Copy number gain of CMTM6 increases the expression of PD-L1 in undifferentiated pleomorphic sarcoma

Undifferentiated pleomorphic sarcoma (UPS) is a sarcoma with a poor prognosis. A clinical trial, SARC028, revealed that treatment with anti-PD-1 drugs was effective against UPS. Studies have reported that UPS expresses PD-L1, sometime strongly ( ≥ 50%). However, the mechanism of PD-L1 expression in UPS has remained still unclear. CKLF-like MARVEL transmembrane domain containing 6 (CMTM6) was identied as a novel regulator of PD-L1 expression. The positive relationship between PD-L1 and CMTM6 has been reported in several studies. The aim of this study was to examine CMTM6 expression in UPS and evaluate the relationship between PD-L1 and CMTM6.


Introduction
Undifferentiated pleomorphic sarcoma (UPS) is a sarcoma with a poor prognosis, the tumors of which do not show speci c differentiation. It is diagnosed by excluding other sarcomas. It is usually treated with excision, chemotherapy, and radiation therapy, but its prognosis is still unfavorable (WHO classi cation of Tumours of Soft Tissue and Bone, 2020). A clinical trial (SARC028) reported that treatment with anti-PD-1 therapy was bene cial for the survival of UPS patients (Tawbi et al. 2017). In addition, some investigations demonstrated that UPS cells express PD-L1 (Boxberg et al. 2018; Keung et al. 2020; Ishihara et al. 2020). In detail, 30% or more of UPS cases were found to express PD-L1 (tumor proportion cut-off ≥ 1%), while approximately 10% of them strongly express PD-L1 (tumor proportion cut-off ≥ 50%) (Boxberg et al. 2018;Ishihara et al. 2020). PD-L1 expression is regulated by IFN-γ secreted by immune cells (Schalper et al. 2017;Burr et al. 2017a;Mezzadra et al. 2017;Ishihara et al. 2020). We reported that IFN-γ induces PD-L1 expression in UPS and that PD-L1 expression (cut-off ≥ 1%) in UPS positively correlated with the level of tumor-in ltrating lymphocytes (TILs) (Ishihara et al. 2020). However, in our previous study, cases with PD-L1 expression (≥ 50%) were not signi cantly associated with TILs (Boxberg et al. 2018), suggesting that factors other than TILs also regulate PD-L1 expression. Boxberg et al. investigated the association between increased copy number of PD-L1 and PD-L1 expression but did not nd a signi cant correlation (Ishihara et al. 2020).
It has been reported that CKLF-like MARVEL transmembrane domain containing 6 (CMTM6) protects PD-L1 from lysosomal degradation and promotes its expression on the cell membrane (Burr et al. 2017b;Mezzadra et al. 2017). CMTM6 acts on PD-L1 by a different route from the stimulation by IFN-γ (Burr et al. 2017b;Mezzadra et al. 2017). A study on non-small-cell lung cancer reported that CMTM6 expression positively correlated with PD-L1 expression and that CMTM6 was a predictor of the therapeutic effect of an anti-PD-1 drug (Koh et al. 2019). In addition, in cancers such as hepatocellular carcinoma and gastric cancer, CMTM6 has been reported to be coexpressed with PD-L1 (Li et al. 2020;Liu et al. 2020).
The purpose of this study is to evaluate the expression of CMTM6 in UPS and to assess the relationship between PD-L1 and CMTM6. In addition, The Cancer Genome Atlas (TCGA) data on CMTM6 were analyzed and the correlation of CMTM6 and PD-L1 was assessed.

