Host-Microbe Multi-omic Profiling Identifies a Unique Program of COVID-19 Inflammatory Dysregulation in Solid Organ Transplant Recipients

Coronavirus disease 2019 (COVID-19) poses significant risks for solid organ transplant (SOT) recipients, who have atypical but poorly characterized immune responses to SARS-CoV-2 infection. We sought to understand and the host immunologic and microbial features of COVID-19 in SOT recipients by leveraging a prospective multicenter cohort of 1164 hospitalized patients. Using multi-omic immuoprofiling, we studied 86 SOT recipients in this cohort, who were age- and sex-matched 2:1 with 172 non-SOT controls. PBMC and nasal transcriptional profiling unexpectedly demonstrated upregulation of innate immune pathways related to interferon (IFN) and Toll-like receptor signaling, and complement activation, in SOT recipients. Longitudinal analyses across the first 30-days post-hospitalization demonstrated persistent upregulation of these innate immunity pathways in SOT recipients. The levels of several proinflammatory serum chemokines, such as CX3CL1 and KITLG, were also higher in SOT recipients at the time of hospitalization, although IFN-gamma levels were lower. We observed differential dynamics of CXCL11, which remained persistently elevated in SOT recipients over the course of hospitalization. Nasal microbiome alpha diversity was higher in SOT recipients versus controls, but no differences in taxonomic abundance beyond SARS-CoV-2 were observed. SOT recipients had higher nasal SARS-CoV-2 viral loads and impaired viral clearance compared to controls. Antibody analysis demonstrated lower anti-SARS-CoV-2 spike IgG levels in SOT recipients upon hospitalization, but no distinctions over time compared to controls. Mass cytometry demonstrated marked differences in blood immune cell populations, with SOT recipients exhibiting decreased plasmablasts and transitional B cells, and increased senescent T cells. Severe disease in SOT recipients was characterized by a less robust induction of inflammatory chemokines, such as IL-6 and CCL7, and a more subtle proinflammatory transcriptional response in the blood and airway. Together, our study reveals distinct immune features and altered viral dynamics in SOT recipients compared to non-SOT controls. We unexpectedly find that SOT recipients exhibit an augmented, predominantly innate immune response in both the blood and upper respiratory tract that remains relatively stable across disease severity, in contrast to non-SOT controls. These findings may relate to the paradoxical observation that SOT recipients have similar COVID-19 mortality rates versus the general population, despite being more susceptible to SARS-CoV-2 infection, remaining infectious longer, and having higher rates of hospitalization. In summary, we find that COVID-19 in SOT recipients is characterized by a biologically distinct immune state, suggesting the potential for unique prognostic biomarkers and therapeutic approaches in this vulnerable population.


Introduction
Coronavirus disease 2019 (COVID-19) has resulted in an enormous societal burden, with a toll of millions of infections and deaths worldwide 1 .Immunocompromised patients who have undergone solid organ transplantation (SOT) are more susceptible to SARS-CoV-2 infection and produce less robust antibody responses following vaccination 2 , although they can achieve effective T cell responses with multiple vaccinations 3 .In addition, they are more likely to be hospitalized, experience adverse clinical outcomes, and have longer durations of infectiousness compared to the general population [4][5][6][7][8] .Surprisingly, however, COVID-19 mortality in SOT recipients is not clearly higher compared to immunocompetent individuals, in propensity-matched studies 5,[8][9][10] .
To prevent organ rejection, SOT depends on immune suppression with a battery of agents including calcineurin inhibitors (e.g., cyclosporin, tacrolimus), cell cycle inhibitors (e.g., mycophenolate mofetil), and corticosteroids.This leads to an altered immunological landscape in SOT recipients, resulting in differing host responses to severe infections, including from SARS-CoV-2.SOT recipients at baseline have decreased frequencies of circulating naïve T and NK cells, and an increased frequency of terminally differentiated and senescent T cells compared with the general population [11][12][13] .Few studies, however, have pro led the immune landscape of SOT recipients in the context of severe infection, and none have yet used a multi-omic approach to assess their responses at the cellular, protein, transcriptional, and microbial levels.
The distinct immune responses of SOT recipients could theoretically be both detrimental and advantageous in the context of COVID-19.On the one hand, impaired innate and adaptive immunity in SOT recipients increases susceptibility to infection and impairs viral clearance 8 , which could lead to worse outcomes.On the other hand, because severe COVID-19 is characterized by a dysregulated, overexuberant in ammatory response [14][15][16] , intrinsic immunosuppression of SOT recipients could confer protection against severe disease.Developing a better mechanistic understanding of this tenuous immune balance in SOT recipients could inform more effective treatment approaches for COVID-19 or other respiratory viral infections, in particular the optimal use of immune modulating therapies 8 .
Here, we leverage a multicenter cohort [16][17][18] of 1164 vaccine-naïve patients hospitalized for COVID-19 to carry out the rst longitudinal immunopro ling analysis of host and microbe in SOT recipients with acute SARS-CoV-2 infection.This cohort afforded the unique opportunity to study immune responses over the course of hospitalization through concurrent analysis of transcriptional, proteomic, cellular, and antibody responses in addition to viral abundance and the airway microbiome.Contrary to expectations, we nd that SOT recipients demonstrate a globally heightened innate in ammatory response compared to non-SOT controls, and observe that established biomarkers of COVID-19 severity do not correlate with disease trajectory in this vulnerable demographic.

