Evolutionary rewiring of host metabolism and interferon signalling by SARS-CoV-2 variants

  1. Riaz-Ali Somji
  2. Jonathan C Brown
  3. Shahid Rowles-Khalid
  4. Tukur Abdullahi
  5. Dongsheng Luo
  6. Natalie Barthel
  7. Michael Mülleder
  8. Markus Ralser
  9. Wendy S Barclay
  10. Charalampos Rallis
  11. Efstathios S Giotis

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

SARS-CoV-2 variants differ in transmissibility and immune evasion, but their effects on host-cell metabolism and signalling remain less defined. Using integrated transcriptomic, phosphoproteomic, and amino acid profiling in primary nasal epithelial cells, we compared early and late host responses to pre-Omicron variants (Alpha, Beta), Delta, and Omicron subvariants (BA.1, BA.5). Pre-Omicron strains broadly suppressed antiviral interferon-stimulated gene expression and reprogrammed metabolism by reducing mitochondrial oxidative phosphorylation and β-oxidation. Delta infection was associated with extensive transcriptional and metabolic remodelling, characterised by activation of stress- and growth-related kinases and selective retention of biosynthetic amino acids, consistent with a host response to stress and viral modulation of interferon-associated signalling. In contrast, Omicron infection elicited a more restrained response dominated by cytokine and survival pathways, with limited metabolic activation and interferon suppression. Together, these findings suggest SARS-CoV-2 has progressively evolved toward a strategy that maintains efficient upper-airway replication while minimising epithelial stress and inflammation.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    1. General Statements [optional]

    We thank the reviewers for their careful evaluation of our manuscript and for their constructive comments. The reviews recognise the relevance of the topic and the value of the multi-omics approach used to investigate host responses to SARS-CoV-2 variants in a physiologically relevant primary nasal epithelial cell model.

    In response to the reviewers' comments, we revised the manuscript to improve clarity of presentation, strengthened the contextualisation of the experimental design, and moderated the interpretation of the results. We also incorporated additional analyses based on existing datasets to better characterise infection burden and host responses across variants.

    Importantly, the MOI reported in the original manuscript (0.01) was a typographical error; all infections were performed at MOI 0.1 as documented in the GEO dataset (GSE271378). This was corrected throughout the manuscript.

    Overall, the revisions were intended to clarify the experimental framework, strengthen the integration of the multi-omics datasets, and ensure that the conclusions accurately reflect the scope of the study as a comparative systems-level analysis of variant-associated host-response signatures rather than a mechanistic dissection of individual pathways.

    2. Description of the planned revisions

    Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

    Reviewer summary: In the submitted work, authors seek to understand the transcriptional and metabolic changes induced by different variants of SARS-CoV-2 infection. They employ a model of pooled, primary nasal epithelial cells (NEC) differentiated within an air-liquid-interface. Subsequently, cultures are infected with isolates representing key variants of SAR-CoV-2 from initial D614G, Alpha, Beta, Delta, and Omicron. Following initial characterization, authors compare transcriptional changes at 24 and 72 hours post infection. Analysis focuses on differentially expressed genes, upregulated Interferon Stimulated genes, and transcripts within known metabolic pathways. Subsequently, infected cultures are also analyzed by phosphoproteomic analysis to identify changes in cell signaling and measured for amino acid content. Throughout, changes in each profile are correlated with different variants of SARS-CoV-2, with Delta and Omicron revealing particular diametrically opposed changes. There are reasonable controls applied, including the use of IFNalpha treatment to "benchmark" ISG production. Overall, authors create a picture that Omicron infections do not suppress IFN signaling as efficiently as Delta variants and further exhibit limited hallmarks of cell stress and metabolic dysregulation. This is a remarkable study that attempts to cross-correlate multiple -omics analyses of cell responses to characterize differences in infection. It is very well written and the data is exemplary. I do have some concerns related to the placement and emphasis of interpretation in the results section that need to be revised. Beyond my stylistic concern, the interpretation of the experimental changes between variants are compromised by the failure to analyze the extent of infection within the NEC model. Using an MOI of 0.01 will produce a dramatically heterogeneous extent of infection at both 24 and 72 hours post infection that will also depend on the extent of viral transmission within the culture. The limited analysis of secreted E-gene detection is insufficient to overcome the inherent unequal comparison of cell responses between variants. There are ways to assuage, but not eliminate, this problem when it comes to comparing and interpreting experimental results. My concerns and suggestions are detailed in the concerns below.

