Pan-cancer analysis of RNA expression of ANGIOTENSIN-CONVERTING ENZYME II (ACE2) and cellular proteases reveals high variability and possible impact on COVID-19 clinical outcomes

Patients with cancer demonstrate particularly poor outcomes from COVID-19. To provide information essential for understanding the biologic underpinnings of this association, we analyzed whole-transcriptome RNA expression data obtained from a large cohort of cancer patients to characterize expression of ACE2, TMPRSS2, and other proteases that are involved in viral attachment to and entry into target cells. We nd substantial variability of expression of these factors across tumor types and identify subpopulations expressing ACE2 at very high levels. In some tumor types, especially in gastrointestinal cancers, expression of ACE2 and TMPRSS2 is highly correlated. Furthermore, we found inltration with T-cell and natural killer (NK) cell inltration to be particularly pronounced in ACE2-high tumors. These ndings suggest that subsets of cancer patients exist with heightened susceptibility to SARS-CoV-2 infection in whom malignant tumors function as viral reservoir and possibly promote the frequently detrimental hyper-immune response in patients infected with this virus.


Introduction
The new severe acute respiratory syndrome coronavirus, SARS-CoV-2, has led to a devastating pandemic affecting large segments of the global population with 11,159,301 con rmed cases and at least 527,827 deaths as of July 4, 2020 1 . The disease resulting from SARS-CoV-2 infection, COVID-19, shows particularly severe courses in men and the elderly, a relationship that is further aggravated by comorbidities 2 . Malignant diseases are comorbidities of interest in this context, as cancer patients, whose frequent treatment with immune-suppressive drugs might add to their vulnerability, have demonstrated particularly severe courses of the disease 3,4 . Despite the similarities of SARS-CoV-2 to other coronaviruses [5][6][7] , the biologic underpinnings of increased COVID-19 morbidity remain unclear.
Interaction with ACE2 is mediated by the viral spike protein, which undergoes proteolytic pre-activation, or priming, by cellular proteases TMPRSS2 and cathepsins (CTSB/CTSL) 5 . The SARS-CoV-2 spike protein also contains a novel cleavage site for the proprotein convertase FURIN that is absent in SARS-CoV 6 .
FURIN cleavage of the SARS-CoV-2 can preactivate the spike protein for cell entry, and TMPRSS2 and CTSB/CTSL have cumulative effects with FURIN on SARS-CoV-2 entry 9 . Interestingly, increased expression of ACE2 and FURIN have also been observed in oral epithelial cells 10 .
Non-small cell lung cancer (NSCLC) patients infected by SARS-CoV-2 have a higher risk of severe symptoms and higher mortality rate than patients without 11,12 . As ACE2 expression is related to the severe acute respiratory syndromes induced by SARS-CoV via mediating the production of cytokines and the resultant adaptive immune response 13,14 , we examined transcriptomic markers of immune and stromal cell populations in the tumor microenvironment (TIME) of NSCLC specimens 15 . Understanding how the tumor microenvironment changes in NSCLC as ACE2 expression increases may provide additional insight into the associated immune system deregulation that may underlie the severe comorbidities observed in NSCLC patients and may provide additional insight into an effective treatment for those suffering from COVID-19.
Initial data suggest that some cancers express high levels of ACE2 16,17 . In such cases, malignant masses could function as viral reservoirs, sustaining and amplifying SARS-CoV-2 infection in patients with cancer. Additionally, increased co-expression of cellular proteases with ACE2 may further identify subsets of cancer patients with increased vulnerability to infection 18,19 . Here, we endeavored to systematically characterize ACE2 and key cellular protease gene expression in normal and malignant tissues, taking advantage of a large database of tumors pro led by whole-transcriptome sequencing (WTS). We hypothesized that ACE2 and protease gene expression would vary across tumor types, with high expression presumed to associate with increased risk of infection and severe course of disease. We found that ACE2 expression varies and correlates with expression of proteases in a tumor type-speci c manner, which may underlie the severe disease progression observed in patients with cancer.

Methods
As part of routine comprehensive molecular pro ling of cancer samples at the Caris Life Sciences laboratory, a NextGen RNA sequencing panel was utilized that uses a hybrid bait-capture method to pull down RNA transcripts and sequence them. The Illumina NovaSeq 6500 was used to sequence the whole transcriptome from patients to an average of 30M paired end reads. Raw data was demultiplexed by Illumina Dragen BioIT accelerator, trimmed, counted, PCR-duplicates removed and aligned to human reference genome hg19 by STAR aligner 20 . NGS RNA sequencing captures 22,192 exonic regions. For transcription counting, transcripts per million molecules (TPM) was used (Salmon expression pipeline 21 ). TPM expression data were pulled from a relational database containing retrospective clinical sample information. In addition to the large cohort of tumor samples tested, a small panel of patients where non-tumor tissue was extracted and tested was examined for ACE2 expression. Pathway analysis was performed using KEGG pathway colored mapping on Z-scores of fold change between groups 22 . All analyses were performed on deidenti ed, retrospective cases. As such, this research was covered under IRB Exemption, reviewed and determined by the Western Institutional Review Board (WIRB). Expression data were used to build data distributions, correlate cohorts, and examine gene expression outliers, and perform statistical tests. Statistical analyses were performed using R version 3.6.1. Pearson's product moment correlations were calculated and associated p-values generated via the cor.test function in the stats package.

