8 and treated the mice with PD-1 antibody for two weeks. Indeed, the treatment

9 enhanced the efficiency of adoptively transferred T cells, as evidenced by retarded

10 tumor growth, less Ki67+ cells and increased TUNEL+ cells (Fig. 6A; Supplementary

11 Fig. S5A). However, dissection of remnant tumor cells revealed that despite a high

12 degree of tumor-specific T cell activation by PD-1 blockade, as evidenced by the

13 upregulation of IFN-γ expression (Fig. 6B), most CD133+ tumorigenic cells survived

14 and few CD133+ tumor cells were killed (Fig. 6C). Since CD133+ cells were softer

15 than CD133- tumor cells (Fig. 6D), these results indicated that tumor-specific T cells

16 were incapable of targeting soft tumor cells in vivo. Then, we treated the mice with

17 Jas to increase the cellular stiffness. We found that the transferred T cells following

18 Jas treatment were capable of killing CD133+ tumorigenic cells (Fig. 6C), and the

19 triple treatment (T cell/anti-PD-1/Jas) resulted in the best treatment outcome (Fig. 6A;

20 Supplementary Fig. S5A). By contrast, softening tumor cells by the administration of

21 Ble attenuated the treatment effect of OT-I CTLs on OVA-B16 tumor (Supplementary

22 Fig. S5B). Notably, in T cell-deficient NSG mice single Jas or Ble treatment had a

23 minor effect on tumor growth (Supplementary Fig. S5C), and Jas or Ble appeared not

24 to affect T cell function (Supplementary Fig. S5D). In a MC-38 tumor model, the Jas

25 treatment also enhanced the neo-antigen-induced endogenous antitumor T cell

21

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 immunity, as evidenced by a remarkable inhibition of tumor incidence

2 (Supplementary Table). Together, these data suggest that stiffening TRCs is capable

3 of improving T cell immunotherapy.

4 Softness attenuates perforin pore formation in cancer patients

5 Finally, we sought to validate our findings with clinical patient samples. CD133

6 and LGR5 can be used as a stemness marker for melanoma and colon cancer,

7 respectively (35-37). We isolated CD133+ and CD133- tumor cells (melanoma) or

8 LGR5+ and LGR5- tumor cells (colon cancer) from patients (n = 5, each). AFM

9 analysis showed that stem cell-like tumor cells (CD133+ or LGR5+) were much softer

10 than their counterparts (CD133- or LGR5-) (Fig. 7A). In addition, when seeding the

11 isolated bulk tumor cells into soft 3D fibrin gels, we found that the formed TRCs were

12 also softer than the bulk tumor cells (Fig. 7 B), more resisted to PI entry (Fig. 7C;

13 Supplementary Fig. S5E), and had smaller and fewer pores (Fig. 7D). To confirm that

14 tumor cells use their softness to evade T cell killing, we co-cultured primary human

15 melanoma or colon cancer cells or their corresponding TRCs with allogenic CD8+ T

16 cells (aCTL) with anti-CD28 antibody. Consistently, CTLs effectively killed bulk

17 tumor cells in a perforin-dependent fashion (14, 38), but had little killing effect on

18 soft TRCs (Supplementary Fig. S5F). Furthermore, a lower perforin-induced drilling

19 force was generated in TRCs relative to bulk cells (70 versus 143 pN in melanoma

20 and 79 versus 249 pN in colon cancer) (Fig. 7E), and neither melanoma nor colon

21 cancer TRCs had pores in their cellular membranes (Supplementary Fig. S5G and

22 S5H). However, strengthening perforin-induced force in Jas-treated TRCs (Fig. 7E)

23 led to enhanced pore formation, as evidenced by enhanced PI staining (Supplementary

24 Fig. S5I), larger pores, and more pores formed (Supplementary Fig. S5G and S5H).

25 On the other hand, treatment of primary human tumor cells with Cyto attenuated

22

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 perforin-mediated force (Supplementary Fig. S5J) and impaired perforin-induced pore

2 formation (Supplementary Fig. S5G, S5H, S5K), suggesting that softness regulates

3 pore formation in human tumor cells. In addition, Jas treatment enhanced

4 CTL-mediated killing of primary melanoma TRCs, while Cyto treatment impeded the

5 killing of the bulk tumor cells by aCTLs (Supplementary Fig. S5L). Given the role of

6 MYH9 in perforin pore formation in murine tumor cells, we also validated this in

7 primary human tumor cells. MYH9 knockout indeed reduced the force induced by

8 perforin (Supplementary Fig. S5M) and resulted in the inhibition of perforin- or

9 SLO-induced pore formation in primary human melanoma cancer cells (Fig. 7F;

10 Supplementary Fig. S5N). In line with this, F-actin levels were much higher in these

11 melanoma cancer cells than their corresponding TRCs (Supplementary Fig. S5O).