Patients and materials
This study was conducted in accordance with the principles of the Declaration of Helsinki. It was also approved by the Ethics Committee of Kyushu University  and consent was obtained from patients who donated tissue. A total of 51 UPS cases, diagnosed as malignant brous histiocytoma or UPS, were retrieved from among soft-tissue tumors registered at the Department of Anatomic Pathology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan, from 1998 to 2019.
To collect de nitive UPS samples, secondary sarcomas, radiation-associated sarcomas, myxo brosarcomas, dedifferentiated liposarcomas, and other sarcomas were excluded (Ishihara et al. 2020). A total of 48 tumors were the same as in this previous study, while three were newly retrieved. Four tumors that had been used in the previous study were not used in this study because the formalin-xed, para n-embedded (FFPE) samples had been used up (Ishihara et al. 2020).
PD-L1 was assessed by determining the proportion of membranous staining-positive tumor cells relative to all tumor cells, as described previously (Ishihara et al. 2020). CMTM6 evaluation was performed in a manner similar to that used for estrogen and progesterone receptor evaluation in breast cancer using the Allred score. Proportion score (PS) was assessed by determining the proportion of tumor cells revealing cytoplasmic positivity for staining relative to all tumor cells on the slide. PS of CMTM6 was de ned as follows: PS 0: no positive cells, PS 1: > 0% and < 1% positive tumor cells, PS 2: ≥ 1% and < 10%, PS 3: ≥ 10% and < 33%, PS 4: ≥ 33% and < 66%, and PS 5: ≥ 66% and ≤ 100%. Intensity score (IS) was assessed using the intensity of the staining as follows: IS 0: negative, not stained, IS 1: weak cytoplasmic staining and no membranous staining, and IS 2: strong cytoplasmic and strong membranous staining. In the Allred scoring system, IS is classi ed into four groups: "negative," "weak," "intermediate," and "strong." However, in this study, cases were classi ed into three groups to maintain reproducibility, because it was di cult to distinguish between "intermediate" and "strong". We set the cut-offs of PS and IS of CMTM6 immunostaining for statistical analysis by drawing ROC curves.

Copy number assay
Copy number assay was conducted using TaqMan quantitative PCR (Thermo Fisher Scienti c Inc.). We used the following primers: CMTM6 (catalog number: 4400291; Thermo Fisher Scienti c Inc.) and RNase P (catalog number: 4403326; Thermo Fisher Scienti c Inc.). For the copy number assay, 13 frozen samples and 14 FFPE samples were used. The copy number assay was not conducted for some cases because the sample volume was small or little sample was left. A normal tissue sample of a UPS case was also tested as a control. Finally, DNA was extracted in 28 cases. PCR was performed with THUNDERBIRD Probe qPCR Mix (TOYOBO), using the following protocol on a Step One Plus Real Time PCR System (Thermo Fisher Scienti c Inc.): pre-denaturing at 95 °C for 10 min; and then 45 cycles of denaturing at 98 °C for 15 s, annealing at 60 °C for 15 s, and extension at 68 °C for 30 s. The obtained relative quantities were processed using CopyCaller (Thermo Fisher Scienti c Inc.) and copy numbers were determined.
Subgrouping in accordance with PD-L1, CMTM6, and tumor-in ltrating lymphocytes We divided the UPS cases into subgroups according to the PD-L1 expression, the level of CD8-positive TILs, and the CMTM6 expression. In this subgrouping, the cut-off of PD-L1 was set as 1% because this is the threshold for administering anti-PD-1 therapy for many cancers. We counted CD8-positive TILs in UPS in ve randomly chosen high-power elds, as previously described (Ishihara et al. 2020). Cases were classi ed as TILs-high or TILs-low according to the cut-off, which was determined by drawing ROC curves (Supplementary Figure 1). If either PS or IS of CMTM6 was evaluated as high, CMTM6 expression was considered to be positive. DNA copy number and mRNA expression pro ling using TCGA data DNA copy number data and mRNA expression data of 50 UPS cases in TCGA (https://www.cancer.gov/tcga) were analyzed using the cBio Cancer Genomics Portal (cBioPortal), an open platform for evaluating genomic data (http://www.cbioportal.org/) (Gao et al. 2014;Cerami et al. 2017).

Statistics
The relationship between the immune expression of CMTM6 and PD-L1 was analyzed by Fisher's exact test. Steel's multiple comparison test was used to analyze the correlation between CMTM6 mRNA and DNA copy number in TCGA data using cBioPortal. The relationship between immunohistochemistry (IHC) data and DNA copy number of CMTM6 was analyzed using the Mann-Whitney U test. Survival curves were constructed using the Kaplan-Meier method. Overall survival and disease-free survival curves were analyzed using the log-rank test. Multivariate analysis was conducted using Cox's proportional hazard model. We used the JMP statistical software package (version 14; SAS Institute) for analysis. A p-value of < 0.05 was considered signi cant in the statistical analysis.