Patient cohort
We conducted a case-control study of patients hospitalized for COVID-19 within the IMPACC cohort, comprised 1164 patients enrolled across the US [16][17][18] between May 2020 and March 2021.86 SOT recipients from 11 medical centers were matched 2:1 by age, sex, and study site with 172 non-SOT controls from the same cohort (Fig. 1, Table 1).The most common transplanted organ type was kidney (Supp.Table 1), with approximately equal representation of heart, liver, and lung.We found no differences between the two groups in terms of ICU admission, intubation status, or COVID-19 severity as measured by ve established COVID-19 outcome trajectory groups (TG) 17 , or as measured by 28-day mortality.Trajectory groups (TG) 1-3 have mild to moderate disease based on hospital stay and level of respiratory support, while TG4 is characterized by longer hospitalizations and prolonged respiratory support requirements and TG5 by death within 28 days 17 .
To investigate host immunologic and microbial features associated with COVID-19 in SOT recipients, we assessed data from mass cytometry (CyTOF), transcriptional pro ling, proteomics, and serologic analyses in the blood, as well as nasal swab transcriptional pro ling and metatranscriptomics at hospital admission (i.e., Visit 1, which is within approximately 72 hours of hospital admission), and longitudinally at up to six timepoints up to approximately 28 days post-hospital admission (Fig. 1).

SOT is associated with increased SARS-CoV-2 viral abundance, and impaired viral clearance
We began our analyses by examining the SARS-CoV-2 viral abundance, as measured in reads per million (rpM) by nasal metatranscriptomic RNA sequencing and N-gene reverse transcription PCR (Supp.Fig. 1).SOT recipients had signi cantly higher SARS-CoV-2 viral rpM at Visit 1 (Fig. 2a, P = 5.87e-9), which could not be explained by differences in time from symptom onset (P = 0.16, Supp.Fig. 2) and did not differ based on the type of transplanted organ (Fig. 2b).Longitudinal analysis revealed that there was a signi cant association between SOT status and viral rpM, with SOT recipients demonstrating impaired viral clearance compared to non-SOT controls (Fig. 2c, P = 0.0022).

Immune cell populations and SARS-CoV-2 antibody levels
To measure immune cell populations in blood, we used mass cytometry (CyTOF) with a panel of 43 antibodies designed to identify cell lineages and markers of functional status.In PBMC samples from Visit 1, we found 5 cell types associated with SOT status (FDR < 0.05) (Fig. 3a).Plasmablasts and transitional B cells were signi cantly less abundant at Visit 1 in SOT recipients compared to controls (Figs.3a,b).Conversely, SOT recipients demonstrated increased proportions of CD4+ T (EMRA CD57hi) and CD4+ T (EMRA CD57low) cells, and CD8+ T (EMRA CD57low) cells, associated with senescence 19 (Figs.3a,b).After adjusting for SARS-CoV-2 viral rpM, only plasmablasts and CD4+ T (EMRA CD57low) cells remained statistically signi cant in terms of proportional differences, suggesting that these two cell types were associated with SOT status in a viral rpM-independent manner (Supp.Fig. 3).
We also compared anti-SARS-CoV-2 spike IgG levels between groups.SOT recipients had lower antibody levels at Visit 1 (P = 0.0004, Fig. 3c), although the rates of increase did not differ based on SOT status (Fig. 3d).

