PD-L1 expression in squamous cervical carcinomas of Mozambican women living with or without HIV

Programmed death-ligand 1 (PD-L1) is overexpressed in squamous cervical cancer (SCC) and can be used for targeted immunotherapy. The highest mortality rates of SCC are reported in sub-Saharan Africa, where Human immunodeficiency virus (HIV) prevalence is high. In Mozambique most SCC patients present at advanced stages. Thus, there is a need to introduce new treatment options. However, immunocompromised patients were frequently excluded in previous clinical trials. Our aim was to determine if PD-L1 expression in SCC is as prevalent among women living with HIV (WLWH) as among other patients. 575 SCC from Maputo Central Hospital were included. HIV status was available in 266 (46%) cases PD-L1 expression was scored through tumour proportion score (TPS) and combined positive score (CPS). PD-L1 was positive in 20.1% of the cases (n = 110), TPS (score ≥ 25%) and in 26.3% (n = 144), CPS (score ≥ 1). Stratifying according to the HIV status, WLWH were TPS positive in 16.7%, compared to 20.9%, p = 0.43, and concerning CPS 21.1% versus 28.7%, p = 0.19, respectively. PD-L1 status was not influenced by stage, Ki-67 or p16, CD8 expression influenced only CPS status. Our data indicates that the documented effect of PD-L1 therapy on SCC should be confirmed in randomized clinical trials in an HIV endemic milieu.

For ICI immunotherapy with some specific drugs as Pembrolizumab and Cemiplimab, the diagnostic companion of PD-L1 positive immunohistochemistry is a must, while for other drugs the PD-L1 assay is optional 21,[37][38][39] , although it is classified as an imperfect biomarker for response prediction 40 , as shown that some patients with high expression of PD-L1 had no response to the ICI´s therapy and some cancers with low expression of PD-L1 had a long lasting response 40,41 .
The PD-L1 scoring cut-offs are not uniform for all types of available clones, leading to variability in the interpretation 21,42,43 , interobserver variability and reliability within the same clone 44 .Furthermore, there are controversy regarding the interchangeability of PD-L1 clones type approved as companion for some specific drugs in specific types of cancer, although some studies demonstrated a strong correlation of immunohistochemical expression between some clones such as 22C3 and SP263 45,46 .
In this study we aimed to determine PD-L1 expression in cervical SCC of Mozambican patients with specific emphasis on the influence of HIV status on PD-L1 expression.This would set the fundamentals for a possibility to perform a randomized clinical trial concerning the efficacy of PD-L1 therapy in women living with HIV (WLWH) with advanced CC.

PD-L1
Twenty-eight (28) cases were excluded from the PD-L1 analysis due to the loss of tissue cores in the TMA construction and/or immunohistochemistry procedures, which resulted in 547 cases evaluable for PD-L1.The
Regarding HIV status, there were no differences of P16 and CD8 expression between the HIV positive and HIV negative groups (Table 2).
In order to evaluate interactions of the biomarkers and clinical data on PD-L1 expression a multiple regression analysis for each of the PD-L1 CPS/TPS variants was performed.Concerning CPS > 1%, an influence of CD8 expression was found (p = 0.041), concerning TPS ≥ 25% and ≥ 50% no statistically significant influence was found, data not shown.
In multiple regression analyses, Ki-67 proliferation of > 30% did not correlate to any other biomarker expression, nor stage or HIV status.