TCGA data analysis
The analyzed data are presented in Figure 1. There was a positive correlation between the mRNA expression of PD-L1 and that of CMTM6 (p = 0.0431 and R 2 = 0.08, Figure 1a). A list of the genes encoding mRNAs coexpressed with CMTM6 is shown in Figure 1b. The genes most closely related to CMTM6 are presented at the top. Many genes located around 3p22 on the short arm of chromosome 3 are listed. The copy number segment of UPS cases is shown in Supplementary Figure 2. The relationship between mRNA expression and copy number variation of CMTM6 is presented in Figure 1c. The mRNA expression of CMTM6 in UPS with CMTM6 copy number gain increased compared with that in UPS with diploid CMTM6 copy number (shallow deletion vs. diploid: p = 0.112, diploid vs. gain: *p = 0.025).

Clinicopathological data analysis
The clinicopathological data of the retrieved UPS cases are summarized in Table 1. A total of 27 (52.9%) cases were younger than 69 years of age and 24 (47.1%) cases were 69 years of age or older. A total of 27 (52.9%) cases were male and 24 (47.1%) were female. Data on tumor size were available for 43 cases: 12 (27.9%) cases had tumors of 5 cm or smaller and 31 (72.1%) had tumors larger than 5 cm. Three (5.9%) cases arose in the soft tissues of the head and neck, nine (17.6%) in the upper extremity, eight (15.7%) in the trunk, and 31 (60.8%) in the lower extremity. A total of 26 (51%) and 25 (49%) cases were evaluated as FNCLCC grades 2 and 3, respectively.

The relationship between CMTM6 and PD-L1 expression in IHC
The relationship between CMTM6 and PD-L1 expression is shown in Supplementary Figure 3. The cutoffs of PS and IS of CMTM6 were determined by drawing ROC curves (Supplementary Figure 4). On the AUC of the curves, it was when CMTM6 surrogated strong PD-L1 (≥ 50%) rather than PD-L1 (> 1%) that PS and IS of CMTM6 became a highly speci c factor. Therefore, cut-offs for determining strong PD-L1 (≥ 50%) were used; the cut-offs for PS and IS were 3 and 2, respectively. The relationships between PS and IS of CMTM6 and PD-L1 are summarized in Table 3. Positive correlations between the expression of PD-L1 (≥ 50%) and PS or IS of CMTM6 were seen (both *p=0.023). The correlations between the expression of PD-L1 (≥1%) and PS or IS of CMTM6 were not signi cant.
CMTM6 copy number analysis CMTM6 copy number in UPS is presented in Figure 3a. Results were obtained for 14/28 cases, with 14 cases not being properly tested due to sample defects. Normal tissue was tested as a control. CMTM6 copy numbers ranged from 1 to 13 (mean 3.6). IHC data for PD-L1 and CMTM6 are shown below. The relationship between CMTM6 copy number, obtained through qPCR, and CMTM6 protein expression, revealed by IHC, was analyzed (Figure 3b and 3c). CMTM6 copy number signi cantly correlated with the PS of CMTM6 (*p = 0.030). CMTM6 copy number was not signi cantly associated with the IS of CMTM6.

Survival analysis
The outcome of survival analysis is shown in Figure 4. Prognostic information for 41 cases was available. Although patients with UPS and higher PS of CMTM6 did not have worse prognosis for overall survival (Figure 4a The relationships between PD-L1, CMTM6, and tumor-in ltrating lymphocytes Figure 5 shows the subgrouping of UPS according to the level of CD8-positive TILs and CMTM6 expression. Among the 20 cases expressing PD-L1 (≥ 1%), 3 cases expressed CMTM6 but had low CD8positive TILs, 11 cases expressed CMTM6 and had high CD8-positive TILs, and 6 cases did not express CMTM6 but had high CD8-positive TILs.