12 These data suggest that cellular softness also prevents perforin from drilling holes in

13 human tumor cells, thus mediating their immune evasion.

23

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Discussion

2 Interaction between T cells and target cells is not only a chemical process but also a

3 mechanical event. It is increasingly clear that CTLs mobilize a mechanical force to

4 facilitate the killing of tumor cells (39), however how tumor cells in turn use force to

5 counteract T cell killing remains unclear. In the present study, we provide evidence

6 that TRCs evade CTL killing through a mechanical softness mechanism underlying

7 impairment of perforin pore formation; downregulating this softness, however,

8 restores CTL-mediated cytolysis of TRCs.

9 Demonstration of the regulation of perforin pore formation by cellular softness

10 provides new insights into how highly tumorigenic cells evade the immune system.

11 Cellular stiffness is known to be decided by tensile stress, which is generated from

12 actin structures. Activation of the myosin-based contraction increases

13 the tensile force in actin filaments, which, in turn, stiffens the F-actin lattice. Thus, the

14 cell stiffness is the collective result of actin polymerization and myosin II mediated

15 contractile activation. Since plasma membrane stiffness is known to be extremely low,

16 on the order of ~0.1-1 Pa (40), and the stiffness of adherent differentiated cells is on

17 the order of 1-10 kPa (16, 41), the contribution from the plasma membrane to cell

18 stiffness is negligible in this study. Perforin actively interacts with membrane

19 molecules and the adhesive force was quantitated by AFM. Thus, we propose to use

20 perforin’s “drilling” force to vividly describe the mechanical process of perforin

21 interacting with membrane molecules including lipids, MYH9, and other

22 membrane-associated molecules, leading to the generation of actin filament tensile

23 stress and the opposite contractile force. Previous studies have reported that Jas has an

24 antitumor effect by targeting the actin cytoskeleton (42, 43). However, given the

25 fundamental function of actin cytoskeleton, Jas treatment is likely to cause side

24

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 effects. Our present study might provide an opportunity for Jas in combination with T

2 cell immunotherapy to generate a better antitumor effect under a condition of the

3 possible low dosage and low side effect.

4 In this study, we found that MYH9 plays a crucial role for mediating

5 perforin-formed pores in tumor cells. Nonmuscle myosin II (NM II) is an important

6 motor protein that has actin cross-linking and contractile properties. MYH9 is a

7 widely expressed encoding nonmuscle myosin heavy chain. Since the knockout

8 of MYH9 makes cell softer and TRCs express less MYH9, we speculate that MYH9

9 affects perforin-induced pore formation through an indirect interaction with perforin.

10 On the other hand, studies have used the artificial lipid monolayer (liposome) to

11 detect the pore formation mediated by perforin (44, 45), suggesting that perforin has

12 the ability to interact with membrane lipid molecules. Therefore, when perforin

13 contacts the plasma membrane of a cell, perforin might first interact with membrane

14 lipid molecule(s), resulting in its conformation change. Following this change,

15 perforin can interact with certain membrane-associated protein(s), which thus binds to

16 and confer MYH9 an increased ATPase enzymatic activity due to its change in

17 conformation. Thus, contractile force, generated from ATP-released energy which

18 propels actin filaments, is enhanced. As a result, the opposite force is increased and

19 transferred to perforin via MYH9. Notwithstanding the elucidation of this mechanical

20 mechanism, previous studies have also demonstrated that perforin pore formation is a

21 chemical process, which involves Ca+ signaling among others (46). Based on the

22 above analyses, we propose that for stiff tumor cells, a perforin-MYH9 interaction

23 induces a strong opposite force, which in turn, affects and changes the conformation

24 of perforin in order to trigger the subsequent chemical process (Figure 7G). In

25 contrast, for soft tumor cells, the interaction induces a weak opposite force, which is

25

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 not able to change the conformation, and therefore cannot trigger the subsequent

2 chemical process. In addition, although MYH9 is traditionally thought to be regulated

3 by the (47), this present study indicates that MYH9 might be

4 mechanically regulated by perforin, thus bypassing chemical signal-triggered myosin

5 light chain regulation. This myosin light chain-independent regulation can explain the

6 inconsistent results from a very recent study by Kim et al, which showed that lower

7 levels of myosin regulatory light chain 9 improved perforin-based killing (48). In

8 addition, in this study, by performing co-immunoprecipitation and mass spectroscopy,

9 we identified several membrane protein molecules including MYH9, FLNA, MYO1C,

10 etc. Whether these molecules directly or indirectly bind to MYH9 is unclear.