    Response: We thank Reviewer #1 for the very positive assessment of the study, which supports a decision to publish, and for the constructive suggestions. We agree that the interpretation of comparative bulk multi-omics data in differentiated NEC cultures should be carefully framed in light of variant-specific infection dynamics. In response, we revised the manuscript substantially. Importantly, the MOI reported in the manuscript (0.01) was corrected, as all infections in this study were performed at MOI 0.1, as correctly described in our GEO submission (GSE271378). We also quantified SARS-CoV-2 reads directly from the RNA-seq libraries to estimate infection burden in each sequenced sample, added infectious virus titre measurements (PFU/ml), expanded the analysis of IFNα-treated samples to include DEG and pathway-level comparisons, improved figure clarity and legends, and substantially tempered the interpretation throughout the Results and Discussion. We believe these revisions address the reviewer's concerns and strengthen the manuscript.

    1. Heterogeneous extent of infection. The MOI of 0.01 used to initiate infection is extraordinarily low for the types of analysis that is employed with the NEC culture. The interpretation of the data does not take into account that there will be infected and uninfected cells, of varying extents, making up the changes observed. Further, the variants likely have differing abilities to spread through the NEC culture, complicating both interpretation of changes and comparison between variants. At a minimum, authors need to evaluate the extent of SARS-CoV-2 infection through either flow cytometry or immunofluorescence analysis against viral protein(s). It is possible that Omicron, while secreted well, has more limited transmission allowing for more cells to mount an IFN response. Delta is a prolifically spreading virus that likely has more extensive infection at 72 hpi than the other variants. These statements are conjecture and highlight how such differences could alter the interpretation of the subsequent experiments. Response: We thank the reviewer for raising this important point. We would first like to clarify that the MOI reported in the manuscript (0.01) was a typographical error. All infections in this study were performed at MOI 0.1, as correctly documented in the RNA-seq dataset deposited in GEO (GSE271378). The manuscript text, Methods, and figure legends was corrected accordingly. MOI values in this range are commonly used for infections of differentiated airway epithelial cultures and allow productive infection while preserving epithelial integrity.

    We agree that infection heterogeneity is an important consideration when interpreting bulk transcriptomic, phosphoproteomic and metabolic measurements in differentiated epithelial cultures. We argue that such heterogeneity is expected in air-liquid interface nasal epithelial models, where SARS-CoV-2 infection occurs within a structured epithelium composed of multiple cell types and infected cells coexist with neighbouring bystander cells responding to paracrine interferon signalling. Bulk multi-omics measurements therefore capture the integrated epithelial response to infection rather than purely cell-intrinsic responses.

    To better contextualise infection burden within the sequenced samples, we included an additional analysis quantifying SARS-CoV-2 reads directly from each RNA-seq library and infectious virus titres (Figure 1). In the revised manuscript, these data are presented together in a new Supplementary Figure 1, which distinguishes intracellular viral RNA abundance from infectious virus production. The viral read analysis shows that intracellular viral RNA increases between 24 and 72 hpi across all variants and becomes broadly similar across lineages by 72 hpi, whereas plaque assays show that BA.1 has the highest early infectious output and Delta reaches the highest infectious titres at 48-72 hpi. We used these data to revise the Results and Discussion so that host-response differences are interpreted in the context of infection burden, while also making clear that intracellular viral RNA abundance, extracellular viral RNA output and infectious virus production are related but distinct measures of variant biology.

    Figure 1: Intracellular viral RNA reads (RNA-seq) (A) and infectious virus titres (PFU mL⁻¹; B) across SARS-CoV-2 variants.

    Further evaluation of IFNalpha treated cells. The paper emphasizes the ISG analysis, but the IFN treated cells should be included in the DEG and metabolic pathway analysis. IFN treatment is known to alter metabolic changes in cells, and it would be valuable to see those changes reflected in your analysis. Consider the evidence presented in the following: Fritsch SD, Weichhart T. Effects of Interferons and Viruses on Metabolism. Front Immunol. 2016 Dec 21;7:630. Heer CD, Sanderson DJ, Voth LS, Alhammad YMO, Schmidt MS, Trammell SAJ, Perlman S, Cohen MS, Fehr AR, Brenner C. Coronavirus infection and PARP expression dysregulate the NAD metabolome: An actionable component of innate immunity. J Biol Chem. Elsevier BV; 2020 Dec 25;295(52):17986-17996. Palmer CS. Innate metabolic responses against viral infections. Nat Metab. 2022 Oct;4(10):1245-1259 Further, It is possible that the changes attributed to Omicron are quite similar to the effects of the IFN treatment, given the extensive ISG detection. The same is true for the phosphor-proteomic analysis and amino acid content. I also have concerns that using a treatment of IFNalpha that impacts all cells as a benchmark for heterogeneous infection is not truly comparable. How was the concentration of IFN chosen? What was the extent of IFN activation in the culture?

