Meflin-positive cancer-associated fibroblasts enhance tumour response to immune checkpoint blockade therapy


 Cancer-associated fibroblasts (CAFs) are an integral component of the tumour microenvironment (TME). Most CAFs shape the TME toward an immunosuppressive milieu and attenuate the efficacy of immune checkpoint blockade (ICB) therapy. However, the detailed mechanism of how heterogeneous CAFs regulate tumour response to ICB therapy has not been defined. Here, we show that a novel CAF subset defined by Meflin, a glycosylphosphatidylinositol-anchored protein marker of mesenchymal stromal/stem cells, is associated with survival and favourable therapeutic response to ICB monotherapy in patients with non-small cell lung cancer (NSCLC). The prevalence of Meflin-positive CAFs positively correlated with CD4-positive T cell infiltration and vascularization within NSCLC tumours. Meflin deficiency and CAF-specific Meflin overexpression resulted in defective and enhanced ICB therapy responses in xenograft tumours in mice, respectively. These findings suggest the presence of a previously unknown CAF subset that promotes ICB therapy efficacy, which adds to our understanding of CAF functions and heterogeneity.


Meflin is a marker of CAFs present in the stroma of invasive NSCLC tumours
We first examined the expression of Meflin in human lung adenocarcinoma (LUAD) tissues. 3 RNA in situ hybridization (ISH) analysis revealed no apparent Meflin + cells in the normal human 4 lung tissue adjacent to the tumours (Fig. 1a, Extended Data Fig. 1). In contrast, there were 5 several Meflin + stromal cells that infiltrated the tumour cells in invasive tumours (INV) with an 6 extensive fibroinflammatory reaction (Fig. 1a, Extended Data Fig. 1). Interestingly, Meflin + 7 stromal cells were not observed in non-invasive tumours (adenocarcinoma in situ; AIS), whereas 8 they were sparsely present in preinvasive lesions (PIL) with a lepidic growth pattern adjacent to 9 invasive tumours (Fig. 1a, Extended Data Fig. 1). Meflin expression was also observed in 10 stromal cells that proliferate in tumours developed in an autochthonous LUAD mouse model (KP 11 mice), harbouring K-ras G12D and p53 null alleles, following the administration of adenovirus-12 expressing Cre recombinase (37, 38), whereas it was hardly detected in the normal mouse lung 13 tissue (Fig. 1b). Statistical analyses showed that the prevalence of Meflin + cells was positively (PDPN). These data showed that Meflin is a marker of CAFs that proliferate in the invasive 20 stages of both human and mouse LUAD. 21 We observed that CAFs proliferating in human NSCLC exhibited variable expression of 22 Meflin and other CAF markers (Fig. 1e). Duplex ISH staining showed that a substantial fraction 23 of CAFs positive for platelet-derived growth factor receptor α (PDGFRα), an established 24 fibroblast marker (39), was also positive for Meflin. In contrast, the expression of α-SMA, a 25 marker of myCAFs (20, 21, 39), was inversely correlated with Meflin expression; approximately 1 12% of CAFs expressing α-SMA were positive for Meflin, indicating that CAFs with high 2 Meflin expression exhibited low or negative α-SMA expression (Fig. 1e). Meflin was expressed 3 in approximately 33% and 12% of FAP + and PDPN + CAFs, respectively. These data suggest that 4 Meflin defines a population of CAFs, PDGFRα +/α-SMA low/neg FAP +/-PDPN low/neg in human 5 NSCLC. CAF-specific expression of Meflin was also confirmed by the analysis of single-cell  (Fig. 1f). These observations suggest that Meflin is a marker of a CAF 9 subset in human NSCLC.  Table 1, 18 Extended Data Fig 3a). As observed in LUAD samples, Meflin expression was specifically 19 observed in CAFs in the stroma of lung squamous cell carcinoma (LUSC) tissues (Fig. 2a). For 20 the quantification of Meflin + CAFs, we assigned all stromal cells with oval-to spindle-shaped 21 nuclei as CAFs, excluding lymphocytes, erythrocytes, endothelial cells, and macrophages, based 22 on their morphologies revealed by haematoxylin counterstain. Following the criteria described 23 previously (30), we divided 88 NSCLC cases into Meflin-high (≥ 20% Meflin + CAFs) and 24 Meflin-low (< 20% Meflin + CAFs) groups (Fig. 2a, b). 25 8 Consistent with the observation that the infiltration of Meflin + CAFs closely parallels the 1 invasiveness of tumours ( Fig. 1a-d), Kaplan-Meier survival analyses using the log-rank Mantel-2 Cox test revealed that the Meflin-high group exhibited significantly poor overall survival (OS) 3 after surgical resection. The pattern of disease-free survival (DFS) was similar to that of OS, 4 albeit not statistically significant (Fig. 2c). Notably, the effect of a high number of Meflin + CAFs 5 on poor patient outcomes was more evident in the LUAD subpopulation than in the LUSC 6 subpopulation (Extended Data Fig. 3c). However, analyses of OS and DFS using the 7 multivariate Cox proportional hazard regression model revealed no significant correlation 8 between Meflin expression and the outcomes (Extended Data Table 2). 9 The impact of the number of Meflin + CAFs on the outcomes of patients with NSCLC was 10 also not evident based on NSCLC cohorts obtained from the TCGA database. TCGA-LUAD and 11 -LUSC patients were divided into Meflin-high and -low groups with an empirically determined 12 cut-off value so that the ratio of Meflin-high and -low patients was comparable to that of the 13 NUH cohort (Fig. 2d, Extended Data Table 3, 4, Extended Data Fig. 3b, c). These data 14 suggested that the number of Meflin + CAFs does not affect the outcomes of patients with 15 NSCLC, which contradicts the results of our previous studies on pancreatic and colorectal 16 cancers (30, 31). Interestingly, however, although not statistically significant, high Meflin 17 expression tended to correlate with better outcomes (OS and DFS) in advanced stage III patients, 18 but not those in early stages (I and II) (Extended Data Fig. 3d).