11 Currently, the identification of the membrane molecules that directly interact with

12 perforin is under investigation.

13 Another important aspect in this study lies in the use of scanning AFM to

14 investigate perforin pore formation. AFM is an established technique of nanometer-

15 scale scanning probe microscopy, which can non-invasively image cells, where a

16 probe is composed of a cantilever with a sharp tip mounted at its end and the probe

17 raster scans the sample to obtain the topography image of cells. AFM with the

18 PeakForce Tapping mode can precisely capture the height profiles, as well as

19 adhesive and stiffness maps of cell membrane pores simultaneously, which is enough

20 for the pore topography recognition. We have reported the method based on

21 PeakForce Tapping mode AFM to visualize pores in the plasma membrane of true

22 immune and tumor cells (21). In this study, we further used this method to measure

23 the perforin-mediated pore formation in tumor cells. Due to the pore-forming property

24 of perforin, very clear signals appeared from the AFM imaging, but in contrast, the

25 almost complete absence of signal occurred in non-perforin treated conditions. In

26

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 addition to AFM, we also used PI staining to evaluate the pore formation, and found

2 that the PI staining was decreased in soft cells. However, the result was inconsistent

3 with a recent study by Basu et al. showing that blebbistatin treatment, which made

4 cells softer, increased pore formation. This might be due to a very high concentration

5 of blebbistatin (100 μM) used by Basu et al, which was very toxic to cells (39). This

6 AFM-based pore visualization can reflect the stiff status of cells. Perforin is easy to

7 form more pores in stiff cells, but is hard to form pores in soft cells. However, use of

8 Jas to stiffen soft cells confer the perforin to form pores with even larger size in the

9 soft cells.

10 TRCs exhibit an interferon-triggered tryptophan-IDO-Kyn pathway (49, 50). Of

11 note, this pathway acts as an important mechanism underlying the induction of PD-1

12 expression in CD8+ T cells (9). In this study, we further identify another extraordinary

13 immunosuppressive feature of TRCs, based on their mechanical softness. Unlike their

14 differentiated counterparts, TRCs are intrinsically soft; however, such softness is a

15 powerful defence against cytolytic T cell killing by hindering perforin pore formation,

16 since certain levels of cellular stiffness appear to be required for perforin to drill a

17 hole in the cell membrane. The elucidation of this mechanism provides deep insights

18 into why clinical tumor relapses exist after efficacious immunotherapy. Based on our

19 finding, it is possible to develop promising approaches that attack those intractable

20 stem cell-like tumor cells. Currently, researchers from different laboratories try to use

21 cancer stem cell (CSC)-specific antigens to induce anti-CSC T cell immunity in order

22 to achieve better tumor treatment outcomes. While such strategies might indeed lead

23 to better targeting and an improved treatment outcome compared to conventional

24 tumor antigens, those CSC-targeting CD8+ T cells might still meet the obstacle of

25 CSC softness, leading to the inability to kill true CSCs which are soft. However, such

27

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 CSC antigen-based T cell therapies, if combined with stiffness enhancer(s), might

2 achieve a better cytolysis of CSCs. Based on this concept, it is worthy of

3 investigation.

4 In summary, This work uncovers an unknown tumor immune evasion mechanism,

5 which may provide a potential explanation on immune escape by malignant cells in

6 CAR T cell- or PD-1 blockade-treated patients.”the data in this study clearly show

7 that perforin-generated force via MYH9 to fewer F-actins in soft TRCs leads to a

8 lower force, thus impairing perforin drilling a pore. This work uncovers an unknown

9 tumor immune evasion mechanism, which may provide a potential explanation on

10 immune escape by malignant cells in CAR T cell- or PD-1 blockade-treated patients.

11

28

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Acknowledgments

2 This work was supported by National Natural Science Foundation of China

3 (81788101, 81773062, 91942314), Chinese Academy of Medical Sciences (CAMS)

4 Initiative for Innovative Medicine (CAMS-I2M) 2017-I2M-1-001 and

5 2016-I2M-1-007.