    Response: In response to this suggestion, we performed pathway enrichment analysis of IFNα-treated samples to evaluate whether interferon stimulation alone induces the metabolic pathway signatures observed during viral infection. IFNα treatment produced the expected transcriptional interferon-stimulated gene programme but did not result in significant enrichment of the metabolic pathways highlighted in the infection comparisons (Figure 2). Specifically, pathways related to glycolysis/gluconeogenesis, glutathione metabolism, fatty acid metabolism, mitochondrial pathways, and oxidative phosphorylation showed only limited or modest negative enrichment scores and did not approach the magnitude of enrichment observed in virus-infected cultures. These results indicate that interferon signalling alone does not reproduce the metabolic pathway signatures associated with variant infection. The IFNα pathway analysis was included in the revised manuscript as supplementary data and referenced in the Results section.

    We agree that IFNα treatment of all cells is not directly equivalent to heterogeneous viral infection within differentiated NEC cultures. The IFNα concentration used was selected based on previous optimisation experiments showing robust induction of canonical ISGs in differentiated airway epithelial cultures. In the revised manuscript we clarified that the IFNα condition is used as a reference for interferon-responsive transcription rather than as a direct surrogate for infected cultures. We provided additional methodological clarification regarding how the IFNα concentration was selected and how interferon activation was benchmarked in NEC cultures.

    Further correlation of transcriptional changes with metabolic changes - While many published works emphasize transcriptional changes as a proxy for metabolic changes, there are robust methods that can be applied to directly analyze metabolite content and changes in the context of viral infection. In particular these studies should be assessed and compared for the interpretation of the presented results: Kramaric, T., Thein, O.S., Parekh, D. et al. SARS-CoV2 variants differentially impact on the plasma metabolome. Metabolomics 21, 50 (2025). Loveday EK, Welhaven H, Erdogan AE, Hain KS, Domanico LF, Chang CB, June RK, Taylor MP. Starve a cold or feed a fever? Identifying cellular metabolic changes following infection and exposure to SARS-CoV-2. PLoS One 2025 Feb 12;20(2):e0305065. Irún P, Gracia R, Piazuelo E, Pardo J, Morte E, Paño JR, Boza J, Carrera-Lasfuentes P, Higuera GA, Lanas A. Serum lipid mediator profiles in COVID-19 patients and lung disease severity: a pilot study. Sci Rep. 2023 Apr 20;13(1):6497. Luke Whiley, Nathan G. Lawler, Annie Xu Zeng, Alex Lee, Sung-Tong Chin, Maider Bizkarguenaga, Chiara Bruzzone, Nieves Embade, Julien Wist, Elaine Holmes, Oscar Millet, Jeremy K. Nicholson, and Nicola Gray, "Cross-Validation of Metabolic Phenotypes in SARS-CoV-2 Infected Subpopulations Using Targeted Liquid Chromatography-Mass Spectrometry (LC-MS)", Journal of Proteome Research 2024 23 (4), 1313-1327

    Response: We thank the reviewer for this important comment and agree that transcriptional pathway enrichment alone cannot establish metabolic flux or enzyme activity. Our intention in this study was to integrate transcriptomic signatures with complementary data layers, including phospho-signalling profiles and targeted intracellular amino acid quantification, to provide a comparative systems-level view of host responses to SARS-CoV-2 variants in nasal epithelial cells.

    We acknowledge that transcriptional enrichment does not necessarily reflect pathway activity and that our amino acid measurements represent a targeted metabolite readout rather than a comprehensive metabolomic or flux-based analysis. In the revised manuscript, we have therefore moderated the language used when describing metabolic changes and refered to pathway enrichment more cautiously as indicative of potential metabolic engagement rather than direct metabolic regulation.

    We have also expanded the Discussion to contextualise our findings with the metabolomic studies suggested by the reviewer and related work examining metabolic responses to SARS-CoV-2 infection.

    Editing to limit interpretation within experimental results. I appreciate that this is a stylistic concern and it is an issue in the paper. Statements in the results are often over-reaching. Some examples include: Line 156 -"suggesting attenuated or delayed early sensing" - The Low MOI and time leaves these results open to various explanations. Better to just state and move on. Line 157 "Delta drove the most extensive" - drove has a lot of assumption. "produced" "resulted in " or something more passive is more appropriate Line 179 "pointing to sustained suppression of interferon responses." - sustained is a leading interpretation. Effective? Comprehensive? again, the MOI is complicating interpretations of global transcript changes. Line 186 "suggesting a weaker activation of interferon signaling" Too much leading interpretation here. You detect fewer ISGs that are differentially regulated. Could be for many reasons.

    Response: We appreciate this comment and agree. We have revised the Results section throughout to make the language more descriptive and less interpretive. The specific examples highlighted by the reviewer were changed accordingly, and similar phrasing elsewhere in the Results was also softened. