SARS-CoV-2 lineage-dependent temporal phylogenetic distribution and viral load in immunocompromised and immunocompetent individuals

Abstract

Objectives

Mutational dynamics of SARS-CoV-2 in immunocompromised hosts, although well documented, remain a relatively unexplored mechanism. This study aims to compare the viral replication load and genetic diversity of SARS-CoV-2 in immunocompromised patients and non-immunocompromised individuals (NICs) from two major hospitals in Paris from January 2021 to May 2023.

Methods

Cycle threshold (CT) values were measured by TaqPath COVID-19 RT-PCR (Thermo Fisher Scientific). The SARS-CoV-2 whole-genomes from 683 immunocompromised patients and 296 NICs was sequenced using Oxford Nanopore Technologies and used to determine lineage and mutational profile.

Results

All immunocompromised patients, but not oncology patients, had lower SARS-CoV-2 viral loads than NICs. The genetic distribution of SARS-CoV-2 was homogeneous between immunocompromised individuals and NICs, with more mutations in immunocompromised patients (IRR = 1,013). Indeed, extensive genomic analysis revealed several mutations specifically associated with immunosuppression status, such as S: T95I, S:N764K, M:Q19E and ORF10:L37F. Conversely, the S: R346K and NSP13:T127N mutations were more common in NICs.

Conclusion

Immunocompromised patients have lower viral loads, probably due to their later diagnosis compared to NICs and oncology patients, who have better access to on-site SARS-CoV-2 testing and follow-up. In addition, mutational profiles differ between the two groups, with immunocompromised hosts accumulating more mutations compared to NICs.

Introduction

Over the course of the pandemic, the general population has developed immunity and protection against severe symptoms through vaccination and previous SARS-CoV-2 infections. Immunocompromised individuals remain at an increased risk of experiencing more severe clinical outcomes and higher mortality rates. Several case reports have suggested that immunocompromised patients may shed SARS-CoV-2 for prolonged periods, leading to the accumulation of mutations in the absence of targeted COVID-19 treatment [1, 2]. These mutations can potentially confer resistance to both naturally acquired and vaccine-induced immunity, as well as to monoclonal antibodies, which could eventually lead to the emergence of new variants [3].

Despite growing interest in the role of immunocompromised individuals in viral evolution, a clear understanding of how their infection differ from those of non-immunocompromised individuals is lacking. In particular, it remains uncertain whether prolonged shedding is associated with increased genetic diversity of the virus within the host and whether viral RNA loads differ significantly between these patient populations. In addition, it is not known whether viruses that infect immunocompromised individuals have similar characteristics to those that infect non-immunocompromised individuals. Addressing these gaps is essential to assess the potential risks related in immunocompromised patients and to inform public health strategies.

This study aims to compare the genetic diversity and RNA loads of SARS-CoV-2 between immunocompromised and non-immunocompromised individuals. In particular, we hypothesize that immunocompromised individuals may have higher intra-host genetic diversity and that their RNA load may differ from those of non-immunocompromised individuals from the onset of the infection. By investigating these factors, we aim to provide insights into the SARS-CoV-2 infection in immunocompromised hosts and their potential future role in viral persistence and adaptation.

Methods

Patients

Our retrospective study is based on the Assistance Publique-Hôpitaux de Paris (AP-HP) ‘SARS-CoV-2 infection of immunosuppressed patients’ (EMERGEN SIID) study (Pitié-Salpêtrière and Bichat-Claude Bernard University Hospitals, France). The design of the work has been approved by the Research Ethics Committee for Infectious and Tropical Diseases (CERMIT; decision number: 2022-05-04). Based on standards currently applied in France individual patient information is not required for internal research.

Immunocompromised patients were selected on the basis of the immunosuppression criteria defined by the French High Council of Public Health (HCSP) on 31 March 2020. The immunocompetent control group consisted mainly of health care workers and patients who did not meet these criteria. The data collection will be carried out continuously from December 2020 to May 2023.

Viral RNA load determination

Viral RNA (80 µL) was extracted from 300 µL of nasopharyngeal samples collected at the time of diagnosis (day 0) using Nuclisens^® Easymag^® (Biomérieux). Viral RNA load was determined using the TaqPath™ COVID-19 RT-PCR kit (Thermo Fisher). Thermal cycling was performed in a LC480 instrument (Roche) with one cycle of reverse transcription at 53 °C for 10 min followed by a cycle of PCR activation at 95 °C for 2 min and finally 40 amplification cycles each consisting of 95 °C–3 s and 60 °C–30 s. Primers and probes targeting N, ORF1ab and S protein were included in the TaqPath™ COVID-19 Assay multiplex reagents.

