key: cord-0839104-30x26ip7
authors: Corley, Michael J.; Sugai, Christopher; Schotsaert, Michael; Schwartz, Robert E.; Ndhlovu, Lishomwa C.
title: Comparative in vitro transcriptomic analyses of COVID-19 candidate therapy hydroxychloroquine suggest limited immunomodulatory evidence of SARS-CoV-2 host response genes
date: 2020-04-14
journal: bioRxiv
DOI: 10.1101/2020.04.13.039263
sha: 156ff08ad386e2d9f7699b3fa68104fb8c5b37f5
doc_id: 839104
cord_uid: 30x26ip7

Hydroxychloroquine (HCQ) has emerged as a potential and controversial antiviral candidate therapy for COVID-19. While many clinical trials are underway to test the efficacy of HCQ as a treatment for COVID-19, underlying mechanisms of HCQ in the setting of COVID-19 remain unclear. Hence, we examined differential gene expression signatures of HCQ exposure, in vitro SARS-CoV-2 infection, and host signatures of COVID-19 in blood, bronchoalveolar lavage, and postmortem lung to evaluate whether HCQ transcriptome signatures associate with restoration of SARS-CoV-2-related host transcriptional responses. Here, we show that 24 hours of in vitro treatment of peripheral blood mononuclear cells(PBMC) with HCQ significantly impacted transcription of 16 genes involved in immune regulation and lipid metabolism. Using transcriptome data from in vitro SARS-CoV-2 infected NHBE and A549 cells and PBMC derived from confirmed COVID-19 infected patients, we determined that only 0.24% of the COVID-19 PBMC differentially expressed gene set and 0.39% of the in vitro SARS-CoV-2 cells differentially expressed gene set overlapped with HCQ-related differentially expressed genes. Moreover, we observed that HCQ treatment significantly impacted transcription of 159 genes in human primary monocyte-derived macrophages involved in cholesterol biosynthetic process and chemokine activity. Notably, when we compared the macrophage HCQ-related gene lists with genes transcriptionally altered during SARS-CoV-2 infection and in bronchoalveolar lavage of COVID-19+ patients, the CXCL6 gene was impacted in all three transcriptional signatures revealing evidence in favor of chemokine modulation. HCQ-related transcriptional changes minimally overlapped with host genes altered in postmortem lung biopsies from COVID-19 participants. These results may provide insight into the immunomodulation mechanisms of HCQ treatment in the setting of COVID-19 and suggest HCQ is not a panacea to SARS-CoV-2 infection.

An outbreak of a novel respiratory disease appeared in Wuhan, China in December 2019 [1] capable of human infection and stealth transmission [2, 3] . Rapid scientific investigation identified the causative pathogen as a coronavirus named SARS-CoV-2 with similarity to SARS. Clinical observations have shown that the disease arising from SARS-CoV-2 infection termed COVID-19 is characterized by a myriad of symptoms including cough, fever, lethargy, and difficulty breathing [3] . Currently, there remains no effective therapy for COVID-19; hence, a variety of clinical trials are under way to identify an effective COVID-19 therapy.

Hydroxychloroquine (HCQ) has emerged as a controversial and potential candidate therapy for COVID-19 [4, 5] ; however, the mechanisms of HCQ antiviral action are unclear. Hence, understanding host biological mechanisms that change during HCQ treatment will be key to understanding HCQ in the setting of COVID-19.

While the efficacy of HCQ as a treatment for COVID-19 remains to be determined, we hypothesized that comparative analyses of HCQ transcriptome signatures and SARS-CoV-2-related host transcriptional responses may provide insight into mechanisms of HCQ treatment. To start to test this hypothesis, we investigated transcriptional signatures of HCQ treated human primary PBMC, human monocyte-derived macrophages, and human plasmacytoid dendritic cells in relation to in vitro SARS-CoV-2 infection and COVID-19 participant PBMC, bronchoalveolar lavage, and postmortem lung tissue biopsy transcriptional signatures. These comparative transcriptome findings do not address the question of the efficacy of HCQ in patients with COVID-19, but begin to explore potential mechanisms of HCQ in the setting of SARS-CoV-2 infection.