Discussion
In the present study we conducted a retrospective analysis of PD-L1 SP263 immunohistochemistry expression in a Mozambican cohort of invasive CC, one of the largest cohorts of SCC from an HIV endemic region.We evaluated the PD-L1 expression based on TPS and CPS for the overall cohort, and further we stratified according www.nature.com/scientificreports/ to the HIV status of the patients.Additional we also assessed the immunohistochemistry expression of p16, CD8 and Ki-67.
It has been suggested from other studies that HIV infection enhances the carcinogenic risk of other viruses through many underlying mechanisms such as an ineffective immune surveillance and facilitates the proliferation of the oncogenic viruses 36 .The chronic inflammatory state of HIV infection results in upregulation of PD-L1 expression, that leads to CD8 + T cells exhaustion, promoting the evasion of HPV infected cells and allowing them to grow and differentiate in an uncontrolled manner, and to develop dysplasia and cancer 47 .The relationship of HIV-HPV infection in carcinogenesis is well established in the cervix, oropharynx, and anal SCC 2,36,48 .
According to our results, the expression of therapeutical target (PD-L1) is not affected by HIV status of the patients, which implies the potential use of PD-L1 targeted therapy both in women living with or without HIV.The same is valid for the presence of CD8 + T lymphocyte population.
PD-L1 expression found in the present study in CPS (26.3%), was less than reported in other studies in CC, ranging from 32.4 to 80% in cervical SCC [49][50][51][52][53][54][55] .Feng et al. that used the TMA approach to perform the PD-L1 immunohistochemistry reported a 32.4% of PD-L1 positive expression 49 .The slightly lower expression of PD-L1 in our study could be related with the heterogenic pattern of expression that have been widely demonstrated in studies using whole slide (WS) preparations, and even though we have sampled three random TMA cores, we could still have missed the spots that were positive in some cases.This indicates that PD-L1 expression still may be systematically underestimated in TMA studies highlighting the fact that TMA is a valuable technique for studies of large cohort but should not be used for the assessment of individual cases.
When comparing the PD-L1 expression in the cases with known HIV status, we found a numerically lower percentage of patients expressing PD-L1 as determined by CPS ≥ 1 (21.1%) and TPS ≥ 25% (16.6%) in the WLWH group compared with the non-HIV women group that expressed little more PD-L1 on CPS ≥ 1 (28.7%) and TPS (20.9%), with non-statistically significant differences of expression between the groups.Previously, differences of PD-L1 expression between WLWH and non-HIV patients have been reported 56 , and PD-L1 expression on the WLWH group was demonstrated on CPS ≥ 1 as 78.6% and TPS ≥ 25% as 31%, thus compared to our results with divergent data concerning especially the CPS, in spite of using identical immunohistochemical assay.A recent study, performed in WLWH and non-HIV patients with in-situ and invasive carcinomas, reported lower (16.7%)PD-L1 expression in both groups using SP263 TPS ≥ 50% and 22C3 CPS ≥ 1 57 .Brito et al. found similar results of PD-L1 expression in cervical SCC (around 16% in both WLWH and non-HIV patients) using a threshold of TPS ≥ 50% and CPS > 1) 57 .Our results were even lower when scored in TPS > 50% showing 14% and 12.3% in non-HIV and WLWH groups respectively, and no statistically significant differences were observed.www.nature.com/scientificreports/ A number of different factors can explain differences between the studies like methodological differences, characteristics of material used, sampling characteristics, anti retroviral therapy (ART) and HIV burden.It has been hypothesized that in WLWH patients under ART, could express less PD-L1 in the tumour cells as result of the therapy 58 , even though in general it has been stated that WLWH patients are more prone to over-express PD-L1 in the tumour cell as a result of the chronic inflammatory status driven by HIV infection itself, as well as genomic alterations induced by HPV/HIV synergetic effect 2,59 .
In the perspective of immunotherapy, has been suggested from other studies, that WLWH patients with invasive CC could benefit from it, as the side effects are similar to non-HIV patients [32][33][34] .Furthermore, it was demonstrated in a melanoma study that CD8 + T cells infiltrate density was more predictive of ICI immunotherapy response 60 .The results of our study, showing that in 88.4% of the cases were CD8 + T lymphocytes infiltrate scored 3 (more than 15 cells) and 4 (uncountable positive cells), mainly intra-tumoral infiltrate, raises the possibility that the majority of the patients of the cohort could have had a good response to the treatment with ICI immunotherapy 40,60 , meaning that in the setting of this study, the ICI therapy would be an alternative to improve the survival rates of CC patients.
In the present study we included a large number of CC of both WLWH and non-HIV patients from a country with high prevalence of both HIV, HPV and CC.Since in this study we have used already existing databases as source demographic and clinical data gathering, this implied a series of limitations to this study such as incomplete clinical data, absence of HIV testing status for the whole cohort and absence of quantitative data on the WLWH regarding the viral load, CD4 + cell count, HIV subtype and data on ART (how long in ART and type of drugs used), and absence of FIGO stage for the majority of the patients.Furthermore, even though some methodological challenges concerning immunohistochemistry 45,61 has been handled, further possible methodological challenges of specimen handling for histopathology 62 may remain in the present low resource setting.
Even though with the outlined limitations, our data indicates that the documented positive effect of PD-L1 on advanced stage of carcinomas 63 could be valid also for WLWH patients with SCC in Mozambique.Thus, there is an urgent need for randomized clinical trials, including optimal biomarker analyses, to demonstrate the feasibility and clinical effect of PD-L1 inhibitors in the HIV endemic milieu of sub-Saharan Africa.