Cytokine and chemokine expression upon hospitalization and over time
Analysis of proximity extension assay (Olink) proteomics data from serum samples identi ed 16 proteins differentially expressed based on SOT status at hospital admission (Fig. 4a).The expression levels of 12 (75%) of these proteins were higher in SOT recipients versus non-SOT controls, including CX3CL1, IL15RA and KITLG, (Fig. 4b).SOT recipients had lower levels of IFN-gamma (IFNG), OSM, TNSF14, and CCL4.To assess whether differences in SARS-CoV-2 viral rpM may contribute to the observed differential protein expression, we repeated the analyses with adjustment for viral rpM.We found that the results changed minimally, suggesting that SARS-CoV-2 viral rpM did not signi cantly affect the differential protein expression between SOT recipients and controls (Supp.Fig. 4a).
Analysis of longitudinal serum cytokine expression dynamics revealed that the IFN-inducible chemokine CXCL11 decreased signi cantly over time in controls, but not in SOT recipients (Fig. 4c).After adjusting for viral rpM differences, CXCL11 dynamics remained signi cantly different between the two groups, along with a more rapid rise over time in CCL3 and CCL4 expression in the SOT recipients compared to the controls (Supp.Fig. 4b).
PBMC gene expression differences upon hospitalization, and over time At the time of hospital admission, differential expression analysis revealed 1047 differentially expressed genes (P adj < 0.05) between SOT recipients and controls (Fig. 5a, Supp.Data 1).Gene set enrichment analysis (GSEA) demonstrated that SOT recipients had increased expression of innate immunity pathways related to type I IFN, TLR signaling, complement activation, IL-1 signaling, and other functions.SOT recipients also exhibited lower expression of B-cell receptor signaling and cell cycle-related pathways (Fig. 5b, c).Adjusting for SARS-CoV-2 viral rpM in the differential expression analysis demonstrated that the increased expression of type I IFN and IL-1 signaling pathways were independent of viral rpM (Supp.Fig. 5a) We next evaluated the dynamics of gene expression over the course of hospitalization in SOT recipients and controls.SOT recipients exhibited increased expression over time of genes related to several immune pathways including types I and II IFN, IL-10, and PD-1 signaling, and CD-28 co-stimulation (Fig 5d, Supp.Data 2).Adjusting for viral rpM did not signi cantly affect results (Supp.Fig. 5b).Some signaling pathways (e.g., interferon signaling) decreased more strongly over time in non-SOT controls compared to SOT recipients (Fig. 5e).For other pathways (e.g., platelet activation, signaling and aggregation), SOT recipients demonstrated pathway upregulation over time, while in the control group downregulation was observed (Fig. 5e, Supp.Fig. 5c, Supp.Data 3).

Upper respiratory tract gene expression differences between SOT recipients and controls
Recognizing that the respiratory tract is the site of active infection in COVID-19, we performed gene expression analyses from nasal swab specimens.Surprisingly, despite the signi cant difference in the nasal viral rpM (Fig. 2a), no differentially expressed genes were identi ed between groups at a false discovery rate (FDR) < 0.05 at the time of hospital admission.GSEA nonetheless demonstrated that SOT recipients exhibited increased expression of genes related to IL-10 signaling, neutrophil degranulation, type I IFN signaling, IL-1, and IL-4/IL-13 signaling in the upper respiratory tract at the time of hospital admission (Fig. 6a, Supp.Data 4), mirroring to some extent our observations in the blood.Most in ammatory pathways differentially upregulated in SOT recipients were unaffected by viral rpM adjustment (Supp.Fig. 6a).
Similar to observations in blood, longitudinal nasal transcriptional pro ling analyses demonstrated increased expression over time of genes related to IFN signaling (Figs.6b, c), TCR signaling (Figs.6b, 6c), PD-1, IL-4, IL-13, and other immune signaling pathways in SOT recipients.In contrast, non-SOT controls demonstrated increased expression over time of genes related to neutrophil degranulation and IL-36 signaling (Fig. 6b, Supp.Data 5).Adjusting for viral load did not signi cantly change results (Supp.Fig. 6b).
Taken together, these results suggested that SOT recipients, in both the upper respiratory tract and the blood compartments, exhibit augmented innate immune responses at the transcriptional level compared to non-SOT controls, with some compartment-speci city to the relevant immune signaling pathways.