29

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Reference

2 1. Borghaei H, Paz-Ares L, Horn L, Spigel DR, Steins M, Ready NE, et al.

3 Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell

4 Lung Cancer. N Engl J Med 2015;373:1627-39.

5 2. Neelapu SS, Locke FL, Bartlett NL, Lekakis LJ, Miklos DB, Jacobson CA, et

6 al. Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B-Cell

7 Lymphoma. N Engl J Med 2017;377:2531-2544.

8 3. Ribas A, Hamid O, Daud A, Hodi FS, Wolchok JD, Kefford R, et al.

9 Association of Pembrolizumab With Tumor Response and Survival Among

10 Patients With Advanced Melanoma. Jama 2016;315:1600-9.

11 4. Zaretsky JM, Garcia-Diaz A, Shin DS, Escuin-Ordinas H, Hugo W,

12 Hu-Lieskovan S, et al. Mutations Associated with Acquired Resistance to

13 PD-1 Blockade in Melanoma. N Engl J Med 2016;375:819-29.

14 5. Majzner RG and Mackall CL. Tumor Antigen Escape from CAR T-cell

15 Therapy. Cancer Discov 2018;8:1219-1226.

16 6. Zou W. Immunosuppressive networks in the tumour environment and their

17 therapeutic relevance. Nat Rev Cancer 2005;5:263-74.

18 7. Nagarsheth N, Wicha MS and Zou W. Chemokines in the cancer

19 microenvironment and their relevance in cancer immunotherapy. Nat Rev

20 Immunol 2017;17:559-572.

21 8. Binnewies M, Roberts EW, Kersten K, Chan V, Fearon DF, Merad M, et al.

22 Understanding the tumor immune microenvironment (TIME) for effective

23 therapy. Nat Med 2018;24:541-550.

24 9. Liu Y, Liang X, Dong W, Fang Y, Lv J, Zhang T, et al. Tumor-Repopulating

30

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Cells Induce PD-1 Expression in CD8(+) T Cells by Transferring Kynurenine

2 and AhR Activation. Cancer Cell 2018;33:480-494 e7.

3 10. Liu J, Tan Y, Zhang H, Zhang Y, Xu P, Chen J, et al. Soft fibrin gels promote

4 selection and growth of tumorigenic cells. Nat Mater 2012;11:734-41.

5 11. Qin Z and Blankenstein T. CD4+ T cell--mediated tumor rejection involves

6 inhibition of angiogenesis that is dependent on IFN gamma receptor

7 expression by nonhematopoietic cells. Immunity 2000;12:677-86.

8 12. Kammertoens T, Friese C, Arina A, Idel C, Briesemeister D, Rothe M, et al.

9 Tumour ischaemia by interferon-gamma resembles physiological blood vessel

10 regression. 2017;545:98-102.

11 13. Voskoboinik I, Whisstock JC and Trapani JA. Perforin and granzymes:

12 function, dysfunction and human pathology. Nat Rev Immunol

13 2015;15:388-400.

14 14. Lopez JA, Susanto O, Jenkins MR, Lukoyanova N, Sutton VR, Law RH, et al.

15 Perforin forms transient pores on the target cell plasma membrane to facilitate

16 rapid access of granzymes during killer cell attack. Blood 2013;121:2659-68.

17 15. Wang N, Butler JP and Ingber DE. Mechanotransduction across the cell

18 surface and through the cytoskeleton. Science 1993;260:1124-7.

19 16. Discher DE, Mooney DJ and Zandstra PW. Growth factors, matrices, and

20 forces combine and control stem cells. Science 2009;324:1673-7.

21 17. Wu P-H, Aroush DR-B, Asnacios A, Chen W-C, Dokukin ME, Doss BL, et al.

22 A comparison of methods to assess cell mechanical properties. Nat Methods

23 2018;15:491-498.

24 18. Tan Y, Tajik A, Chen J, Jia Q, Chowdhury F, Wang L, et al. Matrix softness

25 regulates plasticity of tumour-repopulating cells via H3K9 demethylation and

31

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Sox2 expression. Nat Commun 2014;5:4619.

2 19. Ma J, Zhang Y, Tang K, Zhang H, Yin X, Li Y, et al. Reversing drug resistance

3 of soft tumor-repopulating cells by tumor cell-derived chemotherapeutic

4 microparticles. Cell Res 2016;26:713-27.