Study population, sampling, and samples
Tumour samples as well as demographic, clinical and HIV serology data were obtained from the computer records files of the Pathology Department and from the Hospital-based Cancer Registry databases of the Maputo Central Hospital (MCH), a quaternary and national referral hospital of Mozambique.The HIV serology testing information used in the present study were available in these records.Nationally in Mozambique the HIV testing is performed according to the World Health Organization (WHO) guidelines for high prevalence countries 64 , using rapid diagnostic test (RDT), where the negative result is assigned by one RDT and positive result is followed by second and third different RDTs 65 .The viral load of the patients was not available for any of the cases, as the follow up and treatment for HIV is performed in the primary and secondary level health care facilities.
The Department of Pathology of MCH is responsible for the anatomopathological diagnosis of all specimens coming from the three provinces of the southern region of Mozambique (Inhambane, Gaza and Maputo), and is also a referral for consultation for diagnosis of cases from the whole country.All cases consecutively diagnosed as CC obtained through biopsy or surgical resection, from January 2017 to December 2018 in the Department of Pathology at MCH were retrieved.
All Haematoxylin and Eosin (H&E) stained slides obtained from formalin fixed paraffin embedded (FFPE) tissue blocks were scanned at 40 × brightfield on a Panoramic 250 automatic digital scanner (3D HISTECH Ltd., Budapest, Hungary) at the Department of Clinical Research of Orebro University Hospital (OUH), Sweden, and stored within a storage system from the same supplier at a server facility, fulfilling the general data protection regulation (GDPR).The scanned material and tissue microarray (TMA) slides were assessed and used for evaluation through the Panoramic Case Viewer software version 2.4.0.119028 for Microsoft Windows.
All cases of invasive CC consecutively diagnosed in the study period, and with sufficient tumour material in FFPE block for extraction of at least three core cylinders of 0.6mm for TMA construction and with no depletion risk of the donor block sample were included to the study.Cases with no available tissue, inadequate or insufficient tumour material in FFPE block for TMA construction, non-invasive carcinoma, wrong coded or double entered were excluded.Out of 615 cases available after review of the blocks, 575 were squamous cell carcinomas, which were our final cohort for this study (Fig. 2).
For TMA construction, the H&E digitalized slides were assessed through the above-described software.Selection of tumour area was performed on those slides by a pathologist (L.L.).