20
High infiltration of Meflin + CAFs correlates with favourable response to ICB in NSCLC 21 patients 22 Next, we investigated the involvement of Meflin + CAFs in tumour response to ICB therapy. We 23 conducted a retrospective observational study of 132 patients with NSCLC who had received 24 ICB monotherapy targeting programmed cell death 1 (PD-1) (nivolumab or pembrolizumab) or 25 programmed cell death 1 ligand 1 (PD-L1) (atezolizumab) at NUH (Extended Data Fig. 4a). 1 The patients were divided into Meflin-high (≥ 20% Meflin + CAFs) and Meflin-low (< 20% 2 Meflin + CAFs) groups by ISH analysis (Extended Data Fig. 4b, c). A total of 98 patients were 3 analysed for outcomes including objective response rate (ORR) assessed by immunotherapy 4 Response Evaluation Criteria in Solid Tumours (iRECIST), OS, and Progression-Free Survival 5 (PFS) (Extended Data Fig. 4a-c, Table 1). The exclusion criteria are described in Extended 6 Data Fig. 4a. 7 Interestingly, the data showed that ORR of the Meflin-high group (40.3%, 25 of 62 patients) 8 was significantly higher than that of the Meflin-low group (0%, 0 of 32 patients) (p < 0.0001, 9 Fig. 3a). The threshold of 20% Meflin positivity in all CAFs was found to be the best criterion 10 for predicting response to ICB monotherapy, with an area under the receiver operating 11 characteristic curve (AUC) of 0.632 (95% CI, 0.526-0.738) (Fig. 3b). Kaplan-Meier survival 12 analyses using the log-rank Mantel-Cox test revealed that the Meflin-high group had a 13 significantly favourable prognosis in both OS (p = 0.0281) and PFS (p = 0.0011) than the 14 Meflin-low group (Fig. 3c). The positive correlation between the percentage of Meflin + CAFs 15 and the outcomes was also shown by the multivariate Cox proportional hazard regression model 16 (