Differing relationships between interferon signaling and viral abundance in SOT recipients versus controls
In both the blood and the upper respiratory tract, SOT recipients exhibited increased type I IFN gene expression in a viral rpM-independent manner (Supp.Fig. 5a, S6a).We further explored this by comparing the relationship between IFN-stimulated gene (ISG) expression and viral rpM in SOT recipients versus non-SOT controls (Supp.Fig. 7).In blood, ISG expression strongly correlated with viral rpM in non-SOT controls, but this relationship was weaker in the SOT recipients (Supp.Fig. 7a,c).In contrast, in the upper respiratory tract, ISGs correlated with viral rpM in both groups (Supp.Fig. 7b, d).Together, these results suggested a decoupling of peripheral innate immune responses and SARS-CoV-2 challenge in SOT recipients compared to controls.

Airway microbiome differences between SOT recipients and controls
Next, we used nasal metatranscriptomics to assess whether the composition of the respiratory microbiome differed between SOT recipients and controls upon hospital admission.We found that SOT recipients had greater upper airway microbiome alpha diversity, as measured by Shannon Diversity Index (SDI), compared to controls at Visit 1 (Fig. 6d,e).No differences in bacterial community composition (beta diversity) between groups, as measured by the Bray Curtis Dissimilarity Index, were found at Visit 1 (P= 0.179, Supp.Fig. 8).

Immune correlates of COVID-19 severity differ between SOT recipients and controls
We characterized differences in host correlates of COVID-19 severity [14][15][16] between SOT recipients and non-SOT controls by comparing these groups with respect to cell type frequencies, gene expression, and protein expression differences between patients with severe COVID-19 (TG 4-5) versus those with mild/moderate COVID-19 (TG 1-3).
In both SOT recipients and controls, severe disease was characterized by reductions in several immune cell populations, including conventional dendritic cells (DCs), and intermediate (CD14+CD16+) monocytes (Fig. 7a).Only SOT recipients, however, had signi cantly lower CD11C+ CXCR5-B cells in severe disease.Severe non-SOT control patients, in contrast, had lower proportions of naïve CD4+ T cells and CD8+ NKT cells (Fig. 7a).Severe COVID-19 in controls, but not SOT recipients, was associated with a marked increase in several canonical proin ammatory serum cytokines and chemokines (e.g., IL6, CCL7, and CXCL9) (Fig. 7b).Conversely, serum levels of IFNG and IL12B were signi cantly lower in severe SOT recipients, but not in controls.
A similar analysis of PBMC transcriptomics data revealed that both SOT recipients and controls exhibited greater expression of several immune signaling pathways in severe disease, including neutrophil degranulation, innate immune system signaling, and cellular responses to stress (Fig. 7c).The expression of PBMC genes related to PD-1 signaling decreased in both groups as well.SOT recipients, however, demonstrated lower expression of genes related to TCR signaling, CD28 signaling, and TNF receptorligand interactions (Fig. 7c).
Even more notable differences between groups were observed in severity analyses of the upper airway data.Of note, severe disease in controls was characterized by increased expression of genes related to Toll-like receptor (TLR) signaling, whereas SOT recipients demonstrated the opposite (Fig. 7d).SOT recipients did not demonstrate the increases in expression of genes related to neutrophil degranulation, IL-10, IL-4/IL-13, and innate immune signaling observed in controls.

