Programmed Death Ligand-1 Expression Is Associated With Poorer Survival in Anal Squamous Cell Carcinoma Open Access

After institutional review board approval, we performed a search for cases of ASCC for the period of 2010–2018 within Epic (Epic Systems Corporation; the patient electronic medical record database) of Grady Memorial Hospital, a major urban hospital in Atlanta, Georgia. Pretreatment specimens were selected. In cases with multiple blocks, the block with the best representative section of tumor was selected. Paraffin-embedded blocks were retrieved in 58 cases. Only 2 cases had resections; the remaining cases were pretreatment biopsies. Clinicopathologic features including age, sex, HIV status, HIV viral load, CD4 count, and stage of cancer were collected. HIV viral load and CD4 values selected were those reported at the time of diagnosis of ASCC or the closest available to that date. HIV viral load was subcategorized as negative (<1.6 log₁₀ copies/mL), low (1.6–4 log₁₀ copies/mL), or high (>4 log₁₀ copies/mL). The CD4 count was subdivided into 2 groups: less than and more than 200/μL. Staging in these cases was clinical and based primarily on imaging studies. One resection case had lymph nodes resected. Fine-needle aspiration of lymph nodes confirming metastatic carcinoma was performed on 2 cases. Outcome analysis included disease progression, cancer-specific survival, and overall survival. For the outcome analysis, all cases were reviewed in Epic until March 2020 and patients' survival was followed. Cancer-related deaths included patients who died secondary to metastatic disease and/or complications from treatment (chemotherapy or radiation) such as sepsis. Deaths related to HIV status but not directly to anal carcinoma were included in overall deaths but not in our cancer-related deaths. We use the term disease progression in our study to include persistent disease, regional recurrence, appearance of positive lymph nodes for metastatic disease, and distant metastasis. We used information from all the radiologic imaging reports (computerized tomography, positron emission tomography scan, and/or magnetic resonance imaging) and clinical follow-up visit notes available to determine disease progression. All patients documented with a greater than 60-month (5-year) survival were still alive at the time of the most recent encounter available in our database beyond the 60 months.

Fifty-eight cases were immunostained with PD-L1 mouse monoclonal antibody (22C3, Dako, Carpinteria, California). Four- to 5-micrometer-thick sections of formalin-fixed, paraffin-embedded tissue were tested for the presence of PD-L1 using the Dako PD-L1 IHC 22C3 pharmDx kit. The sections were deparaffinized and rehydrated to deionized water. They were then antigen retrieved in a low pH target retrieval solution using Dako PT link module. All slides were loaded on an automated system (Dako Auto Stainer Link48) and incubated with a mouse monoclonal antibody to PD-L1 or the negative control reagent. This was followed by a mouse linker composed of rabbit anti-mouse antibody and then incubation with a ready-to-use visualization reagent consisting of a goat secondary antibody and horseradish peroxidase molecules coupled to a dextran polymer backbone. The enzymatic conversion of the subsequently added diaminobenzidine chromogen results in precipitation of a visible reaction product at the site of antigen. The color of the chromogenic reaction is modified by a chromogen enhancer. The specimens were then counterstained with hematoxylin and coverslipped. All incubations were performed at room temperature; between incubations, sections were washed with Tris-buffered saline. Results were interpreted using a light microscope. Of these, 51 cases could be evaluated for PD-L1 expression, and 7 cases had to be excluded because of the loss of or insufficient tumor tissue. Of the 51 cases, 49 were biopsies and 2 cases had resection. We evaluated PD-L1 expression using both combined positive score (CPS) and tumor proportion score (TPS). Combined positive score (CPS = [No. of PD-L1 staining cells (tumor cells, mononuclear inflammatory that are within the tumor nests and/or adjacent supporting stroma)/total No. of viable tumor cells] × 100) of 1% or higher was defined as PD-L1 positive (PD-L1⁺) in the Keynote-028 clinical trial¹⁶  and has also been used in a few other studies.^16,18–20  Because other studies evaluating PD-L1 in anal cancer have used TPS, we also evaluated TPS.^21–25  Tumor proportion score is the percentage of viable tumor cells showing partial or complete membrane staining and of any intensity (TPS = [No. of PD-L1⁺ tumor cells/total No. of PD-L1⁺ cells + PD-L1 negative (PD-L1⁻) tumor cells] × 100). Any nuclear staining or granular cytoplasmic staining should not be interpreted as PD-L1 positivity. Mononuclear immune cells with complete membrane staining and tumor cells with partial or complete membrane staining were interpreted as PD-L1⁺. There can be heterogeneity in the staining, and we determined the overall percentage of PD-L1 positivity. Overall, average CPS of 1% or higher and TPS 1% or higher was interpreted as PD-L1⁺. Scoring was first done by one of the authors and subsequently by another author. At the end of the first round of scoring, overall agreement between the 2 observers in assigning PD-L1 as negative or positive was 94%, with 100% agreement on PD-L1⁻ cases. Three cases with borderline overall TPS and CPS scores between 0 and 3 were reviewed together to decide the final score. In other PD-L1⁺ cases the differences in scores were between 5 and 10 and the mean score was recorded. PD-L1 expression was correlated with age, sex, HIV status, HIV viral load, CD4 count, stage of cancer, disease progression, cancer-specific survival, and overall survival.