Due to the immunomodulation effects reported HCQ [6] we surveyed whether in vitro treatment of peripheral blood mononuclear cells from healthy people with hydroxychloroquine (10 µM) impacted immune cell gene transcription. We used transcriptome profiling data from three heathy participants PBMC cells that had been either cultured ex vivo for 24 hours or treated with hydroxychloroquine for 24 hours (Supplemental Table 1 ).

We examined differential gene expression from gene counts using edgeR [7] and corrected for multiple comparisons using the FDR method and observed 16 genes that significantly differed between 24 hour untreated and 24 hour hydroxychloroquine conditions at an adjusted P value of P < 0.05 (Fig.1a) . The most significant difference was observed for the IL1R2 gene linked to immune regulation [8] , which was downregulated with HCQ treatment. Transcription of IL1R2 has been reported to be significantly upregulated in monocytes from individuals with chronic graft-versus host disease [9] suggesting HCQ may exert antiinflammatory effects. Notably, we observed that the chemokine CCL22 gene was significantly down regulated after 24 hours of HCQ treatment as compared to the untreated condition (Fig.1a) . CCL22 codes for a CC cytokine that is involved in chemotactic activity for monocytes, dendritic cells, natural killer cells, and activated T cells. Other genes that were significantly differently expressed comparing the untreated and HCQ treated conditions included SEMA3G, FN1, MSMO1, ARHGEF28, SLC1A2, PHOSPHO1, ENC1, CPNE6, SCD, NDRG2, LDLR, A2M, AQP9, and CYP51A1. Pathway analyses of this gene set using the Enrichr gene enrichment analysis tool [10] showed that the top pathways involved lipid and lipoprotein metabolism, SREBF and miR-33 in cholesterol and lipid homeostasis, lipoprotein metabolism, and steroid biosynthesis (Supplemental Data 1). These in vitro data suggest that 24 hour in vitro treatment with HCQ minimally impacts immune gene transcription and support findings that HCQ has potential immunomodulatory and lipid metabolism effects [11] . Specifically, previous work from in vivo studies has reported that HCQ treatment significantly decreased serum TNF-A and increased IL-10 levels in women following treatment with 400mg/per day orally [12] and inhibited IL-1β production from stimulated human neutrophils [13] . In the context of COVID-19 infection, clinical reports indicate severe patients exhibit a well-known immune cytokine storm that leads to multi-organ failure [14, 15] . Ongoing clinical trials will see if in addition to hypothesized in vivo viral impacts of HCQ whether this drug also averts the likelihood for a cytokine storm.

While there remains a lack of treatments for COVID-19, HCQ has received increased attention due to data from in vitro experiments suggesting antiviral activity [4] . In an effort to examine whether the HCQ-related transcriptional changes in PBMC cells related to in vitro SARS-CoV-2 impacted gene transcription upon infection, we examined the overlap of the 16 differentially expressed HCQ genes with recent transcriptional data from in vitro SARS-CoV-2 infection of NHBE cells and A549 cells (Supplemental Table 1 ). This comparative analysis identified only two overlapping genes (Fig.1b) , NDRG Family Member 2 (NDRG2) and