Discussion
Pharmacologic immunosuppression is necessary to prevent rejection following SOT, but comes at the expense of increased vulnerability to infection.While it is well known that SOT recipients can exhibit clinically atypical responses to respiratory infections including COVID-19 20 , the molecular features of these differences have remained unclear.Here, we performed comparative host/microbe systems immunopro ling of SOT recipients and matched non-SOT controls to address this key knowledge gap.
Unexpectedly, we found that COVID-19 in SOT recipients is not characterized by globally suppressed systemic immune signaling, but instead by augmented innate immune responses and more subtle differences across states of COVID severity (Fig. 8).
In the peripheral blood of SOT recipients, augmented innate immune signaling was characterized by higher expression of genes related to type I IFN, IL-1, and complement system pathways.Throughout the course of hospitalization, SOT recipients demonstrated consistent increases in these in ammatory signaling pathways, as well as in PD-1 and CD28 signaling.At the protein level, SOT recipients had higher levels of a few proin ammatory cytokines, such as CX3CL1, a potent chemoattractant of T cells and monocytes, and KITLG, which plays a role in hematopoiesis.In addition, SOT recipients demonstrated an increase in CXCL11 levels over time.Together, these results highlight an unexpected state of activated innate immune signaling in SOT recipients at the time of hospitalization, complemented by stable to increased expression of PD1 and other genes related to T cell signaling and exhaustion over the course of hospitalization.
We found that this state of innate immune activation was driven in part by higher SARS-CoV-2 viral load in SOT recipients, as adjustment for SARS-CoV-2 rpM impacted the magnitude of expression differences for some proin ammatory signaling pathways and cytokines.Most notable was type I IFN signaling, re ecting the proportional induction of many ISGs in response to viral RNA levels 21,22 .The lower proportions of naïve B and T cell populations in SOT recipients that we observed by mass cytometry presumably contributed to impaired control of viral replication, and the downstream activation of innate immune pathways.
In longitudinal analyses, however, even after adjusting for viral rpM differences, SOT recipients demonstrated consistently greater induction of innate immune signaling pathways in the peripheral circulation, including type I IFN signaling, compared to non-SOT controls.Furthermore, while ISG expression in the blood strongly correlated with viral rpM in non-SOT controls, this relationship was not consistently observed in SOT recipients, suggesting a partial decoupling of IFN signaling and viral pathogen burden.
In the upper airway, transcriptional differences between SOT recipients and controls were subtle, although GSEA did reveal important distinctions between groups at the pathway level.Most notably, as in the blood, SOT recipients demonstrated evidence of upregulated innate immune responses in the airways characterized by increased expression of genes related to type I IFN signaling, IL-1 signaling, and complement activation.In contrast to blood, expression of ISGs in the upper airway was strongly correlated with SARS-CoV-2 viral rpM in both non-SOT controls and SOT recipients.
In line with many prior studies [14][15][16] , higher expression of proin ammatory cytokines such as IL-6 correlated with disease severity in non-SOT control patients.In contrast, we found that the expression of most in ammatory cytokines in SOT recipients minimally differed between mild/moderate and severe disease.In addition, while controls exhibited marked severity-associated increases in the expression of canonical proin ammatory genes, this was not observed in SOT recipients.Instead, severe disease in SOT recipients was associated with lower T cell signaling gene expression in the blood, as well as less robust induction of TLR and IL-6 signaling pathways in the upper airway (Fig 7).These observations suggest a profound difference in the immune milieus in SOT versus non-SOT patients, depending on severity.The lack of association between increased proin ammatory serum cytokines in SOT patients and severe COVID-19 may have important implications, and suggests that the clinical utility of immune modulatory therapies, such as IL-6 inhibitors (e.g., tocilizumab), or JAK inhibitors (e.g., barcitinib) may not be the same in SOT recipients as in the general population.
Despite their increased susceptibility to SARS-CoV-2 infection, and comparatively poor outcomes with other respiratory infections 23 , SOT recipients paradoxically have comparable COVID-19 mortality versus the general population, in propensity matched studies 5,[8][9][10] .Our observation that severe disease in SOT recipients is not characterized by a marked increase in mortality-associated in ammatory cytokines such as IL-6, offers a potential explanation.Our ndings could also simply re ect higher levels of innate immune signaling in SOT recipients versus controls across many states of disease severity, possibly representing a compensatory effect of immunosuppressant medications that predominantly target the adaptive immune system.
These observations lead to a model of immune perturbation in COVID-19 with very different pro les in SOT recipients compared to non-SOT controls.In SOT patients on chronic immune suppression, increased senescent T cells and decreased plasmablast and B cells are unable to effectively clear virus, leading to increased and persistent viral replication.Perhaps in a compensatory effort, innate immune responses, such as type I interferon, are upregulated, and fail to attenuate appropriately over time.This dysregulated state of impaired B and T cell immunity, delayed viral clearance, and augmented innate immune signaling has parallels with aging-associated in ammatory changes 24 and may have important implications for management of immunosuppression in SOT patients with acute infection.
Our study has several strengths.These include a large, comparative immunopro ling study of SOT recipients during their rst encounter with a novel viral pathogen without the complication of variable vaccination histories.In addition, our multicenter design, and assessment of both host and microbe using a diverse range of assays are strengths.Our study also has limitations including an insu cient sample size to assess differences based on type of transplanted organ, limited clinical data regarding immunosuppressant dosing, a small sample size of intubated patients with severe COVID-19, and a lack of data speci c to allograft function.Further work is needed to determine whether our ndings in SOT recipients with COVID-19 also apply to other types of viral, bacterial, and fungal respiratory infections.
Taken together, we nd that COVID-19 in SOT recipients is characterized by a biologically distinct immune state with augmented innate signaling but lower proportions of certain adaptive immune cell populations.
The distinct immune state of SOT recipients lacks the dynamic induction of genes and cytokines associated with severe COVID-19 in the general population, suggesting a role for unique prognostic biomarkers and therapeutic approaches in this vulnerable population.