5 20. Hirano K, Hosoi A, Matsushita H, Iino T, Ueha S, Matsushima K, et al. The

6 nitric oxide radical scavenger carboxy-PTIO reduces the immunosuppressive

7 activity of myeloid-derived suppressor cells and potentiates the antitumor

8 activity of adoptive cytotoxic T lymphocyte immunotherapy.

9 Oncoimmunology 2015;4:e1019195.

10 21. Liu Y, Zhang T, Zhou Y, Li J, Liang X, Zhou N, et al. Visualization of

11 perforin/gasdermin/complement-formed pores in real cell membranes using

12 atomic force microscopy. Cell Mol Immunol 2019;16:611-620.

13 22. Khazen R, Muller S, Gaudenzio N, Espinosa E, Puissegur MP and Valitutti S.

14 Melanoma cell lysosome secretory burst neutralizes the CTL-mediated

15 cytotoxicity at the lytic synapse. Nat Commun 2016;7:10823.

16 23. Reymond N, Im JH, Garg R, Vega FM, Borda d'Agua B, Riou P, et al. Cdc42

17 promotes transendothelial migration of cancer cells through beta1 integrin. J

18 Cell Biol 2012;199:653-68.

19 24. Rohatgi R, Ma L, Miki H, Lopez M, Kirchhausen T, Takenawa T, et al. The

20 interaction between N-WASP and the Arp2/3 complex links Cdc42-dependent

21 signals to actin assembly. Cell 1999;97:221-31.

22 25. Pecci A, Ma X, Savoia A and Adelstein RS. MYH9: Structure, functions and

23 role of non-muscle myosin IIA in human disease. Gene 2018;664:152-167.

24 26. Mermall V, Post PL and Mooseker MS. Unconventional in cell

25 movement, membrane traffic, and signal transduction. Science

32

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 1998;279:527-33.

2 27. Sun Z, Guo SS and Fassler R. Integrin-mediated mechanotransduction. J Cell

3 Biol 2016;215:445-456.

4 28. Cai Y and Sheetz MP. Force propagation across cells: mechanical coherence of

5 dynamic . Curr Opin Cell Biol 2009;21:47-50.

6 29. Praper T, Besenicar MP, Istinic H, Podlesek Z, Metkar SS, Froelich CJ, et al.

7 Human perforin permeabilizing activity, but not binding to lipid membranes, is

8 affected by pH. Mol Immunol 2010;47:2492-504.

9 30. Straight AF, Cheung A, Limouze J, Chen I, Westwood NJ, Sellers JR, et al.

10 Dissecting temporal and spatial control of cytokinesis with a myosin II

11 Inhibitor. Science 2003;299:1743-7.

12 31. Wang N, Naruse K, Stamenovic D, Fredberg JJ, Mijailovich SM,

13 Tolic-Norrelykke IM, et al. Mechanical behavior in living cells consistent with

14 the tensegrity model. Proc Natl Acad Sci U S A 2001;98:7765-70.

15 32. Yadav M, Jhunjhunwala S, Phung QT, Lupardus P, Tanguay J, Bumbaca S, et

16 al. Predicting immunogenic tumour mutations by combining mass

17 spectrometry and exome sequencing. Nature 2014;515:572-6.

18 33. Yost KE, Satpathy AT, Wells DK, Qi Y, Wang C, Kageyama R, et al. Clonal

19 replacement of tumor-specific T cells following PD-1 blockade. Nat Med

20 2019;25:1251-1259.

21 34. Ribas A and Wolchok JD. Cancer immunotherapy using checkpoint blockade.

22 Science 2018;359:1350-1355.

23 35. Li Y, Luo S, Ma R, Liu J, Xu P, Zhang H, et al. Upregulation of cytosolic

24 phosphoenolpyruvate carboxykinase is a critical metabolic event in melanoma

25 cells that repopulate tumors. Cancer Res 2015;75:1191-6.

33

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 36. Shimokawa M, Ohta Y, Nishikori S, Matano M, Takano A, Fujii M, et al.

2 Visualization and targeting of LGR5(+) human colon cancer stem cells. Nature

3 2017;545:187-192.

4 37. de Sousa e Melo F, Kurtova AV, Harnoss JM, Kljavin N, Hoeck JD, Hung J, et

5 al. A distinct role for Lgr5(+) stem cells in primary and metastatic colon

6 cancer. Nature 2017;543:676-680.

7 38. Trapani JA and Smyth MJ. Functional significance of the perforin/granzyme

8 cell death pathway. Nat Rev Immunol 2002;2:735-47.