Low Density Lipoprotein Receptor (LDLR) genes, which were transcriptionally altered in the HCQ treatment of PBMC transcriptional dataset (Fig.1c ,f) and in the SARS-CoV-2 in vitro infection of NHBE cells transcriptional dataset (Fig.1d,g) . Interestingly, the NDRG2 and LDLR genes were not altered upon SARS-CoV-2 in vitro infection of A549 cells (Fig.1e,h) , suggesting different in vitro SARS-CoV-2 infection models vary in transcriptional responses which will be critical for identifying relevant host genes impacted by viral infection. The NDRG2 gene has been identified as a novel factor for the regulation of IL-10 in myeloid cells [16] and a factor in macrophage polarization [17] , suggesting HCQ may impact myeloid cell inflammation that has been hypothesized as a driver of severe COVID-19 [18] . We observed that NDRG2 was significantly upregulated in the SARS-CoV-2 in vitro infection dataset (Fig.1d ) and significantly downregulated in the in vitro hydroxychloroquine treatment dataset (Fig.1c) , suggesting that HCQ may restore host transcription of NDRG2 and exert anti-inflammatory effects. We found that the LDLR gene was significantly upregulated in both the in vitro hydroxychloroquine treatment and SARS-CoV-2 infection datasets ( Fig.1f ,g), suggesting both HCQ treatment and SARS-CoV-2 infection alter lipid metabolism transcription. We acknowledge these observations are associative and further casual studies are needed. Previous work has shown HCQ impacts lipid profiles [19] by potentially altering intracellular pH leading to lipid uptake and increased receptor expression.

Whether HCQ exerts immunomodulator effects by altering transcription of cell lipid metabolic pathways needs further study.

Next, we sought to perform a more direct comparison of transcriptional differences observed by HCQ treatment of primary PBMC cells with transcriptional differences in a dataset of PBMC samples from SARS-CoV-2 infected individuals compared to uninfected controls. We analyzed RNA-seq data of PBMC samples from three male individuals infected with SARS-CoV-2 and three control individuals without SARS-CoV-2 infection obtained from China (Supplemental Table 1 ) [20] . We first sought to confirm the observed immunological changes observed during COVID-19 disease in peripheral blood by others [15, [21] [22] [23] [24] . We utilized the bulk transcriptome PBMC data and estimated the abundances of 22 immune cell types using the CIBERSORT analytical deconvolution tool [25] . Similar to other reports, we observed a drastic expansion of monocyte cells in peripheral blood of COVID-19 participants (75.3%, 46.4%, and 57.4%) compared to controls (19.8%, 5.6%, and 26.8%) using an estimation of member cell types in PBMC based on gene expression data ( Fig. 2a, Supplemental Figure 2) . Also, we saw that CD4 memory resting cells and NK cells resting drastically reduced in all three COVID-19 participants compared to controls (Fig. 2a, Supplemental Figure 2 ). differentially expressed genes related to COVID-19, we identified that only 5 genes overlapped between the gene set of HCQ-related transcriptional changes and COVID-19-related transcriptional changes in PBMC (Fig.   2b ). This accounted for only 0.24% of the host related COVID-19 transcriptional changes in PBMC, suggesting HCQ minimally restores/impacts COVID-19-related transcriptional immune changes. In the HCQ PBMC transcriptional dataset, we observed that the IL1R2, FN1, and A2M were significantly downregulated and LDLR and SCD gene were significantly upregulated (Fig.2c) . In the COVID-19 PBMC transcriptional dataset, we observed that the LDLR gene was significantly upregulated similar to the SARS-CoV-2 in vitro infection dataset for NHBE (Fig.2d) . Additionally, in the COVID-19 PBMC dataset, the IL1R2, FN1, and SCD genes were significantly upregulated compared to a significant downregulation of the A2M gene (Fig.2d) . These results should be taken with caution as studies are needed comparing PBMC cells from COVID-19 participants before and after HCQ treatment. Additionally, whether the directionally of transcriptional change in PBMC cells due to HCQ treatment or related to COVID-19 has any immunomodulatory effect leading to clinical benefits requires further study. This comparative analysis is limited and does not address whether HCQ has antiviral effects by reducing SARS-CoV-2 virus entry through impacting cell lipid metabolism pathways since PBMC cells are not a main target for SARS-CoV-2 infection and viral replication. We preliminarily confirmed that PBMC cells do not harbor SARS-CoV-2 virus by mapping RNA-Seq reads from COVID-19 PBMC samples and identifying no significant reads aligning to the SARS-CoV-2 reference (Supplemental Figure 3b) . This was also observed by Xiong et al. [20] . In contrast, we observed viral mapping of reads in the Blanco-Melo et al. in vitro SARS-CoV-2 infection dataset of A549 and NHBE cells (Supplemental Figure 3c,d) . Notably, the in vitro SARS-CoV-2 infection dataset contains a 3' bias of read alignment compared to the COVID-19 PBMC dataset. Recent in vitro work suggest SARS-CoV-2 infects T lymphocytes [26] . Whether this is observable in COVID-19 patient remains to be determined. We assume HCQ treatment of A549 or NHBE cells during SARS-CoV-2 infection would dramatically reduce the viral mapping detected by RNA-Seq. Our observations of overlapping transcriptional changes provides candidate genes to further study for addressing host immune-related mechanisms of action of HCQ in the setting of COVID-19.