Patient enrollment and sample collection
This study leveraged data from the IMPACC cohort 17,18 , which enrolled 1286 participants from 20 hospitals across 15 medical centers in the United States between May 5th, 2020 and March 19th, 2021.
Eligible participants were participants hospitalized with SARS-CoV-2 infection con rmed by RT-PCR and symptoms or signs consistent with COVID-19.Solid organ transplant (SOT) patients were identi ed by review of medication list for immunosuppressive medications.Patients identi ed as SOT recipients were con rmed by chart review to verify transplant status and organ type.We conducted a case-control study of patients within the IMPACC cohort, matching all 86 solid organ transplant (SOT) recipients 2:1 by age, sex and study site with 172 immunocompetent controls.The detailed study design and schedule for clinical data and biologic sample collection, and shared core platform assessments were previously described 16,18 .Detailed clinical assessments and sampling of blood and upper respiratory tract were performed within ~72 hours of hospitalization (Visit 1), and on approximately Days 4, 7, 14, 21, 28 after hospital admission.As previously described 18 , biological sample collection and processing followed a standard protocol utilized by every participating academic institution.

Ethics
NIAID staff conferred with the Department of Health and Human Services O ce for Human Research Protections (OHRP) regarding potential applicability of the public health surveillance exception [45CFR46.102(l)(2)] to the IMPACC study protocol.OHRP concurred that the study satis ed criteria for the public health surveillance exception, and the IMPACC study team sent the study protocol, and participant information sheet for review, and assessment to institutional review boards (IRBs) at participating institutions.Twelve institutions elected to conduct the study as public health surveillance, while three sites with prior IRB-approved biobanking protocols elected to integrate and conduct IMPACC under their institutional protocols (University of Texas at Austin, IRB 2020-04-0117; University of California San Francisco, IRB 20-30497; Case Western reserve university, IRB STUDY20200573) with informed consent requirements.Participants enrolled under the public health surveillance exclusion were provided information sheets describing the study, samples to be collected, and plans for data de-identi cation, and use.Those that requested not to participate after reviewing the information sheet were not enrolled.Participants did not receive compensation for study participation while hospitalized, and subsequently were offered compensation during outpatient follow-up.

Common statistical analyses framework
Deidenti ed quality assured raw data was obtained from the IMPACC study and made publicly available 17,18 .All data analyses employed R v4.0.2.For each data type, we investigated the behavior of features both at Visit 1 (within 72 hours of hospital admission for most of the participants) and longitudinally for scheduled visits (Visits 1-6, up to 30 days post-hospital admission, both inpatient and outpatient samples, and excluding escalation samples).For Visit 1 analyses, we used linear modeling with age as a continuous variable and controlled for sex and baseline respiratory severity.Longitudinal measures of the WHO 7-point severity ordinal scale over time were clustered into ve trajectory groups (TG) using group-based trajectory modeling, a likelihood-based approach commonly used to group time series of clinical data, as described previously 16 .For the severity analysis, we de ned mild participants as those with trajectory group (TG) 1-3, and severe participants as those with TG 4-5.
For longitudinal analysis of SARS-CoV-2 nasal viral rpM and serum anti-Spike IgG, we used generalized additive models with mixed effects from the package gamm4 (v0.2.6) to evaluate the effects of age while controlling for sex and TG.Generalized additive modeling was preferred for these features due to their non-linear trajectories as previously reported.For all other data types, we used linear mixed effects models from the package lme4 (v1.1.25).P values in all analyses were adjusted with Benjamini-Hochberg correction.