Meflin expression in CAFs correlates with high infiltration of CD4 + T cells and tumour
1 vessel area 2 We next examined the correlation between Meflin expression in CAFs and the profiles of 3 tumour-infiltrating lymphocytes (TILs). Multiplex immunofluorescence (IF) staining of the 4 specimens of 32 surgically resected NSCLC tumours who received ICB monotherapy revealed 5 that the number of CD4 + T cells infiltrating the stroma (interstitium), but not the tumour 6 parenchyma, was significantly higher in Meflin-high patients than in Meflin-low patients (Fig.   7 4a). In contrast, the numbers of CD8 + T cells and CD4 + FoxP3 + regulatory T cells were 8 comparable between the two groups ( Fig. 4a). There was also no difference in the numbers of 9 CD45RO + memory CD4 + T cells, CD45RO + memory CD8 + T cells, or CD20 + B cells in both the 10 stroma and tumour parenchyma between the two groups (Extended Data Fig. 6).

11
Our previous study showed that tumours developed in the pancreas of Meflin knockout 12 (KO) mice exhibited a decrease in tumour vessel area accompanied by changes in collagen 13 configuration (30). Higher tumour vascularity is also associated with better tumour responses to 14 ICB therapy in mouse models (43). Immunostaining of the NSCLC tumour samples with anti-15 CD31 antibody showed that the Meflin-high group tumours had greater tumour vessel area than 16 those of the Meflin-low group (Fig. 4b, Extended Data Fig. 7). These data suggest the 17 possibility that the infiltration of Meflin + CAFs is associated with increased tumour vessel 18 perfusion.

20
Defective response of tumours to ICB therapy in Meflin-KO mice 21 Next, we determined whether Meflin expression in CAFs is crucial for the response of tumours 22 to ICB using C57BL/6J wild-type (WT) mice and Meflin-KO mice. We previously reported that 23 Meflin-KO mice displayed decreased spleen weight compared to WT mice (44). Therefore, we 24 first examined the immunophenotype of lymphocytes isolated from the spleen of Meflin-KO 1 mice to compare it with that of WT mice. The data showed no differences in the proportions of 2 CD4 + , CD8 + , and regulatory T cells (Extended Data Fig. 8a, b). 3 We then subcutaneously transplanted syngeneic MC-38 colorectal cancer (CRC) cells, a 4 well-established cell line that is used to study the anti-tumour effect of ICB therapy (2, 3), into 5 WT mice and Meflin-KO mice, followed by intraperitoneal administration of anti-mouse PD-1 6 (mPD-1) antibody or isotype control IgG on day 4, 7, and 10 after transplantation (Fig. 5a). WT 7 mice treated with anti-mPD-1 antibody, but not isotype control IgG, had a statistically better 8 prognosis than Meflin-KO mice (Fig. 5b). The effect of Meflin deficiency on the anti-tumour 9 effect of mPD-1 antibody was also evaluated using a linear mixed-effect model with restricted 10 maximum likelihood estimates, which showed that the suppressive effect of anti-mPD-1 11 antibody on tumour growth was significantly weakened by Meflin-KO (p = 0.0041), although 12 Meflin-KO itself did not exhibit altered tumour growth (p = 0.901, Fig. 5c). 13 We then explored the status of TILs in tumours developed in WT and Meflin-KO mice 14 (Extended Data Fig. 9a). Unfortunately, the infiltration of CD4 + T cells, CD8 + T cells, and 15 CD25 + FoxP3 + regulatory T cells varied across two independent experiments. Therefore, we 16 concluded that, contrary to the analysis of human NSCLC tissues, T cell infiltration was similar 17 between tumours developed in WT and Meflin-KO mice (Extended Data Fig. 9b). Interestingly, 18 we found that the expression of immune checkpoint molecules in some subsets of TILs was 19 higher in tumours of WT mice than that of Meflin-KO mice, which included T-cell 20 immunoglobulin and mucin-domain-containing molecule 3 (TIM-3) on CD8 + and regulatory T 21 cells, PD-1, CD25, cytotoxic T-lymphocyte associated protein 4 (CTLA-4), and inducible T-cell 22 co-stimulator (ICOS) on CD8 + T cells (Extended Data Fig. 9c-e). Previous studies have 23 indicated that several molecules such as PD-1 and ICOS on CD8 + T cells are associated with the 24 activation of anti-tumour immunity and favourable clinical responses to ICB therapy (9, 45-47).