All data were analyzed using the SPSS statistical software package (version: 26.0; IBM Corporation). The χ² and Fisher exact probability tests were used to analyze the differences between qualitative data. The Mann-Whitney U test was used to compare the medians. Survival curves were plotted by the Kaplan-Meier method and compared using the log-rank test. Univariate and multivariate Cox regression analyses were performed to identify significant prognostic factors. Only variables with a P value of <.5 in the univariate analysis were taken for the multivariate analysis. A 2-tailed P value of less than .05 was considered statistically significant.

Forty-nine of the 51 cases were pretreatment biopsies. Only 2 cases had resections. PD-L1 was positive in 24 of 51 cases (47%) by CPS and in 18 of 51 (35%) by TPS. Staining was heterogeneous in all PD-L1⁺ cases; invasive foci ranged from being completely negative for PD-L1 to foci with 100% positivity. The overall average CPS and TPS ranged from 1% to 50%. Of the 2 resection cases, 1 was PD-L1⁺. There was no difference in staining between the deep and superficial aspects of the tumor in this case. The majority of PD-L1⁺ cases had overall low scores of less than 10 (67% [16 of 24] of CPS and 67% [12 of 18] of TPS). Thirteen percent (3 of 24) of CPS and 17% (3 of 18) of TPS were higher than 25 (Figure 1, A through D).

The distributions of age, sex, and HIV status in the study population by CPS and TPS are given in Table 1. Age range in our patient population was 25 to 68 years. Age ranges for PD-L1⁻ and PD-L1⁺ cases were similar: 25 to 67 years and 29 to 68 years, respectively, with no significant difference in the median age of PD-L1⁻ and PD-L1⁺ groups by CPS and TPS. There were 40 men and 11 women, with a male to female ratio of 3.6:1. The majority of patients were male, and the proportion was not significantly different between the PD-L1⁻ and PD-L1⁺ groups. A majority of all patients in our study were African American (AA) (74%; 38 of 51) and HIV positive (71%; 36 of 51). HIV-positive status was not significantly different between the PD-L1⁻ and PD-L1⁺ groups.

The distribution of stage of disease in the 2 groups is given in Table 1. Distribution of cancer stages 1 through 4 was similar and not significantly different between the PD-L1⁻ and PD-L1⁺ groups (P = .99 by CPS and P = .97 by TPS).

Distribution of HIV viral load and CD4 counts are given in Table 1. HIV viral load for the categories of negative and low combined versus high was not significantly different between the PD-L1⁺ and PD-L1⁻ patients. A CD4 count of higher than 200/mL was significantly higher in PD-L1⁺ patients compared with PD-L1⁻ patients, with a greater difference by TPS compared with CPS (83.3% versus 42.8%, P = .03 by TPS; 75% versus 41.2%, P = .049 by CPS).

Eighty-six percent (44 of 51) of all patients underwent radiation therapy (RT). The majority (80%; 41 of 51) of all patients also received chemotherapy. Clinical outcomes are summarized in Table 2. Only 2 patients were lost to follow-up for unknown reasons, giving us an excellent retention rate of 96%. Fifty percent (25 of 51) of all patients had disease progression, with 1 patient of unknown status because of loss to follow-up. Disease progression was not significantly different between the PD-L1⁻ and PD-L1⁺ groups. By CPS, 47.8% (11 of 23) of PD-L1⁺ patients and 37% (10 of 27) of PD-L1⁻ patients had disease progression, P = .57. By TPS, 42.4% (14 of 33) of PD-L1⁻ patients and 41.2% (7 of 17) of PD-L1⁺ patients had disease progression, P > .99. Progression-free survival in months was also not significantly different between PD-L1⁻ and PD-L1⁺ patients (23 versus 22 months, P = .89 by CPS; 24 versus 18 months, P = .53 by TPS).

Overall, 41.2% (21 of 51) of patients died, with 19 deaths (90% of all deaths) secondary to the cancer diagnosis. Two deaths, 1 in the PD-L1⁻ and 1 in the PD-L1⁺ group, were due to disseminated tuberculosis and cardiac disease, respectively. The median cancer-specific survival was significantly lower in PD-L1⁺ patients compared with PD-L1⁻ patients (24 versus 48 months, P = .02 by CPS; 22 versus 48 months, P = .01 by TPS). The ranges for disease-specific survival for PD-L1⁺ and PD-L1⁻ patients were 3–60 and 6–60 months, respectively. The Kaplan-Meier curve for 5-year overall survival was significantly lower in PD-L1⁺ patients compared with PD-L1⁻ patients (42% versus 72%, P = .03 by CPS; 41% versus 64%, P = .02 by TPS) (Figures 2 and 3).