Analysis of nasal metatranscriptomics data
Taxonomic alignments from nasal metatranscriptomics data were obtained from raw fastq les using the CZ-ID pipeline 25 , which rst removes human sequences via subtractive alignment against human genome build 38, followed by quality and complexity ltering.Subsequently, reference-based taxonomic alignment at both the nucleotide and amino acid levels against sequences in the National Center for Biotechnology Information (NCBI) nucleotide (NT) and non-redundant (NR) databases, respectively, is carried out, followed by assembly of the reads matching each taxon.Taxa were aggregated to the genus and higher phylogenetic levels from NCBI for analyses 26 .For all analyses using SARS-CoV-2 viral rpM, log transformation of total reads per million (rpM) aligned to the Beta-coronavirus genus was used.Alpha diversity (Shannon Diversity Index) was calculated using the vegan package v.2.6 in R. Differential abundance analyses for Visit 1 samples were performed using linear mixed effect modeling (using the R package nlme v3.1-162) to evaluate SOT effect on individual taxon levels at the genus, family, class, order, phylum, and superphylum levels (rpMs from lower taxon levels were summed to create higher phylogenetic level rpM), using the following formula: Taxon_abundance ~ transplant_status with random effects of (1|enrollment_site) Principle coordinate analysis (PCoA) of Bray-Curtis dissimilarity index was performed on Visit 1 nasal metatranscriptomics samples, with signi cance calculated with Adonis using the R package vegan (v2.6).

Analysis of SARS-CoV-2 viral abundance
SARS-CoV-2 viral abundance was calculated as log10(rpM+1), where rpM is the reads per million of SARS-CoV-2 as measured by nasal metatranscriptomics.The viral rpM in each organ transplant type was compared using a likelihood ratio test on the null and alternative models: Where viral_rpm was the log10-transformed viral rpM, and organ_type was the organ transplant type.Longitudinal analysis of SARS-CoV-2 viral rpM was performed using the gamm4 function from the gamm4 package (v0.2.6), using the following formula: viral_rpm ~ s(event_date, bs="cr") + s(event_date, bs="cr", by=transplant) + transplant with random effects (1|pid), ie, participant random intercept.In the formula, viral_rpm was the log10transformed viral rpM as described above, event_date was the number of days post hospitalization, transplant is the transplant status.P-value was calculated using the Chi-squared test on the gam component's reference degrees of freedom and F-statistics.

Analysis of SARS-CoV-2 antibody titers
Antibody levels against the recombinant SARS-CoV-2 spike protein receptor-binding domain (RBD) were measured using a research-grade enzyme-linked immunosorbent assay (ELISA) as described 16 .The optical density (OD) was measured and the area under the curve (AUC) was calculated, considering 0.15 OD as the cutoff.For the Visit 1 analysis, we log2-transformed the AUC values and modeled them with linear regression.For the longitudinal analysis, we also log2-transformed the AUC values, and used the linear mixed effects models (to t the null and alternative models: Where z is the log2-transformed AUC, event_date is the number of days post-admission, transplant is the transplant status, and (1|pid) is the participant random intercept.The P-values were calculated using likelihood ratio test, and adjusted with Benjamini-Hochberg correction.For visualization of longitudinal antibody levels, data was t to a third-order polynomial.
Analysis PBMC and nasal RNA-seq data RNA was extracted from PBMC and inferior nasal turbinate swabs and gene expression levels were quanti ed by RNA-seq as previously described 17 .For all RNA-seq analyses, we retained protein-coding genes that had a minimum of 10 counts in at least 20% of the samples.We normalized the gene counts using the voom function (normalize.method= "quantile") from the limma package v3.46.0, tted a linear model for the gene expression with lmFit function (default settings), calculated the empirical Bayes statistics with eBayes function (default settings), and calculated the P values for differential expression controlling for FDR.We controlled for log-transformed viral rpM in certain analyses when indicated.
For longitudinal analyses, we accounted for repeat measures from the same individual using duplicateCorrelation from the limma package, and modelled the interaction between days post-admission and transplant status as follows: z ~ event_date + transplant + event_date:transplant Where z is the log-transformed normalized expression count, event_date is the number of days postadmission, and transplant is the transplant status.
Fold-change values for Visit 1 analyses, representing the fold-change of transplant patients over control patients, and longitudinal analyses, representing the interaction term of days post-admission and transplant status, were used as the input for Gene Set Enrichment Analysis (GSEA).We used the gsePathway function from the ReactomePA v1.42.0 package to search for enriched pathways in the Reactome database, with minimum and maximum geneSet sizes of 3 and 1000, respectively.
For analysis of the relationship between interferon signaling and viral rpM at Visit 1, we rst subset the total PBMC and nasal RNA-Seq data to genes within the Reactome Interferon Signaling pathway (R-HSA-913531, n=308).We then split the data by transplant status and modelled the relationship between interferon signaling gene expression and log2-transformed viral rpM for controls and transplant recipients separately, using the approach described above.