Whole -genome sequencing

Whole-genome sequencing of SARS-CoV-2 was performed on samples with a Ct < 28 using an Oxford Nanopore GridION instrument according to a previously established protocol [4]. Sequences with < 90% coverage were excluded.

Statistical analyses

We analyzed SARS-CoV-2 genomes using the Jaccard index to compute distances based on the mutation presence/absence. Multiple correspondence analysis (MCA) visualized the distance matrix, focusing on 167 upstream mutations present in at least 1% of the sequences. The first four principal components captured maximum inertia. A generalized Poisson model assessed the number of mutations per genome, taking into account lineage and group interactions. Chi-squared or Fisher’s exact tests with FDR correction identify mutations with significantly different frequencies between immunocompromised and NICs, visualized in a volcano plot. One-way ANOVA analyzes viral load and mutation count by immunosuppression status. Kruskall-Wallis and Dunn’s post-hoc tests were performed to determine which immunocompromised group had a higher number of mutations.

Results

Patient’s characteristics

A total of 641 hospitalized immunocompromised patients and 281 NICs were enrolled (Table 1A). The NICs were mainly healthcare workers (253/281), but also included patients being treated for conditions not affecting their immune system (28/281). The most common lineage of SARS-CoV-2 infecting immunocompromised patients and NICs was Omicron BA.1 (n = 185 [29.0%]) and Omicron BA.2 (n = 99 [35.0%]), respectively.

SARS-CoV-2 viral load

We observed that immunocompromised patients had a significantly lower median viral load than NICs on the day of diagnosis (20.72 [17.76–23.18] vs. 19.60 [17.12–22.38], p = 0.01).

One-way ANOVA model showed that patients receiving oncologic treatment had a significantly lower viral load compared to other immunosuppression status (Table 1B). This difference is expected to be 1.4 CT lower than other immunosuppression status.

Table 1A Baseline patients’ characteristics

Table 1B Immunocompromised characteristics and viral loads

Mutational profile of immunocompromised patients and controls

We identified 167 majority mutations that occur in at least 1% of the individuals included in the study.

Based on the distance matrix, MCA did not reveal a significant difference between the mutational profiles of immunocompromised patients and NICs, as these two groups did not appear to form distinct clusters, regardless of the time period during which they may have been infected and the variant with which they were infected (Fig. 1A). Furthermore, the sequences clustered logically according to the mutational profile of their SARS-CoV-2 lineages, with the exception of lineage BA.2 and its sublineages BA.5 and BQ.1, whose viral genomes did not appear genetically distant enough to form distinct clusters (Fig. 1B).

Fig. 1

Whole-genome analysis and single mutation analysis (A) Multiple correspondence analysis shows no distinct mutation profile between controls and immunocompromised patients. Sequences form clusters according to their lineage and their similarities from one lineage to another, such as BA.2 and its sub-lineage BA.5 and BQ.1 (B). Single mutation analysis display on the volcano plot, showing three amino acid substitutions in the Spike protein and three substitutions in the M, ORF10 and NSP3 proteins. Substitutions S: R346K and NSP3:T127N are positively associated to NICs group and S: T95I, S:N764K, ORF10: L37F and M: Q19E are positively associated to immunocompromised patients group (C)

Mutations count study

Based on a robust generalized Poisson model, the modeled mutation count was evaluated as a function of SARS-CoV-2 variant and clade. All clades had a significant effect on mutation count, with the effect increasing as the clade became more recent. Furthermore, being an immunocompromised host was associated with a higher mutation count than controls (IRR = 1.013) (Table 2). To refine our analysis, the group of immunocompromised patients was stratified according to their pathology and differences were observed (p < 0.001). Patients admitted to intensivtically different from all other groupse care were statis (Table 3) and had the highest number of mutations (Fig. 2).