Discussion
UPS cells express PD-L1 and the PD-L1 expression is regulated by IFN-γ secreted by immune cells (Schalper et al. 2017;Burr et al. 2017a;Mezzadra et al. 2017;Ishihara et al. 2020). IFN-γ induces PD-L1 expression and that PD-L1 expression (cut-off ≥ 1%) positively correlate with the level of TILs. The same phenomenon was noted in UPS. However, our previous study suggested that factors other than TILs also regulate PD-L1 expression. In the current study, we revealed the positive correlation between PD-L1 expression and CMTM6 expression.
CMTM6 expression was here shown to be positively correlated with PD-L1 expression in UPS. This nding is similar to those in previous reports on lung cancer, hepatocellular carcinoma, and gastric cancer (Koh et al. 2019;Li et al. 2020;Liu et al. 2020). CMTM6 has been reported to inhibit its lysosomal degradation, possibly by preventing its ubiquitination, and to promote PD-L1 expression on the cell membrane (Burr et al. 2017a;Mezzadra et al. 2017). In the current study, the same mechanism was also considered to be involved, in that CMTM6 prevented the lysosomal degradation of PD-L1 and assisted its membranous expression. Interestingly, CMTM6 expression closely and positively correlated with the strong expression of PD-L1 (≥ 50%). All ve UPS cases with strong PD-L1 expression (≥ 50%) exhibited high CMTM6 expression. In UPS, like in other cancers, PD-L1 is induced by IFN-γ (Burr et al. 2017b;Schalper et al. 2017;Mezzadra et al. 2017;Ishihara et al. 2020). However, our ndings suggest that strong PD-L1 expression in UPS requires high CMTM6 expression, in addition to the stimulation of IFN-γ.
CMTM6 copy number was found to be positively correlated with CMTM6 expression in the current study. Analysis of TCGA data revealed that the mRNAs encoded by some genes located around 3p22 were coexpressed with CMTM6 mRNA. UPS is known to have a complex karyotype, with chromosome numbers ranging from haploid to polyploid (WHO classi cation of Tumours of Soft Tissue and Bone, 2020). Data on mRNA expression suggested that the copy number of the genes located around 3p22 including CMTM6 often varies because of the chromosomal instability of UPS and that the impact of copy number variation of CMTM6 on its mRNA expression would be greater in UPS than that of other factors. It is possible that UPS cases have CMTM6 copy number variation, sometimes exhibit copy number gain, have upregulated CMTM6 expression, and strongly express PD-L1.
Tumors with positivity for CMTM6 and a low level of CD8-positive TILs were identi ed in the present study. It is generally considered necessary for tumor immunity that tumor cells express neoantigen (Li et al. 2020) (Gao et al. 2014) (Cerami et al. 2017). However, it is unclear whether or not UPS cases with PD-L1 expression, high CMTM6 expression, and low CD8-positive TILs express neoantigen; therefore, the effect of anti-PD-1 drugs on UPS should be evaluated from the perspective of CMTM6 expression. In addition, when using anti-PD-1 drugs for such tumors, it may be desirable to use drugs that promote cancer immunity, such as anti-CTLA4, as activators of the immune system. CMTM6 protein is suggested as a useful prognostic marker for the effect of anti-PD-1 drug in patients with non-small-cell lung cancer (Schalper et al. 2017). The same would also be true for UPS, a tumor with copy number variation of CMTM6. The current study could provide a basis for predicting the e cacy of anti-PD-1 therapy in UPS by evaluating CMTM6 expression.
Analysis of TCGA data revealed that the levels of PD-L1 and CMTM6 mRNAs were signi cantly positively correlated in UPS, which is consistent with the ndings in lung cancer (Schalper et al. 2017). However, this correlation cannot be explained by the mechanism that CMTM6 assists PD-L1 to avoid lysosomal degradation. Instead, it is possible that there is some transcriptional regulator in common between PD-L1 and CMTM6, but this remains unknown. Strong immunostaining of CMTM6 was a poor prognostic factor in the current study. Therefore, CMTM6 inhibition may improve the prognosis of patients with UPS.
However, CKLF-like MARVEL transmembrane domain containing 4 (CMTM4) was reported to interact with PD-L1 and restore PD-L1 expression in CMTM6-de cient melanoma cells (Mezzadra et al. 2017). Thereby, when using an anti-CMTM6 drug, the use of an anti-CMTM4 drug in combination is suggested.