Previous research in SARS suggested that SARS-CoV regulates immune function-related gene expression in myeloid cells [27] and a report in a SARS-CoV-infected mouse model showed that dysregulated inflammatory and interferon responses of monocyte/macrophage cells caused lethal pneumonia [28] . Clinical data about COVID-19 suggest that peripheral innate immune monocyte cells and lung tissues macrophage cells are critically involved in host response and severe disease course [29] . Together, these findings suggest that modulation of myeloid cells may have an impact on SARS-CoV-2-related host responses. Interestingly, research on HCQ suggest that it may impact TLR signaling, antigen presentation, and alter macrophage function [30] . HCQ is also proposed to indirectly have anti-inflammatory effects upon the immune system.

Hence, we examined transcriptional differences in human primary CD14+ monocyte MCSF-differentiated macrophage conditions treated with HCQ for 24 hours (Supplemental Table 1 ). Differential gene expression analyses showed that 159 genes were differentially expressed comparing the untreated and HCQ treated conditions corrected for multiple comparisons using the FDR method at P < 0.05 (Supplemental Data 1).

Among these 159 genes, gene enrichment analyses showed the top GO biological process was cholesterol biosynthetic process, GO molecular function chemokine activity, and GO cellular component integral component of plasma membrane (Fig.3a-c, Supplemental Data 1) . Specifically, HCQ altered expression of LDLR, CXCL6, CCL18, CCL5, and CXCL3 genes (Fig.3d-h) . To extend our findings of HCQ impacting immune chemokine activity in innate immune cells, we examined transcriptomes of plasmacytoid dendritic cells (pDC) from four healthy individuals that had been either stimulated with RNA-IC for 6 hours only or stimulated with RNA-IC for 6 hours in the presence of HCQ (Supplementary Table 1) . We observed that CXCL10, CXCL11, CCL4, and CXCL9 genes were significantly downregulated in stimulated HCQ treatment compared to the only stimulated condition (Supplemental Figure 4) . Together, our findings in macrophages and pDC cells suggest HCQ impacts certain cell type specific chemokine signaling pathways and likely modulates immune inflammation.

In our comparative analyses for the macrophage HCQ dataset, we first examined whether the HCQ- Table 1 ). This comparative analysis identified 15 overlapping genes including LDLR, CXCL6, GPNMB, HMGCS1, IDI1, C1S, FDPS, MMP1, PLAU, MAFF, ITGB3, BHLHE40, TNC, CXCL3, SLC7A8 (Fig.4a) . Next, we sought to compare the HCQ-related transcriptional changes in human primary macrophage cells with transcriptional data from COVID-19 participants and in a similar cell type. Hence, we analyzed transcriptome data from bronchoalveolar lavage samples obtained from two individuals with COVID-19 and three uninfected individuals (Supplemental Table 1 ). Bronchoalveolar lavage (BAL) samples predominantly contain macrophages with only a minor population of neutrophils, lymphocytes, and eosinophils. We used CIBERSORT to confirm that the majority population in the transcriptome dataset was macrophage cell types (Supplemental Figure 5) . Additionally, we confirmed that the BAL transcriptome dataset from COVID-19 participants contained high levels of SARS-CoV-2 virus by mapping reads to the viral reference (Supplemental Figure 3a) . The bronchoalveolar lavage COVID-19 transcriptome dataset contained 5,304 differentially expressed genes comparing COVID-19 to uninfected individuals. Our comparative analysis of the macrophage HCQ-treated gene set with the COVID-19 lavage transcriptional dataset identified 45 overlapping genes accounting for only 0.84% of differentially expressed genes related to COVID-19 (Fig.4a) . Seven of the 45 genes overlapped with the in vitro SARS-CoV-2 infection transcriptional dataset and included CXCL6, IDI1, MMP1, PLAU, MAFF, BHLHE40, and SLC7A8 genes (Fig.4a-d) . Our data suggest that studies examining the regulation of these genes and relationship to SARS-CoV-2 infection and host immune responses are warranted.