Analysis of CyTOF data
PBMCs were phenotyped on the Fluidigm Helios mass cytometer using a panel of 46 surface and intracellular markers, and the cell types were annotated using an automated annotation pipeline as previously described 16 .Prior to analysis, we removed cells identi ed as red blood cells, multiplets, debris, and those that were not identi able with high con dence.These counts were converted to proportions per sample, by dividing each cell type count by the total cell count.The minimum proportion per cell type across all samples was added to each sample prior to log2-transformation, to avoid taking the logarithm of zeros.
For the Visit 1 analysis, the log2-transformed cell type proportions were modeled with linear regression.
For the longitudinal analysis, the log2-transformed cell type proportions were modelled with linear mixed effects models (to t the null and alternative models: Null: z ~ event_date + transplant + (1|pid) Alternative: z ~ event_date + transplant + event_date:transplant + (1|pid) Where z is the log2-transformed cell type proportion, event_date is the number of days post-admission, transplant is the transplant status, and (1|pid) is the participant random intercept.The P-values were calculated using likelihood ratio test, and adjusted with Benjamini-Hochberg correction.

Analysis of in ammatory protein (Olink) data
All samples were processed with the Olink multiplex assay in ammatory panels (Olink Proteomics), according to the manufacturer's instructions and as previously described 16 .This in ammatory panel included 92 proteins associated with human in ammatory conditions.Target protein quanti cation was performed by real-time micro uidic qPCR via the Normalized Protein Expression (NPX) manager software.Data were normalized using internal controls in every sample, inter-plate control and negative controls, and correction factor and expressed as log2 scale proportional to the protein concentration.For additional quality control, we set any NPX measurements below the assay's limit of detection (LOD) to zero.Next, we excluded proteins that were detected in fewer than 20% of samples, resulting in 84 proteins for analysis.
For the Visit 1 analysis, we standardized the NPX values and modeled them with linear regression, with and without adjusting for SARS-CoV-2 viral rpM.The viral rpM was also calculated as log10(rpM+1).We then t the following linear model: z ~ transplant + transplant:TG Where z is the standardized protein level, transplant is the SOT status, and transplant:TG is the interaction term between SOT status and disease severity.This formulation allows us to nd two separate coe cients (i.e., two separate log fold-change values) for the effects of severity, one for the SOT group and one for the control group.
For the longitudinal analysis, we also standardized the NPX values, and used the linear mixed effects models to t the null and alternative models: Null: z ~ event_date + transplant + (1|pid) Alternative: z ~ event_date + transplant + event_date:transplant + (1|pid) Where z is the standardized protein level, event_date is the number of days post-hospital admission, transplant is the transplant status, and (1|pid) is the participant random intercept.The P-values were calculated using a likelihood ratio test, and adjusted with Benjamini-Hochberg correction.

Figures Figure 1
Figures

Figure 3 Compared
Figure 3

Figure 7 Host
Figure 7

Table 1 .
Clinical and Demographic Features of Cohort.