Nature and Signicance of Stromal Differentiation, PD-L1, and VISTA in GIST: Shifting Current Paradigms

The nature and signicance of stromal differentiation (SD), program death-ligand 1 (PD-L1), and v-domain Ig suppressor of T cell activation (VISTA) in gastrointestinal stromal tumor (GIST) is largely unknown. Looking forward, the assessment of SD and immune check point inhibition will become more ubiquitous in surgical pathology. Immature, myxoid stroma has been found to be a poor prognostic signature in many cancer subtypes (colon, breast, cervix, esophagus, stomach); although little is known regarding its signicance in GIST. For immune check-point inhibition, studies have demonstrated high PDL1 and expression to be associated with patient outcomes in numerous cancer subtypes. The present body of work aims to discover the role of SD, PD-L1 and VISTA: both in terms of its nature, and its signicance in a clinical setting. Here we found PD-L1 expression in immune cells (IC) and immature SD to be associated with worse cancer free survival, while positive VISTA expression was found to be associated with improved outcomes. High-grade immature SD had the highest propensity for death/recurrence and was the only variable found to have prognostic signicance on multivariate analysis. Our ndings support the evaluation of SD, PD-L1 and VISTA in GIST, with clinical practice implications for pathologists. Ultimately, we hope our ndings lead to improved prognostication, further optimization of therapeutics, and improved outcomes in a true clinical environment. For GIST, tumoral differentiation (TD) may be performed in isolation by pathologists, but cautiously; PD-L1 and VISTA could be on the cutting edge of an immunotherapeutic revolution, and SD may be the answer to clinical heterogeneity in GIST. We hope that this work can serve as a lexicon and a guide for discovering the essence of stroma in GIST, while also being a catalyst for pathologists to shoulder the adoption of SD in cancer. disease-free survival rate as a function of time. The two tailed Fisher exact test was used to assess relationships between stromal differentiation, immune checkpoint inhibition and the clinicopathological prole. Statistical Analysis was performed using Prism graphpad 8.4.2. Cox-regression was performed on SPSS 1.0.0.1508. A p value of <0.05 was used to indicate statistical signicance.


Introduction
Gastrointestinal stromal tumor (GIST) is the most common mesenchymal neoplasm of the gastrointestinal tract and harbors a propensity for malignant transformation 1 . Risk strati cation is based on tumor size, mitotic rate, location, and perforation status; however, stromal differentiation 2 (SD) and immune check-point inhibition 3 are garnering signi cant attention in the world of oncology. Today, more than 35 years after Dvorak 4 colorfully termed cancer "a wound that does not heal," our understanding of the stroma may be coming of age.
It has been long understood that biological processes controlling wound response also control developing tumoral stroma 5 . Current research paradigms are transitioning to the extracellular matrix, deciphering cellular components of the tumoral microenvironment, and how they are in uenced by each other. Now we are beginning to truly understand the signi cance of this biology, perhaps in the form of SD.
In the eld of histopathology, SD has been recently described in many notable cancer subtypes: colon, breast, cervix, esophagus and stomach [6][7][8][9][10] . In the breast, SD has been found to predict clinical outcomes, immune pro les and molecular phenotypes 9 . Overall, immature stroma is a bad player, being associated with reduced survival and higher pathologic stage 2 .
The evaluation of immune check-point inhibition is becoming more and more ubiquitous to the practice of surgical pathology. Here protein expression is evaluated from immunohistochemically stained slides, and positive expression can determine therapeutic candidacy: such as seen with Atezolizumab targeting the program death-ligand 1 (PD-L1) in triple negative breast cancer 11 .
In 2015, Bertucci et al. 12 was the rst to evaluate PD-L1 expression in GIST, they found PD-L1 expression to predict metastatic relapse. Low PD-L1 expression was found to corelate with higher metastatic risk independent of the AFIP classi cation and the KIT mutational status 12 . This suggests that evaluating PD-L1 in GIST may help improve outcomes and better tailor adjuvant imatinib; in the metastatic setting, Importantly, PD-L1 may be able to guide the use of immune check-point inhibition.
It is important to remember that immune escape in cancer is not restricted to the program death ligand axis. Markers such as v-domain Ig suppressor of T cell activation (VISTA) are on the cutting edge and as of now 13 . studies have described the suppressive effect of VISTA, with a presumed e cacy for anti-VISTA therapy. However, VISTA is highly controversial: it acts as a ligand on antigen-presenting cells, while also serving as a receptor on T cells 13 . Clinical trials with anti-VISTA are ongoing (Clinicaltrials.gov) and studies correlated with the expression and VISTA as a poor prognostic signature will be important in justifying its use.
This body of work is one of the larger studies on GISTs to date. Prior to this, no studies have evaluated SD in GIST, and for immune check point inhibitors we are living in the equivalent to the wild west era of the American frontier. It can certainly be said that the expression of VISTA in GIST and its relationship to the clinical pro le is not understood, as it has never been evaluated before. Innovations here could lead to improved prognostication, further optimizing therapeutics and improved patient outcomes in a true clinical environment.
Looking forward, GIST patients may bene t from a more holistic diagnostic approach, one which incorporates the underlying tumoral microenvironment. The current state of art as proposed by Fletcher et al. 14 is somewhat rudimentary in nature, being primarily based off mitotic indices, tumor size and kit immunopositivity 14 , these variables do not allow one to discover the essence of stroma, so we must shift current paradigms.