Immune regulation and defenses against viral pathogens occur within in the lung. We sought to examine whether the differential gene expression signatures of HCQ in immune cells associated with a COVID-19related transcriptional profile in the lung. We analyzed a limited transcriptomic data from postmortem lung biopsies obtained from two male participants with COVID-19 over the age of 60 years and two uninfected agematched control male participants (Supplemental Table 1 lung compared to uninfected lung (Supplemental Figure 7) . This deconvolution analysis showed some interesting initial differences of interest for further study. First, the relative fraction of CD4 memory resting T cells was 15% and 19% in the uninfected lungs compared to only 3% and 4% for the COVID-19 lung. Second, the relative fraction of neutrophils was 23% and 13% in COVID-19 lung compared to 0% and 8% for the uninfected lung. These results support the notion of continual neutrophil influx into the lung in the setting of COVID-19 during sustained and persistent inflammatory cytokine and chemokine secretion [31] . Lastly, in support of the possibility of mast cell activation in severe cases of SARS-CoV-2 infection, we observed that the relative fraction of mast cells activated was 59% and 45% in COVID-19 lung compared to 37% and 0% for uninfected. While these results are limited by the low sample size, they do provide evidence for certain areas of future investigation. We suspect that transcriptome data from HCQ-treated COVID-19 individual's lung is needed to conclude that HCQ does not dramatically restore SARS-CoV-2 induced lung transcriptome alterations.

In this study, we observed that in vitro HCQ treatment of immune cells alters gene transcription which is likely cell type-specific and context dependent. Notably, these alterations were more apparent in macrophage cells compared with bulk PBMC cells treated with HCQ. Our study is a start to address mechanisms of action for whether HCQ is an effective therapy for COVID-19. While these findings are of interest to studies of HCQ and COVID-19, transcriptional studies need to be conducted on in vivo treatment of COVID-19 participants before and after HCQ treatment. The mechanisms of action of HCQ in the setting of viral infection remains an area still under investigation and a research area containing contrasting findings. This is supported by previous HCQ research in the setting of HIV that reported HCQ treatment drastically reduced immune activation in HIVinfected individuals [32] . In contrast to these findings, a randomized double blind placebo controlled trial of 400 mg of HCQ failed to find any difference in T cell activation and reported increased viral replication [33] . While in vitro data of SARS-CoV-2 provide compelling rationale for suggesting clinical efficacy of HCQ [4, 34] , in vivo data and randomized double blinded placebo COVID-19 studies of HCQ are needed to account of contextdependent effects of the host being administered HCQ. Further work will need to determine whether the mechanism of action of HCQ is directly impacting SARS-CoV-2 or indirectly impacting the immune system and hence indiscriminately improving COVID-19 clinical outcomes. Unfortunately, at this moment HCQ is not a panacea for COVID-19 and alternative therapies are urgently needed.

MJC and LCN equally contributed to the concept and writing of the manuscript. CS contributed to RNA-Seq analyses and SARS-CoV-2 viral detection in bulk RNA-Seq data. MS edited manuscript and contributed to concept. RES contributed postmortem RNA-Seq data for analysis and edited manuscript.