Patient and tumor characteristics and correlations with ACE2 expression
A total of 38,628 tumors specimens, representing 36 tumors types, that underwent comprehensive molecular pro ling at Caris Life Sciences between 2019 and 2020 were retrospectively reviewed. Molecular pro les of adjacent normal (non-cancerous) tissue from 127 tumor specimens, representing 24 tumor types, were reviewed as a control. Key demographic characteristics of these cohorts are shown in Table. Average
Tumor type-speci c deviations in ACE2/protease co-expression were less prominent for CTSB, CTSL, and FURIN compared to TMPRSS. Notably, ACE2/protease co-expression in GI cancers deviated from the primary cluster due to increased ACE2 expression (CTSB: r=-0.02, m=-0.0001; CTSL: r=-0.04, m=-0.02; FURIN: r=0.03, m=0.02). Together, this suggests the co-expression of ACE2 and protease genes may be more predictive of susceptibility to viral infection by SARS-CoV-2 than ACE2 expression alone, with GI cancers exhibiting the highest potential susceptibility.

ACE2 expression and association with cell population abundance in the TME in NSCLC
In addition to its function as a receptor for SARS-CoV-2, ACE2 is suggested to be involved in various aspects of post-infection processes, and the expression of ACE2 in lung has been shown to associate with innate and acquired immune responses 23

Discussion
The study presented here provides, to our knowledge, the rst systematic assessment of RNA expression of key molecules involved in the infectious process leading to COVID-19 in advanced cancer using data from a large cohort of patients who have undergone routine molecular pro ling of their malignancies. We found extensive variability of expression of the main receptor for SARS-CoV-2, ACE2, in particular in NSCLC and gastrointestinal cancers. Our ndings are of immediate relevance for an improved understanding of the dynamics of the current COVID-19 pandemic since recent studies highlight the particular vulnerability of patients with cancer in terms of increased mortality but also worse severity of the disease 3,4 . While there are manifold potential pathophysiologic underpinnings for this concerning nding, including immune-suppression and smoking history, differences in expression of the primary viral receptor in normal and malignant tissues might contribute. Knowledge of such systematic differences could lead to improved risk prediction for patients with cancer.
Our analysis of normal tissue samples demonstrates signi cantly higher ACE2 levels in males and in patients of age less than 55 compared to females and older patients, respectively. This nding mirrors the clinical observation of increased mortality and morbidity in male patients in Asian and Western populations while not in agreement with the worse outcomes from the infection in older patients. 2, 25, 26 . Thus, it is conceivable that differences in expression levels of ACE2 in normal tissues contribute to differences in risk of infection and disease severity in men but also highlight the possible impact of comorbidities in the elderly. Interestingly, multiple components of the renin-angiotensin system appear to be co-regulated with ACE2 in normal tissues, indicating that there is a potential complex interplay between the infection and pre-existing conditions, which might contribute to the worse outcome for patient with hypertension and other cardiovascular diseases 25 .
Examination of our extensive database of molecular pro les obtained from patients with advanced malignancies revealed wide-ranging variability in expression of ACE2 in samples from advanced cancers. Highest ACE2 RNA expression levels were observed in colorectal, renal, and NSCLC cancers as well as in cholangio-and gastric carcinomas. These ndings are in agreement with previous observations in the TCGA dataset, which consists mostly of data from resected, non-metastatic tumors 17 while our dataset is overwhelmingly derived from patients with advanced cancers. Our analysis of expression levels of several proteases known to function as coreceptors demonstrates variable correlation with ACE2.
Correlation of ACE2 expression with protease expression was particularly pronounced in GI cancers suggesting these patients may have heightened susceptibility for SARS-CoV-2 infection and a more severe course of COVID-19. From a clinical perspective, of particular interest is our nding that subgroups of patients with intensely increased ACE2 expression levels exist in several cancer types, including NSCLC and breast cancers, as well as in glioblastoma and melanoma. While, in general, little is known about the interaction of human viruses with malignant tumors, data from studies of oncolytic viruses demonstrate that viruses can replicate in malignant masses and sustain ongoing viremia if viral receptors are present on tumor cells 27 . Thus, it is conceivable that malignant tumors with particularly high ACE2 and protease expression function as viral reservoirs leading to sustained viremia and increased disease severity. This hypothesis needs to be tested clinically since it would suggest intensi ed isolation measures or even prophylactic antiviral treatment for patients with tumors that express ACE2 at very high level. In addition, our nding of increased presence of various immune-cell lineages, including T cells and NK cells, in tumors with high ACE2 expression highlights the possibility that such tumors are not only susceptible to SARS-CoV-2 infection but could also promote the severe in ammatory response observed in many patients with COVID-19. On the other hand, intratumoral injection of u vaccine can promote conversion "cold" tumors to hot, generates systemic CD8 + T cell-mediated antitumor immunity, and sensitizes resistant tumors to checkpoint blockade 28 . Thus, one could speculate that in some cases SARS-CoV-2 infection might promote tumor regression in the context of immune checkpoint inhibitors.
In summary, our investigation demonstrates signi cant differences in ACE2 and protease expression in normal and malignant tissues with a subgroup of patients expressing very high levels of ACE2 RNA. These ndings, together with the increased presence of in ammatory cells in tumors displaying high ACE2 levels might contribute to the developing complex pathophysiologic picture of COVID-19 and help guide prophylactic measures in patients with solid malignancies. Table   Table 1 Table 1: ACE2 expression associated with patient demographics of normal control and cancer cohorts. Pvalues were calculated using the Kruskal-Wallis rank sum test for numeric variables and Pearson's Chisquared test for categorical variables.
Supplementary Table   A supplementary table was

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.