25
These data suggest that Meflin expression in CAFs is associated with TIL activation in mice, but 1 not their recruitment or infiltration into tumours.

2
The role of Meflin in promoting the anti-tumour effect of anti-mPD-1 antibody was also 3 confirmed in another experimental setup, in which we orthotopically transplanted syngeneic 4 EO771 breast cancer (BC) cells (2, 43) into the fourth right mammary fat pad of WT mice and 5 Meflin-KO mice, followed by intraperitoneal administration of anti-mPD-1 antibody or control 6 IgG on day 6, 9, and 12 after the transplantation (Fig. 5d). On day 19, a suppressive effect of 7 anti-mPD-1 antibody on tumour volumes was observed, which was abrogated in Meflin-KO Brunner-Munzel test with Holm-Bonferroni correction (Fig. 5e). The importance of Meflin 10 expression in CAFs upon anti-mPD-1 antibody treatment was measured by the effect size (Cliff's 11 delta) of -0.796 (95.0% CI -1.00 --0.184) (Fig. 5e). Consistent with the analysis of human 12 NSCLC samples, the tumour vessel area in EO771 tumours developed in WT mice was greater 13 than that in Meflin-KO mice (Fig. 5f). Taken together, Meflin expression in CAFs might 14 facilitate the anti-tumour effect of anti-mPD-1 antibody by increasing the tumour vascular bed. (hereafter referred to as Meflin-TO), we administered doxycycline in drinking water (2 mg/mL) 1 to Meflin-TO mice and transplanted MC-38 and EO771 cells subcutaneously and orthotopically, 2 respectively, followed by the analysis of Meflin expression (Extended Data Fig. 10a). 3 Quantitative PCR (qPCR) of the tumour tissue samples revealed that doxycycline induced Meflin 4 expression in Col1a1 + stromal cells, representing CAFs, in tumours developed in Meflin-TO 5 mice administered doxycycline, but not in control mice that lack the Meflin-Cre allele 6 (Extended Data Fig. 10b). ISH and qPCR also confirmed the induced Meflin expression in 7 cultured CAFs isolated from tumours developed in Meflin-TO but not that of control mice 8 (Extended Data Fig. 10c). 9 Consistent with the analysis of tumour vessel area of human NSCLC tissues and tumours 10 developed in WT and Meflin-KO mice (Fig. 4b, 5f), the area of vasculature in tumours 11 developed in doxycycline administered-Meflin-TO was significantly larger than that in Meflin-