Public Data was obtained from de-identified human participants for this IRB exempt study.

Not applicable.

L.C.N. has consulted for AbbVie and on the scientific advisory board of ViiV. R.E.S. is on the scientific advisory board of Miromatrix Inc. The authors declare no competing interests.

This work was supported by NIH NHLBI (K01HL140271 to MJC).

An Overview of Their Replication and Pathogenesis

A pneumonia outbreak associated with a new coronavirus of probable bat origin

Clinical Characteristics of Coronavirus Disease 2019 in China

Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro

No Evidence of Rapid Antiviral Clearance or Clinical Benefit With the Combination of Hydroxychloroquine and Azithromycin in Patients With Severe COVID-19 Infection

Hydroxychloroquine is associated with impaired interferon-alpha and tumor necrosis factor-alpha production by plasmacytoid dendritic cells in systemic lupus erythematosus

Differential Expression Analysis of Complex RNA-seq Experiments Using edgeR

IL-1 receptor 2 (IL-1R2) and its role in immune regulation

Upregulation of IFN-Inducible and Damage-Response Pathways in Chronic Graft-versus-Host Disease

Enrichr: a comprehensive gene set enrichment analysis web server 2016 update

Mechanisms of action of hydroxychloroquine and chloroquine: implications for rheumatology

Immunomodulatory effects of hydroxychloroquine on Th1/Th2 balance in women with repeated implantation failure

Hydroxychloroquine inhibits IL-1β production from amyloid-stimulated human neutrophils

COVID-19: consider cytokine storm syndromes and immunosuppression

Longitudinal characteristics of lymphocyte responses and cytokine profiles in the peripheral blood of SARS-CoV-2 infected patients

NDRG2 is one of novel intrinsic factors for regulation of IL-10 production in human myeloid cell

Loss of NDRG2 in liver microenvironment inhibits cancer liver metastasis by regulating tumor associate macrophages polarization

The Role of Cytokines including Interleukin-6 in COVID-19 induced Pneumonia and Macrophage Activation Syndrome-Like Disease

Associations of hydroxychloroquine use with lipid profiles in rheumatoid arthritis: pharmacologic implications

Transcriptomic characteristics of bronchoalveolar lavage fluid and peripheral blood mononuclear cells in COVID-19 patients

Clinical and immunologic features in severe and moderate forms of Coronavirus Disease

Study of the lymphocyte change between COVID-19 and non-COVID-19 pneumonia cases suggesting other factors besides uncontrolled inflammation contributed to multi-organ injury

Restoration of leukomonocyte counts is associated with viral clearance in COVID-19 hospitalized patients

Immune Cell Profiling of COVID-19 Patients in the Recovery Stage by Single-Cell Sequencing. medRxiv

Robust enumeration of cell subsets from tissue expression profiles

SARS-CoV-2 infects T lymphocytes through its spike protein-mediated membrane fusion

SARS-CoV regulates immune function-related gene expression in human monocytic cells

Dysregulated Type I Interferon and Inflammatory Monocyte-Macrophage Responses Cause Lethal Pneumonia in SARS-CoV-Infected Mice

The landscape of lung bronchoalveolar immune cells in COVID-19 revealed by single-cell RNA sequencing. medRxiv

Antimalarial Drugs as Immune Modulators: New Mechanisms for Old Drugs

Contribution of neutrophils to acute lung injury

Hydroxychloroquine drastically reduces immune activation in HIV-infected, antiretroviral therapy-treated immunologic nonresponders

Effects of hydroxychloroquine on immune activation and disease progression among HIV-infected patients not receiving antiretroviral therapy: a randomized controlled trial

Structural and molecular modeling studies reveal a new mechanism of action of chloroquine and hydroxychloroquine against SARS-CoV-2 infection

STAR: ultrafast universal RNA-seq aligner

We would like to acknowledge the authors that deposited raw data for analysis.