Materials And Methods
Institutional Review Board: Institutional Review Board approval from the O ce of the Human Research was obtained for all experiments. All methods were carried out in accordance with relevant guidelines and regulations. Patient consent was not required by the institutional review board (IRB) committee due to the retrospective nature of the study (Northwell Health IRB #: 19-1066).

Design:
This study was retrospective in nature and included patients diagnosed at the Northshore University and Long Island Jewish Hospitals, part of the Northwell Health system. Routine hematoxylin and eosin (H&E) stained slides and representative cancer blocks were collected for GIST cases between 2010 to 2017. We collected resection specimens which lacked neoadjuvant chemo/radiation therapy. One representative block was selected per case from a single-slide containing the tumor's deepest invasion. Both spindle cell and epithelioid GIST cancers were collected (FIG 1). Clinicopathological data was collected and reviewed for each case including age, gender, tumor site, pathologic stage, tumor size, tumor site/surgery, as well as the histopathology. Patients were retrospectively followed for cancer free survival (CFS): de ned as any recurrence, second primaries or death by the Northwell Health Cancer Registry.

Digital Slides:
We utilized digital slides for the purpose of analyzing histology and immunohistochemistry for this study.
Whole slide images were accessed through the Aperio vendor agnostic whole slide image viewer and slides were scanned on a Leica Aperio AT2 (Leica Biosystems, Buffalo Grove, Illinois, USA) whole slide scanner at 20×. Digital PHH3 was calculated on whole slide images (WSI) in the tumoral hotspot containing the greatest mitotic activity. The number of digital PHH3 positive cells was counted in 0.5 mm2 automatically using the Ventana Virtuoso software (Roche Diagnostics, North America).

Stromal Differentiation:
Scoring was based on the 3-tier grading system proposed by Hacking et al. 2 (TAB 1). Mature stroma (SD1) was composed of mature collagen bers, strati ed into multiple brous bands at tumoral front. We assessed immature stroma in a semi-quantitative manner based on the degree of myxoid, amorphous SD. Immature, myxoid stroma contained basophilic to grey extracellular matrix and often intermingled with hyalinized collagen. Low grade stroma usually contains high variability in absolute difference in intensity between stromal matrix regions (mosaic pattern), less contiguous areas of myxoid stroma (FIG  2). Low grade should contain a 40x eld of myxoid stroma and intermediate stroma containing "keloid like" can usually be categorized as low-grade. High grade usually contains low variability in absolute difference in intensity between stromal matrix regions, with stromal cells surrounded by contiguous regions of myxoid stroma, being the predominate stromal pattern (>50%) at the extramural tumor front.

Histologic Grade:
We based our grading off Fletcher et al. 14 who creased the NIH classi cation, which was the rst to classify GIST. It determined the risk of strati cation based on a mitotic count higher than 5/50 high power elds (HPF). conventional counting of mitotic cells was performed consecutively in HPFs, which allowed for a mitotic rate per 5 mm 2 surface area calculated according to a 400x eld diameter.
Statistical Analysis: The primary objective of the statistical analysis was to assess the relative prognostic importance of stromal differentiation and immune checkpoint inhibition in GISTs. Descriptive statistics such as frequencies and percentages were calculated for the categorical variables. The Kaplan-Meier method was used to evaluate the disease-free survival rate as a function of time. The two tailed Fisher exact test was used to assess relationships between stromal differentiation, immune checkpoint inhibition and the clinicopathological pro le. Statistical Analysis was performed using Prism graphpad 8.   Figure 4(a) and bar graphs for PD-L1 and VISTA scoring is presented in Figure 4(b) for both the TC and IC components. Based on non-linear regression data in Figure 4(c), a combined score of 3 was used as a cutoff for PD-L1 and VISTA.