9 39. Basu R, Whitlock BM, Husson J, Le Floc'h A, Jin W, Oyler-Yaniv A, et al.

10 Cytotoxic T Cells Use Mechanical Force to Potentiate Target Cell Killing. Cell

11 2016;165:100-110.

12 40. Waugh R and Evans EA. Thermoelasticity of red blood cell membrane.

13 Biophys J 1979;26:115-31.

14 41. Chen; J and Wang N. Tissue cell differentiation and multicellular evolution via

15 cytoskeletal stiffening in mechanically stressed microenvironments. Acta

16 Mechanica Sinica 2018;

17 42. Takeuchi H, Ara G, Sausville EA and Teicher B. Jasplakinolide: interaction

18 with radiation and hyperthermia in human prostate carcinoma and Lewis lung

19 carcinoma. Cancer Chemother Pharmacol 1998;42:491-6.

20 43. Senderowicz AM, Kaur G, Sainz E, Laing C, Inman WD, Rodríguez J, et al.

21 Jasplakinolide's inhibition of the growth of prostate carcinoma cells in vitro

22 with disruption of the actin cytoskeleton. J Natl Cancer Inst 1995;87:46-51.

23 44. Law RH, Lukoyanova N, Voskoboinik I, Caradoc-Davies TT, Baran K,

24 Dunstone MA, et al. The structural basis for membrane binding and pore

25 formation by lymphocyte perforin. Nature 2010;468:447-51.

34

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 45. Leung C, Hodel AW, Brennan AJ, Lukoyanova N, Tran S, House CM, et al.

2 Real-time visualization of perforin nanopore assembly. Nat Nanotechnol

3 2017;12:467-473.

4 46. Pipkin ME and Lieberman J. Delivering the kiss of death: progress on

5 understanding how perforin works. Curr Opin Immunol 2007;19:301-8.

6 47. Vicente-Manzanares M, Ma X, Adelstein RS and Horwitz AR. Non-muscle

7 myosin II takes centre stage in cell adhesion and migration. Nat Rev Mol Cell

8 Biol 2009;10:778-90.

9 48. Kim JS, Kim B, Lee HK, Kim HS, Park EJ, Choi YJ, et al. Characterization of

10 morphological changes of B16 melanoma cells under natural killer cell attack.

11 Int Immunopharmacol 2019;67:366-371.

12 49. Liu Y, Lv J, Liu J, Liang X, Jin X, Xie J, et al. STAT3/p53 pathway activation

13 disrupts IFN-beta-induced dormancy in tumor-repopulating cells. J Clin Invest

14 2018;128:1057-1073.

15 50. Liu Y, Liang X, Yin X, Lv J, Tang K, Ma J, et al. Blockade of

16 IDO-kynurenine-AhR metabolic circuitry abrogates IFN-gamma-induced

17 immunologic dormancy of tumor-repopulating cells. Nat Commun

18 2017;8:15207.

19

35

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Figure legends

2 Figure 1. Soft TRCs are resistant to perforin/SLO-induced pore formation. A,

3 OVA-B16 cells were co-incubated with OVA-CTLs at different ratios for 4 hr. Cell

4 apoptosis was determined by flow cytometry. B, Stiffness of OVA-B16 cells before

5 (n = 340) or after (n = 160) the co-incubation was determined by AFM. C, The

6 stiffness of OVA-B16 or B16 cells or TRCs was determined by AFM. D,

7 OVA-CTLs were co-cultured with OVA-B16 cells or TRCs for 4 hr. Cell apoptosis

8 was analyzed by flow cytometry. Iso indicates isotype control. E, OVA-B16 cells

9 or TRCs were treated with perforin isolated from T cells (PFR-mix) or recombinant

10 perforin (rPFR) for 10 min, and PI was added to the medium. PI+ cells were

11 analyzed by flow cytometry. F, OVA-B16 cells or TRCs were co-cultured with

12 OVA-CTLs for 4 hr, followed by the addition of PI. PI positive cells were analyzed.

13 G, OVA-B16 cells or TRCs were co-cultured with OT-1 T cells for 4 hr. GrB+ cells

14 were determined by flow cytometry. H-I, OVA-B16 cells or TRCs were treated

15 with 50 U rPFR (H) for 10 min, or were co-cultured with OVA-CTLs for 4 hr (I).

16 Cells were imaged by AFM. **p<0.01, *** p<0.001, by 1-way ANOVA (A, E, F,

17 G, H and I) or Student’s t test (B-D). The data represent mean ± SEM (A, D, E, F

18 and G) or mean ± SD (B, C, H and I) of three independent experiments.