12
TO mice not administered doxycycline and control mice that lacked the Meflin-Cre allele 13 (Extended Data Fig. 10d).

17
Genotyping of the mice was done on day 19 (Fig. 6b). The genotypes of the mice were blinded 18 to the investigators during data acquisition and analysis. The results showed that Meflin-TO 19 administered doxycycline and anti-mPD-1 antibody exhibited a more favourable prognosis and 20 response than control mice (Fig. 6c, d). These data supported the notion that Meflin is a CAF  In the present study, we focused on the role of Meflin, a recently identified rCAF marker in 2 pancreatic and colorectal cancers (30, 31), in tumour response to ICB therapy through the 3 analyses of human NSCLC samples and xenograft tumour mouse models. Our data suggest that 4 Meflin expression in CAFs correlates with favourable tumour response to ICB therapy, leading 5 to the hypothesis that Meflin + CAFs promote the host anti-tumour immune response. Previous 6 studies have shown that immunosuppressive CAFs, such as α-SMA + FAP + CAFs, LRRC15 + 7 CAFs, and TGF-β-activated CAFs, suppress anti-tumour immunity and are associated with ICB 8 therapy failure (24, 27-29). We propose that a balance between the immunosuppressive CAFs 9 and Meflin + rCAFs is crucial for determining the net response to ICB therapy (Fig. 6e). 10 Given our initial data that Meflin expression in CAFs correlated with the favourable 11 outcomes of NSCLC patients treated with ICB, it was an unexpected finding that the repertoire 12 of TILs was almost comparable between Meflin-high and -low groups, except for CD4 + T cells.   Consistent with this, the present data showed that the tumour vessel areas correlated with Meflin 8 expression in CAFs in both human NSCLC tissues and mouse models. Although not proven in 9 the present study, an intriguing hypothesis is that Meflin-mediated suppression of tissue fibrosis 10 or decrease in interstitial pressure facilitates tumour vessel perfusion and therapeutic antibody 11 delivery. In addition, the involvement of Meflin in controlling the enhanced permeability and 12 retention (EPR) effect, which refers to the ability of macromolecules such as anti-PD-1 13 antibodies to accumulate in the tumour tissue (57, 58), will be a subject of future research.
14 An appealing feature of Meflin is that none (0%) of the NSCLC patients in our institution 15 with low Meflin expression in CAFs responded to ICB therapy. These data suggest that the 16 number of Meflin + CAFs could be a marker for identifying patients who will not benefit from 17 ICB therapy. The present study also showed that the induced expression of Meflin in CAFs  immune checkpoint inhibitors and chemotherapeutic agents (clinicaltriaol.gov) (60). It would be 1 interesting to study how Meflin is involved in vitamin D-mediated remodelling of the TME and 2 the increased efficiency of ICB therapy in clinical settings.

3
In conclusion, we identified a CAF subset marked by Meflin expression and found that its 4 prevalence is associated with a favourable response to ICB therapy in NSCLC patients and 5 xenograft tumour mouse models. Induction of Meflin expression in CAFs augmented the tumour 6 response to ICB therapy in mice. Together with other studies that identified CAF subsets that 7 suppress anti-tumour immunity and are associated with ICB treatment failure, we propose that 8 the heterogeneity of CAFs determines the net response of tumours to ICB therapy.                                                     and pan-cytokeratin (Pan-CK), followed by imaging with a multispectral imaging system. Boxed        NSCLC who underwent surgery (surgery cohort, n = 147; Extended Table 1). The other was a 7 cohort of patients with advanced or recurrent NSCLC who received programmed cell death 1 8 (PD-1) or programmed cell death 1 ligand 1 (PD-L1) antibody-based immune checkpoint 9 blockade (ICB) monotherapy (ICB cohort, n = 132, Table 1). All patients consented to the 10 Institutional Review Board-approved protocols permitting specimen collection. Meflin-KO mice, and Meflin-Cre mice have been described previously (30, 37, 44). 19 We generated a transgenic mouse line carrying a third-generation tetracycline-response volumes of MC-38 tumours were measured and calculated two to three times per week using the 2 modified ellipsoid formula as follows: 1/2 × (length × width 2 ). Mice with tumour volumes 3 >2,000 mm 3 were sacrificed. Animals whose tumours were ulcerated with bleeding before 4 progression were terminated and included in the study. was four days after tumour inoculation at a dose of 200 μg/body, followed by subsequent 10 antibody administration on day 7 and 10 at the same dose. For EO771 orthotopic tumour models, 11 antibody administration was initiated on day 6 after palpable tumours had formed, followed by 12 antibody administration on day 9 and 12 at the same dose via the same route.

14
Tumour growth/tumour volume analysis A linear mixed-effects model was used to examine 15 repeated measurement data to investigate the effect of genotype (G: WT or Meflin-KO), 16 treatment (T: control or PD-1), and interaction between G and T on tumour volume over time.

22
Here, β0, β1, β2, β3, β12, β13, and β123 are the coefficients of the fixed effects. b0 is the random 1 effect of the intercept, and b1 is the random effect of the slope. σ 2 B0 is the variance of the 2 individual difference at the baseline, σ 2 B1 is the variance of the individual difference of the slope,