Clinical characteristics, HIV status, and molecular biomarkers in squamous cell carcinoma of the conjunctiva in Ghana

Abstract Background and aims Conjunctival squamous cell carcinoma (CSCC) varies in incidence geographically from 0 to 1 case per 100 000 per year globally. Additionally, the incidence of CSCC is known to increase 49% for every 10° decrease in latitude. Since the onset of the AIDS epidemic, there has been a trend of increasing incidence of CSCC in Africa, and despite relatively stable levels of ultraviolet (UV) exposure, there is an observed 12 times greater risk of developing CSCC when individuals are infected with HIV. In this study, we aim to analyze the clinical characteristics and biomarkers of CSCC in Ghana. Methods In this study, a registry review of patients from January 2011 to May 2016 with CSCC at Komfo‐Anokye Teaching Hospital in Kumasi, Ghana, was performed (n = 64). Tumor blocks of the CSCC were analyzed for the expression of various biomarkers. Results In this study, the median age of onset of CSCC is 46.5 years old (range of 20–90 y old). Fifty one and a half percent (n = 33) of the cohort is female. There is a low rate of smoking and alcohol use in our CSCC cohort. Thirty‐nine percent (n = 12) of Ghanaian men with CSCC are HIV−, while only 12% (n = 4) of women are HIV−. Fifteen patients had metastasis to lymph nodes or other tissues, and we observed a statistically significant relationship between HIV infection and metastasis (P = 0.027, chi‐squared test). We observed no statistically significant relationship between known prognostic CSCC biomarkers and HIV status, age, or tumor stage. Conclusion Better characterization of CSCC could have a profound impact on the prevention, early identification, and treatment of CSCC in Africa. A retrospective chart analysis and collection of tumor samples can be challenging in this region due to methods of record keeping and stigma attached to clinical data such as HIV testing and smoking and alcohol use. As a result, in this study, data were often incomplete leading to inconclusive results and analysis that should be interpreted with caution. Future studies should consider a prospective study design that gathers clinical data in a standardized format and ensures fresh tissue from CSCC tumors.


| INTRODUCTION
Conjunctival squamous cell carcinoma (CSCC), a subset of ocular surface squamous carcinoma, is a cancer of the eye with an incidence rate of 0 to 1 case per 100 000 per year globally. 1 In the United States, this cancer typically effects elderly men (median age of 77) with high levels of ultraviolet (UV) exposure. 2,3 In fact, Newton et al showed that the incidence of CSCC increased by 49% for every 10°decrease in latitude, supporting the role of UV exposure as a major risk factor. 4 Consistent with this finding, there are higher incidence rates of CSCC in equatorial areas of sub-Saharan Africa than in the rest of the world (0-5.3 cases per 100 000 vs 0-1 case per 100 000, respectively). 5 More recently, studies have also noted an apparent trend of increasing incidence of CSCC in sub-Saharan Africa. For example, in Uganda, there were four cases of CSCC reported between 1960 and 1971 and 66 cases 6,7 reported between 1994 and 1997. Even accounting for better access to health care and increasing reporting of cancer in hospitalbased registries, these differences appear disproportionately larger than increased reporting alone would predict. In light of relatively small increase in levels of UV radiation exposure in the region, such increase in incidence has been putatively associated with the onset of the AIDS epidemic. 5,8 Consistent with the parallel increases in the incidence of both HIV and CSCC, a study also observed twelvefold greater risk of developing CSCC in patients infected with HIV. 9 Although relationships have been drawn between CSCC and HIV/AIDS in countries in East Africa like Kenya and Uganda, 6,7,[10][11][12][13] very few studies on CSCC in West Africa have been published. Additionally, while a relationship between CSCC and HIV/AIDS has been established in sub-Saharan Africa, the pathogenesis and natural history of CSCC cases in HIV+ patients in Western Africa have not yet been described, limiting our understanding of causative risk factors for the disease. One potential causative factor is infection with human papillomavirus (HPV), although the literature has shown disparate findings regarding the role of HPV in the pathogenesis of CSCC in HIV+ patients. In one set of studies done in both Africa and the United States, evidence of HPV infection was found in most CSCC tumor samples. [13][14][15][16][17][18][19][20] In contrast, another set of studies from Africa as well as from other regions of the world such as the United States, Germany, Brazil, and Mexico has found no evidence of HPV in CSCC. 10,[21][22][23][24] In addition, prognostic tumor biomarkers like cyclin D, p16, estimated glomerular filtration rate (EGFR), and p53 have been described in CSCC in the United States, but these are not well studied in CSCC in Africa. [25][26][27] This study seeks to establish the clinical characteristics of HIV+

| Marker detection by immunohistochemistry
Fifty-nine CSCC tumors in paraffin-embedded block or slides from Ghanaian patients seen and/or treated at KATH were analyzed at the University of Michigan, under IRB approval. Twenty seven out of fifty nine (45.8%) of the samples have corresponding clinical information. For immunohistochemistry (IHC), formalin-fixed paraffinembedded (FFPE) blocks were submitted to the University of Michigan Histology Core for cutting and staining for p53, p16, cyclin D, and EGFR using an in-house protocol. 30 Reagents used include citrate buffer, pH 6.0 (Sigma, C9999), 3,  Bethlehem, PA 18020, USA). Protocol 5.2 was followed using the xylene and ethanol to remove paraffin wax from tissue. We were unable to obtain high-quality RNA from these blocks. Tissue digestion was carried out at room temperature overnight followed by supplementation with an additional 10 uL of Proteinase K provided by Qiagen kit and followed by the manufacturer's recommended protocol. DNA samples were first analyzed using a Nanodrop ND-1000 for spectrum and Qubit High-Sensitivity dsDNA (Invitrogen, Cat Q32854) for final concentration.
To determine the presence of HPV and identify the specific viral subtypes in each sample, genomic DNA extracts were assessed in quadruplicate using the multiplex competitive polymerase chain reaction (PCR), followed by amplicon-specific single-base extension as used in the Head and Neck Squamous Cell Carcinoma (HNSCC) Cancer Genome Atlas (TCGA) project. 25 Briefly, amplification of the heterogeneous E6 region of 15 high-risk HPV types (HPV 16,18,31,33,35,39,45,51, 52, 56, 58, 59, 66, 68, and 73), three low-risk HPV types (HPV 6, 11, and 90), and a human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene, as a control, was performed, followed by a shrimp alkaline-phosphatase quenching. Multiplex single-baseextension reactions of type-specific E6 probes create a mass differential for each extension product, which was analyzed on a matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometer, an analysis, which was performed at the University of Michigan DNA Sequencing Core.