19 Figure 2. Tumor cell softness regulates perforin pore formation. A, OVA-B16,

20 A375 or MCF-7 cells were treated with cytochalasin D (Cyto, 1 μM) for 4 hr. Cell

21 stiffness was determined by AFM. B, OVA-B16 cells, pretreated with latrunculin A

36

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 (Lat-A, 1 μM) for 12 hr or Cyto for 4 hr, were then treated with 50 U rPFR for 10

2 min. Cells were imaged by AFM, pore size and number were calculated. C and D,

3 Cyto-pretreated OVA-B16 or A375 cells were treated with PFR-mix (C) or rPFR

4 (D) for 10 min. PI was added and the PI+ cells were determined by flow cytometry.

5 E and F, OVA-B16 TRCs or A375 TRCs were treated with 50 nM jasplakinolide

6 (Jas) for 12 hr. Part cells were used to determine the stiffness (E). Some OVA-B16

7 cells were treated with PFR or SLO for AFM (F). G and H, OVA-B16 TRCs or

8 A375 TRCs were pretreated with 50 nM Jas for 12 hr, and then treated with

9 PFR-mix (G) or rPFR (H) for 10 min. The PI+ cells were analyzed. ** p<0.01, ***

10 p<0.001, by Student’s t test (A, D-F and H) or 1-way ANOVA (B, C and G). The

11 data represent mean ± SEM (C, D, G and H) or mean ± SD (A, B, E and F) of three

12 independent experiments.

13 Figure 3. MYH9 mediates perforin pore formation. A, Binding between PFR-mix,

14 rPFR, rPFRm or negative PFR to the cell lysate of OVA-B16 or MCF-7 was

15 measured by biolayer interferometry. B, The same as (A), except that OVA-B16

16 cells were treated with rPFR and immunoprecipitates were analyzed with

17 anti-MYH, MYO or PFR antibody. C, SGCTRL- or MYH9-SGs- OVA-B16 or

18 MCF-7 cells were lysed and the binding of rPFR or rPFRm to the cell lysate was

19 analyzed by BLI. D and F, The stable knockout of Myh9 or Myo1c in OVA-B16

20 cells was analyzed by western blot. The cells were treated with PFR-mix and PI

21 was added for PI+ cell analysis (D) or confocal microscopy (F). Bar, 10 μm. E,

22 SGCTRL- or MYH9-SGs- OVA-B16 or MCF-7 cells were treated with 50 U rPFR

37

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 for 10 min, and PI was added for PI+ cell analysis. G, SGCTRL- or

2 Myh9-SGs-OVA-B16 cells were treated with 50 U PFR-mix for 10 min, and then

3 imaged by AFM. The formed pore size and number were calculated. ** p<0.01, by

4 1-way ANOVA (D-G). The data represent mean ± SEM (D-F) or mean ± SD (G) of

5 three independent experiments.

6 Figure 4. Perforin drilling force is generated via interaction with MYH9. A and

7 B, SGCTRL-, MYH9-SG1- and MYH9-SG2-MCF-7 or OVA-B16 cells or TRCs

8 were treated with PFR and imaged by AFM with anti-PFR-conjugated tips. The

9 density of adhesion force from data points (n = 500) was shown. The positive

10 control indicated the direct measure of adhesion force from PFR and anti-PFR

11 antibody. C, MCF-7 or OVA-B16 cells or TRCs were pretreated with 25 μM Bleb

12 for 12 hr, and then treated with PFR. Cells were imaged by AFM with

13 anti-PFR-conjugated tips. The frequency of adhesion force distribution from data

14 points (n = 500) was shown. D, F-actin and MYH9 in OVA-B16 or MCF-7 cells or

15 TRCs was determined by confocal microscopy. The quantitation was done with

16 Image J software. Bar, 10 μm. E, The stiffness of SGCTRL or MYH9-SGs-

17 OVA-B16 or MCF-7 cells were measured by AFM. F, Plasma membrane proteins

18 were isolated from OVA-B16 or MCF-7 cells or TRCs and MYH9 was determined

19 by western blot. ** p<0.01, *** p<0.001, by Student’s t test (F) or 1-way ANOVA

20 (E). The data represent mean ± SEM (A-C and F) or mean ± SD (E) of three

21 independent experiments.