On univariate analysis, only stage and PD-L1⁺ status by both CPS and TPS were significant prognostic factors for overall survival (Table 3). On multivariate analysis, PD-L1⁺ statuses by both CPS and TPS were independently significant for worse overall survival with hazard ratio = 2.85 (95% CI, 1.06–7.67; P = .04) by CPS and hazard ratio = 3.4 (95% CI, 1.18–9.76; P = .02) by TPS (Table 4).

In summary, factors that were not significantly different between the PD-L1⁺ and PD-L1⁻ groups were age, sex, HIV status, HIV viral load, stage, number of patients with cancer progression, and progression-free survival. The median cancer-specific survival and 5-year overall survival were significantly lower in PD-L1⁺ patients compared with PD-L1⁻ patients. PD-L1 positivity was an independent prognostic factor for worse overall survival.

PD-L1 expression on tumor cells has been found in various malignancies, including head and neck, gastric, colorectal, breast, renal cell, non–small cell lung cancers, and melanomas.^12,26–30  Its expression by tumor cells has been associated with both favorable and unfavorable prognosis. Thus far, only a few studies have investigated PD-L1 expression in ASCC, with controversial results.

In this study, we investigated the correlation of PD-L1 expression in ASCC with clinical outcomes, clinical characteristics, and laboratory data in a unique patient population of predominantly male, HIV-positive AAs. The 5-year mortality rate in our study was about 41% compared with 31% for anal cancer based on SEER data from 2011 through 2017. However, 46% of cases in the SEER data were in stages 3 and 4 combined, whereas 55% of our patients were in stages 3 and 4 combined.¹  Seventy-six percent of our study cohort were AAs receiving care in an urban public hospital. Racial disparity in stage distribution and survival is multifactorial and is reported to be due to socioeconomic factors, insurance status, comorbidities, and tumor characteristics.³¹  Studies by Iseas et al,¹⁹  Wessely et al,²²  Balermpas et al,³²  and Chamseddin et al³³  showed better survival rates from ASCC in PD-L1⁺ patients. Mitra et al²⁴  did not find PD-L1 to be prognostic but found a different biomarker, indoleamine 2,3 dioxygenase 1 (IDO1), to be associated with poor outcome. Zhao et al¹⁸  and Govindarajan et al²¹  were the first to suggest that PD-L1 positivity was associated with worse recurrence and mortality rates in ASCC. Steiniche et al²⁰  also found PD-L1 positivity to be associated with shorter overall survival. In our patient population, the median cancer-specific survival and 5-year survival were significantly lower in PD-L1⁺ patients than in PD-L1⁻ patients. Even though these findings are in concordance with those of Zhao et al¹⁸  and Govindarajan et al,²¹  there are significant differences in our patient population demographics and HIV status compared with the population sample of those 2 studies. Although in both of those studies there was a large proportion of female patients (49% and 85%) with only 15% of AA and no HIV or immunocompromised patients included in the study, our study population was composed mostly of male, AA, and HIV-positive patients (78%, 74%, and 71%, respectively). The other studies that showed improved survival in PD-L1⁺ ASCC also examined a distinctly different patient population with a predominance of female, White, and nonimmunocompromised patients.^19,22,32,33  The PD-L1 positivity rate in several studies was wide, ranging from 22% to 82.5%, with Koncar et al²³  reporting 22% PD-L1 positivity using TPS and Steiniche et al²⁰  reporting 82.5% PD-L1 positivity using CPS. Our PD-L1 positivity is within this range. Although differences in patient population could be a potential explanation for variation in PD-L1 positivity, the method used for establishing PD-L1 positivity is also different in different studies. Some studies used CPS^18,19 ; some used TPS^21,22,25 ; some used greater than or equal to 1%,^19,20,22  greater than 1%,^24,33  or greater than 5% as a cutoff value¹⁸ ; and others used “any positivity” as evidence of PD-L1 expression.²¹  In our study the criterion of an overall average TPS of greater than or equal to 1% of tumor cells with partial or complete membrane staining as PD-L1⁺ resulted in lower PD-L1 positivity than using CPS of greater than or equal to 1%. Also, several different commercially available PD-L1 antibodies are available. Each uses a different staining protocol, threshold, and scoring algorithm (Table 5); therefore, there is a pressing need to standardize PD-L1 scoring in ASCC. The association of HIV status with local antitumor immune response has been poorly understood. Yanik et al³⁴  observed no difference in PD-L1 expression of ASCC between HIV-positive and HIV-negative patients, with 49% of total tumors demonstrating tumor cell PD-L1 expression. Their findings are in concordance with our results in which HIV-positive status was similar in PD-L1⁻ and PD-L1⁺ groups (73% and 71%, P > .99). In addition, our study showed that a CD4 count of higher than 200/μL was associated with higher tumoral expression of PD-L1. Yanik et al³⁴  showed no association of peripheral CD4 count with expression of immune checkpoint molecules, including PD-1, PD-L1, and LAG-3. CD4 count correlated positively with PD-L1 CPS higher than 1% and negatively with TPS higher than 25% in HIV-positive patients with cervical squamous cell carcinoma.³⁵  Such discrepancies may be attributed to a relatively small number of cases tested and different methodology in evaluating PD-L1 and different tumor microenvironments of cervical and anal SCC. Besides, it has been postulated that a decrease in CD4 count may result in decreased PD-L1 expression on tumor cells and antigen-presenting cells, secondary to decreased interleukins (IL-2) and interferon-γ in the microenvironment. In HIV-infected patients, even after prolonged antiretroviral therapy, fewer CD4-positive cells may prevent IL-2 and IFN-γ secretion, resulting in decreased PD-L1 expression in antigen-presenting cells and tumor cells. On the other hand, when CD4 count increases, the immune system is enhanced and produces more cytokines, increasing PD-1/PD-L1 expression on immune cells and tumor cells.³⁵  More studies are needed to better understand the relationship of CD4-positive cells and PD-L1 expression in HIV-positive patients and the effect of antiretroviral therapies.