TMA construction
The areas of interest (AOI) were identified, and the TMA circles markers of 0.6 mm diameter were placed.For each slide, three TMA circle markers were placed.The tissue cylinders punches were performed using the TMA Grand Master 2.8 (3D HISTECH Ltd., Budapest, Hungary), an automated tissue microarrayer.The tissue cylinder cores punched from the donor block were automatically transferred into a new paraffin block called recipient block, according to the manufacturer specifications.
All TMA blocks were sectioned in 4µm thickness and transferred to the glass slide.At least one section per paraffin block was stained with H&E to confirm the histology and to certify the availability of tumour tissue in the slides by a pathologist.Other slides were submitted to immunohistochemistry for PD-L1, CD8, P16 and Ki-67.The sectioning for immunohistochemistry were performed the day before the immunostaining 45,61  www.nature.com/scientificreports/

Immunohistochemistry performing and scoring
PD-L1 SP263 and CINtec p16 (E6H4) were performed on Benchmark Ultra (Roche diagnostics, Switzerland) according to the manufacturer's instructions.CD8 (C8/144B) and Ki67 (MIB-1), was performed on DAKO OMNIS (Agilent, Santa Clara, USA), with a polymer-based detection system, according to the manufacturer's instructions (Agilent, Santa Clara, USA).Positive and negative control tissues were included for all antibodies used in each run.Control tissues were included in each run, tonsil, negative lung SCC and positive lung SCC for PD-L1, tonsil and appendix for CD8 and tonsil for p16 as recommended by the manufacturer.Additional, for PD-L1 SP263, an isotype specific negative control was included.The cases were considered evaluable if at least one TMA core was present.
The PD-L1 expression was scored according to Ventana algorithm for PD-L1 SP263 66 .The tumour proportion score (TPS) is calculated by counting PD-L1 stained tumour cells divided by the total number of tumour cells, PD-L1 was considered positive when tumour cells had partial or complete circumferential brown staining on the cellular membrane at any intensity.The TPS was further categorized into two groups: high/positive when ≥ 25% of tumour cells exhibit membrane staining; and low/negative when < 25% are stained 56,66,67 , a ≥ 50% cutoff was also used 68 and the combined positive score (CPS) was also assessed using the cutoff of ≥ 1.The CPS was defined by sum of the tumour cells and mononuclear immune cells stained with PD-L1 divided by the total number of tumour cells, then multiplied by 100 39 .
P16, was scored as positive when both cytoplasmatic and nuclear brown staining was present.According to the extent of positivity, p16 was scored on a four-grade scale as: strong diffuse (3 +), moderate intensity (2 +) and weak sporadic or mild intensity (1 +) or no visible staining (0).Depending on the percentage of positive cells, staining was graded in a semi-quantitative manner, as: Score 0: negative cells; Score 1 -less than 10% of positive cells; Score 2-10 to 50%); Score 3-> 50% 69 .
Ki67 was scored as positive when brown staining was present in the nucleus at any intensity.The positive staining was further categorized in three-grade scale in a semi-quantitative manner as: Score 1: 10-30% stained cells (Low proliferation); Score2: 30-50% stained cells (moderate proliferation); Score3: > 50% (high proliferation) 70 .
For all challenging cases, scoring was discussed, and consensus was reached.

Statistical analysis
Descriptive statistics and assessment of differences between groups, the later using Fishers exact test as well as Multivariate Logistic Regression Analyses with statistical significance considered to be significant with p < 0.05.SPSS (IBM SPSS v 28.0.0.0).

Ethical considerations
All the experimental protocols were approved by the National Bioethics Committee for Health (CNBS) board of Mozambique under registration number 114/CNBS/2019 and Swedish Ethical Review Authority under registration number 2023-01674-01-423571, according to the Helsinki Declaration in human subject research.In Mozambique, the Ethical board (joint committee of the Faculty of Medicine, Eduardo Mondlane University and Maputo Central Hospital) at the university reinforced the project, but since the projects included the use of human biological material an approval from the governmental authority, National bioethical committee, was requested.In Sweden, ethical approvals are granted by a governmental authority which is separate and

Figure 2 .
Figure 2. Flowchart of exclusion process within the overall cohort of carcinoma cases.

Table 1 .
Overall patient clinical data.

Table 2 .
PD-L1 expression according to HIV status.