| Statistical analysis
The data were analyzed with SPSS software, version 23.0 (SPSS, Inc., Chicago, IL), and a chi-squared analysis was used on categorical data sets. Logistic regression was employed to analyze the relationship of various factors with age, applying a significance level of P < 0.05. A Bonferroni correction was applied for multiple testing of the biomarkers against clinical characteristics (six attributes, significance level of P < 0.0083 required).

| Patient characteristics and cancer risk factors
Sixty-four patients were diagnosed with CSCC from January 2004 to May 2016, but 49/64 (76.6%) of these cases were diagnosed between January 2010 and May 2016. The demographic information is outlined in Table 1. The mean age of the male patients is 47.2 ± 3.04 years old, and the mean age of the female patients is 51.9 ± 51.9 years old. We observed a significant relationship between HIV status (positive, negative, and unknown) and sex (chi-squared test, P = 0.035). Thirty five out

| Clinical characteristics and natural history of conjunctival cancers in Ghana
Patients first presented to the hospital 3 weeks to 5 years after the initial symptom was noted (median: 1 y; interquartile range [IQR]: 0.35-1.5). After diagnosis, the most common symptoms patients complained of are shown in Figure 2. The different types of discharge     Therefore, p16 immunostaining did not seem to correlate with HPV status in our CSCC cohort, as seen in Figure 4. However, these results once again must be interpreted with caution as only a small subset of patients had tumors available for testing and the samples that were successfully analyzed did not always yield results for every biomarker.

| DISCUSSION
This study was the first to assess the demographics, clinical character- Additionally, other studies of CSCC in Africa report a higher rate of HIV infection in patients with CSCC than in the general population,  Immunostaining does not correlate with human papillomavirus (HPV) status in our conjunctival squamous cell-carcinoma (CSCC) cohort. Four independent replicates of immunohistochemical staining with parallel positive and negative control tissue were performed per sample, and representative images from p16 and p53 staining from an HPV positive case and HPV negative case are shown. We expected to observe p53 expression in the HPV negative cases and p53 suppression in HPV positive cases as p53 is known to be suppressed in HPV infection. Indeed, p53 expression is seen in the HPV negative cases, and p53 suppression is seen in the HPV positive case. Additionally, we expected to see p16 expression in the HPV positive case as p16 is a frequent marker of HPV infection in cells. However, p16 is negative in both the HPV positive case and HPV negative case. It is possible that p16 expression is not an accurate marker for HPV infection in these samples Additionally, this analysis is limited by the fact that only 35/64 (54.7%) of patients had information on alcohol use recorded and only 34/64 (53.1%) had smoking data available, leaving the data set open to selection bias. Analysis with more complete patient data is required to draw further conclusions.

The clinical presentation of CSCC in HIV+ and HIV− patients in
Ghana was also examined in this study. Most patients had symptoms for 1 year before presentation, and the most common symptom at presentation was a mass. After diagnosis, the most common symptoms included discharge, eye pain, and headache. This presentation is consistent with CSCC in the United States as well as in Africa. 26

| CONCLUSION
In conclusion, this study highlighted the equal incidence of CSCC in males and females in Ghana as well as the 15.6-fold higher HIV infection rate in the CSCC cohort compared with the general Ghanaian population. We also found a significant relationship between female sex and HIV infection in this cohort. Smoking, a known risk in upper aerodigestive-tract HNSCC in the United States, has no greater incidence in CSCC cohort than in the general Ghanaian population, suggesting that smoking is not a major risk factor in CSCC in Ghana.
However, only 53.1% of patients had smoking data recorded, leaving the data set susceptible to selection bias. Additionally, HIV+ CSCC patients are significantly more likely to have metastasis and a history of constitutional symptoms than the HIV− CSCC patients, suggesting that HIV+ CSCC may be a more aggressive cancer than HIV− CSCC.
Furthermore, we postulated that there would be a relationship between HIV and HPV, p53, cyclin D, EGFR, or p16, but we found no relationship between these tumor biomarkers and sex, stage, or HIV status. However, these data should be interpreted with caution as patient information was often incomplete and only a small subset of patients had tumor samples available for analysis. This results in possible selection bias, and the small sample size raises the possibility that significant relationships may have been overlooked. More conclusive results could be obtained in a prospective study in which patients' HIV status is noted and fresh tissue from their tumor sample is collected for biomarker analysis.

ACKNOWLEDGEMENT
We thank H. Walline and B. Jewel for technical advice.

CONFLICTS OF INTEREST
No potential conflict of interest was reported by the authors. The supporting funds had no involvement in the study design, collection, analysis and interpretation of data, the writing of the report, or the decision to submit for publication.