Discussion
The present study elucidates the role of stromal differentiation and immune check-point inhibition (PD-L1 and VISTA) in GISTs. In addition to having clinical practice implications for patients with GIST, our ndings may also be transferable to other cancer subtypes.
The differentiation of the extracellular matrix leads to characteristic immature, myxoid stroma seen on routine histologic evaluation. In the present study we found immature stroma to be associated with higher anatomical extent of disease, higher tumoral grade, higher mitosis, negativity for and worse CFS. This is reminiscent of what was seen in breast cancer 9 , where mature stroma was also found to be associated with PD-L1 expression. Importantly, high grade differentiation (SD3) was found to have the highest risk of recurrence and was the only variable in GIST found to be statistically signi cant on multivariate analysis. Looking forward, the Hacking Classi cation of Stromal Differentiation could be used in clinical practice, identifying patients with poor prognostic outcomes, allowing them to be treated more aggressively.
Understanding the peculiarities of the ECM in GIST could lead to therapeutic opportunities; myxoid degeneration of the ECM has been demonstrated to decrease the physical barrier in the cancer microenvironment, which can improve the delivery of therapeutics, nutrients, and immune cells 16 . Therefore, mature SD could act as barrier, which is supported by preclinical studies showing ECM degradation to improve drug uptake and response 16 In this setting, myxoid stromal degeneration could be good and bad: while it may have an improved therapeutic response, these patients do have the highest risk of recurrence.
For PD-L1 and VISTA, this is the second thorough study to explore PD-L1 12 and the rst to explore VISTA in GIST. Similar to Bertucci et al. 12 we also found PD-L1 expression to be associated with poor prognostic outcomes and recurrence in GIST. In mouse models, PD-1 expression on T cells has been found correlate with imatinib treatment in GIST 17 . Meanwhile, treatment with imatinib inhibited the IFNγ-induced upregulation of PD-L1 via STAT1 inhibition 17 , suggesting a role for PD-L1 based combination therapy.
It is interesting that VISTA was found to be protective in GIST when positive in ICs. While interpreting these prognostic ndings, it is important to understand the complexity of VISTA as an immune check point regulator and the tumoral microenvironment in general. VISTA is expressed on both CD4+/Foxp3+ regulatory T cells and myeloid-derived suppressor cells (MDSCs) 18  These ndings may be secondary the effect of VISTA in converting naïve T cells into FoxP3 expressing T cells 22 . FoxP3 has been shown to improve survival in patients with colorectal cancer 2,23 but the prognostic effect varies in other cancer types 20 . FoxP3 plays a role as a transcriptional repressor for the proto-oncogene SKP2, and in-turn regulates the cell cycle in the G2/M phase 2,24 . Decreased FoxP3 expression promotes SKP2 and cell proliferation. This may explain why manual mitotic count and digital PHH3 was decreased in patients with signi cant VISTA expression in immune cells.
Our ndings do support that VISTA may be immune-protective for patients with GIST. However, VISTA was found to be a poor prognostic signature in melanoma 21 , suggesting a multifaceted role of VISTA in different tumor subtypes.
PD-L1 and VISTA are molecules that can be targeted by synthesized antibodies and for GIST, clinical trials using immune checkpoint inhibition in combination with anti-KIT and chimeric antigen receptor (CAR) T-cells have shown promising results 22 . The treatment of GIST and the use of immunotherapy is evolving; robust, well-designed clinical trials will have the ultimate say in determining therapeutic response.
Our study did not come without pitfalls, rstly this study was retrospective, which allows for the potential of bias and secondly we were unable to perform molecular testing for SD, PD-L1, and VISTA. Thirdly, we were unable to perform multiplex immunolabeling, this would have allowed us to identify PD-L1 and VISTA expression in different immune cell subtypes.
The use of digital image analysis has been shown to outperform manual analysis 23 ; in our study we also found digital PHH3 to be comparable to manual mitotic count. Coupling our whole slide images with deep learning 24 approaches could have improved prognostication; future studies should integrate computational approaches to access the tumoral microenvironment.
To conclude, SD, PD-L1 and VISTA evaluation may help tailor the need for adjuvant imatinib therapy, which can be given alone or in combination with immune check point inhibition. The status of a patients SD may be the most important factor for predicting recurrence, and for GIST patients with high-grade SD, clinical trials will be needed to determine whether aggressive therapeutics are bene cial. For GIST, tumoral differentiation may be performed in isolation by some pathologists, but cautiously: PD-L1 and VISTA could be on the cutting edge of an immunotherapeutic revolution, and SD may nally be the answer to clinical heterogeneity in GIST we have been searching for.

Data Availability
Pathology data and the statistical analyses for the current study are available from the corresponding author upon reasonable request.

Funding
No funding was provided for the production of this manuscript.

Disclosure
The authors report no con icts of interest in this work.