38

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Figure 5. Tumor cell softness regulates perforin pore formation in vivo. A,

2 Schematic of the mouse model is shown. OVA-B16 melanoma-bearing mice were

3 adoptively transferred with OT-I cells, followed by treatment with Ble or Jas. B-E,

4 Isolated CD45-CD133- or CD45-CD133+ tumor cells were used for flow cytometric

5 analysis of GrB+ cells (B and D, n = 7) or imaged by AFM (C and E, n = 8). F and

6 G, C57BL/6 (F, n = 10) or NSG (G, n = 10) mice were inoculated with SGCTRL or

7 Myh9-SGs-OVA-B16 cells, and the tumor growth was recorded. H, C57BL/6 mice

8 were inoculated with SGCTRL- or Myh9-SGs-OVA-B16 cells. Mice with 5×5 mm

9 tumor size were adoptively transferred with or without OT-I T cells (4×106 cells)

10 once per 3 days for three times. Tumor growth (left) and tumor size (right) were

11 analyzed (n = 10). I and J, mice (n = 10) were immunized with peptide-Adpgk plus

12 anti-CD40 and poly (I:C) as adjuvant (Adj + Adpgk) or adjuvant alone (Adj),

13 followed by inoculating with SGCTRL- or Myh9-SGs-MC38 cells, outlined in the

14 schematic (I). Tumor growth was analyzed (J). K, mice were immunized, followed

15 by MC38 tumor cell inoculation and treatment with PBS or Bleb. Tumor growth

16 was recorded (n = 15). * p<0.05, ** p<0.01, *** p<0.001, by 1-way ANOVA (B-H).

17 The data are shown as mean ± SD (B-E) or mean ± SEM (F-H).

18 Figure 6. Tumor cell softness affects CTL killing in vivo. A, OVA-B16

19 melanoma-bearing mice (n = 10) were adoptively transferred with OT-I T cells,

20 concomitant with the treatment of anti-PD-1 or Jas for two weeks. Tumor growth

21 (left) and mouse size (right) were recorded. B-D, mice bearing OVA-B16

22 melanoma were adoptively transferred with OT-1 T cells (once per 3 days) for three

39

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 times, and were also treated with PBS, anti-PD-1 (once per two days) or Jas (once

2 per two days) for two weeks. IFN-γ levels in tumor tissues were measured by

3 ELISA (B, n = 5). CD133+ tumor cells were counted by gating the CD45- cells

4 isolated from tumor tissues (C, n = 8). The stiffness of CD45-CD133+ cells and

5 CD45-CD133- cells was determined by AFM (D, n = 20 cells from 5 mice). *

6 p<0.05, ** p<0.01, *** p<0.001, by 1-way ANOVA (A-C), Student’s t test (D).

7 The data are shown as mean ± SD (B-D) or mean ± SEM (A).

8 Figure 7. Cell softness attenuated PFR-induced pore formation in patients. A,

9 CD133+ and CD133- tumor cells (MP-1) or LGR5+ and LGR5- tumor cells (HCC)

10 were isolated and the stiffness was determined by AFM. B, The stiffness of TRCs

11 and the control cells was measured by AFM. n = 50. C and D, The above bulk

12 tumor cells or TRCs were treated with PFR, and PI was added for flow cytometry

13 (C). Melanoma cancer cells and the TRCs were imaged by AFM (D). E, MP-1 and

14 HCC TRCs pretreated with PBS or Jas (50 nM) for 12 hr were treated with PFR,

15 and imaged with anti-PFR-conjugated functional AFM tip. The density of adhesion

16 force distribution was shown. F, MYH9-SGCTRL and MYH9-SGs-MP-1 melanoma

17 cells were treated with PFR or SLO (50 U) for 10 min. PI was added and the PI+

18 cells were analyzed by flow cytometry. G, The schematic diagram of perforin

19 forming pore in bulk cell or TRC. ** p<0.01, *** p<0.001, by Student’s t test (A-D)

20 or 1-way ANOVA (F). The data represent mean ± SD (A-D and F) or mean ± SEM

21 (E) of three independent experiments.

40

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on November 9, 2020; DOI: 10.1158/0008-5472.CAN-20-2569 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Cell softness prevents cytolytic T cell killing of tumor-repopulating cells

Yuying Liu, Tianzhen Zhang, Haizeng Zhang, et al.

Cancer Res Published OnlineFirst November 9, 2020.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-20-2569

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2020/11/07/0008-5472.CAN-20-2569.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet Manuscript been edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2020/11/07/0008-5472.CAN-20-2569. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2020 American Association for Cancer Research.