Table 2 Robust generalized Poisson model parameters

Table 3 Pairwise comparisons of mutations count between patient groups stratified by type of immunosuppression using Dunn’s post-hoc test

Fig. 2

Boxplot comparing the number of mutations at diagnosis across different patient groups stratified by immunosuppression type. The groups include patients with autoimmune or inflammatory diseases, hematological oncology, HIV infection, intensive care admission, oncology, respiratory diseases, treatment with rituximab, and solid organ transplantation, as well as controls. The boxplots represent the median, interquartile range (IQR), and the range of observed values, with outliers shown as individual points. Notably, patients admitted to intensive care exhibit a higher number of mutations compared to most other groups, while controls have a lower average mutation count

Single mutation analysis

Focusing on the single mutation level according to lineage and group, 6 amino acid substitutions showed different frequencies between the two study groups. Indeed, the substitutions S: T95I (p = 0.016), S: N764K (p = 0.044), M: Q19E (p = 0.028) and ORF10:L37F (p = 0.022) are positively associated with the immunocompromised group, with ORs ranging from 3 to 6 depending on the mutation (Fig. 1C). Among these mutations, S:T95I is associated with anti-CD20 treatment including the rituximab group (p = 0.002), while ORF10:L37F is associated with the haemato-oncological group (p = 0.004).

Conversely, the NSP3:T127N (p = 0.019) and S: R346K (p = 0.045) substitutions appeared to be positively associated with the NICs group, with ORs of 0.2 and 0.4 within the immunocompromised group, respectively.

Discussion

In this study, we compared the SARS-CoV-2 genetic diversity and viral load in 683 immunocompromised patients and 296 NICs. The results showed a lower viral load in immunocompromised patients and the presence of different signature mutations compared to controls.

Counterintuitively, we observed lower SARS-CoV-2 viral loads in immunocompromised patients, with the exception of oncology patients. We hypothesize that the delay between infection and diagnostic testing differs between immunocompromised patients and NICs. Healthcare workers, who were often tested asymptomatically, are likely to have received earlier diagnoses, whereas immunocompromised patients were tested at more advanced stages of infection. Oncology patients, who were regularly tested during hospital visits, also had earlier diagnoses. This explains the higher viral loads in NICs and oncology patients.

Whole-genome analysis did not reveal distinct mutational clusters between the controls and the immunocompromised patients, suggesting no significant genetic diversity in SARS-CoV-2 between these populations at the time of diagnosis. However, the higher number of mutations in immunocompromised patients suggests prolonged viral replication prior to hospitalization and testing, leading to mutational accumulation even at the time of delayed hospital testing. In particular, patients admitted to intensive care had a significantly higher number of mutations. This observation may be explained by a possible advanced stage of the disease in these patients, which facilitates the accumulation of mutations [5,6,7].

Single mutation analysis revealed six mutations that differed between patients and NICs, for which there are no specific data in the literature allowing us to interpret these results, except for the S: T95I and S: R346K mutations. S: R346K was found to be positively associated with Omicron BA.1-infected NICs. This finding is likely due to the 14-day delay in the sampling of BA.1-infected NICs compared to BA.1-infected patients. Indeed, during this period we observed a rapid evolution from BA.1 to BA.1.1 carrying this particular substitution.

In our study, S:T95I showed a higher frequency in Delta-infected immunocompromised patients. Located in the NTD domain, S:T95I occurred in 30% of Delta sequences in France and is a signature mutation of the Iota variant, which has emerged in New York City and has shown resistance to multiple monoclonal antibody therapies, but was not circulating in France during the Delta era. S: T95I is also a signature mutation of Omicron BA.1. This suggests that S: T95I may have first emerged in immunocompromised patients during the Delta wave and then spread to the general population with the emergence of BA.1. As shown in our previous study, immunocompromised patients had a higher rate of minor mutations. Some signature mutations of newer variants were present as minor mutations in viruses infecting immunocompromised patients before they circulated in the global population [8].

Our study has several limitations. Due to its retrospective nature, the study may suffer from sampling bias, as immunocompromised patients were often diagnosed later in their disease course, which may affect viral load comparisons and mutation identification. Indeed, it has been shown that a longer disease course may result in more mutations [5,6,7]. Clinical outcomes (symptomatic or asymptomatic status at enrollment, the number of COVID-19 vaccine doses received for example), which we were unable to obtain, could have contributed to a broader understanding of the impact of different viral loads and mutational profiles. There was also a gender imbalance in the control group, probably due to the fact that it consisted mainly of nurses, a predominantly female profession. However, the effect of this imbalance is likely to be minimal, as the study focuses on baseline (D0) parameters, where the influence of sex differences in immune response is limited.

Conclusion

Finally, these findings provide valuable reference points for comparing the characteristics of SARS-CoV-2 infection between immunocompetent and immunocompromised individuals. A clearer understanding of these differences will contribute to ongoing discussions regarding infection control measures, treatment strategies, and surveillance of emerging viral variants.