Furthermore, an immunosuppressive subset of CD4 T cells (Tregs), characterized by the expression of a master regulatory transcription factor, forkhead box P3 protein (FOXP3), correlates with poor prognosis and low survival rates in some cancers and is favorable in others.³⁶  There is accumulating evidence suggesting that an increased number of Tregs is associated with resistance against cancer immunotherapy.³⁷  PD-L1 expression on tumor cells has been found to have a direct association with the presence of FOXP3⁺ Treg density at the tumor site, which could indicate a coordinated immune suppressing action and worse prognosis.³⁶  In our study, we did not investigate the density of tumor-infiltrating T cells and their subsets. Further studies are needed to elucidate FOXP3-positive Tregs' prognostic significance as a tool for overcoming Treg-mediated tumor resistance and maximizing the therapeutic efficacy and immune response in patients with ASCC.

Many immune-suppressive factors such as PD-1 and PD-L1 are stimulated or enhanced by radiation in the absence of immunotherapy. Therefore, it has been demonstrated that the combination of radiotherapy and PD-1/PD-L1 checkpoint blockade synergistically enhances antitumor immune activity and increases the treatment efficacy of either therapy alone.³⁸ 

Scientists have not reached a consensus concerning the effect of RT on Tregs. The numerical and functional abnormality of Tregs may be influenced by radiation dose, the form of radiation, or tumor type.^39,40  A previous study of HIV-positive patients with cervical squamous cell carcinoma showed that PD-L1⁺ tumor cells are more radiosensitive and that those patients show better response with RT.³⁵  High CD8:FOXP3 ratio combined with negative PD-L1 expression was an independent and significant favorable predictive factor for local control. It is reported that it might be a useful biomarker of radiosensitivity in patients with laryngeal SCC receiving definitive RT.³⁹  In our study, most patients received RT, and the subset of PD-L1⁺ patients could benefit from a combination of radiotherapy and anti–PD-L1 treatment. Further studies on CD8/FOXP3 expression on ASCC are needed to explore future therapeutic venues.

Our study involved a patient population that was predominantly male, AA, and HIV positive. The incidence of anal cancer in HIV-infected adults is more than 30-fold greater than in the general population, and our patient cohort reflects this. Although this may be viewed as a limitation, the results show that even in this unique cohort, PD-L1 expression is associated with worse outcomes.⁴¹ 

In summary, we have described PD-L1 expression in a unique population of mostly male, AA, and HIV-positive patients with ASCC. We found that the median cancer-specific survival and 5-year overall survival were significantly lower in PD-L1⁺ patients than PD-L1⁻ patients with ASCC. PD-L1 expression is an independent prognostic factor for worse overall survival and does not correlate with age, sex, tumor stage, HIV status, HIV viral load, and progression-free survival. HIV-positive patients with higher CD4 count were more likely to be PD-L1⁺.

As we have learned the lesson from efforts to standardize tissue preparation, staining, and scoring of HER2, a prognostic and therapeutic marker in breast carcinoma, we should also strive to develop strict standardization and accurate staining and scoring of PD-L1 in ASCC. More studies are needed to explore the combination of RT, PD-L1 expression, and other potential targeted therapies and immune checkpoint blockade as new venues for the treatment of ASCC.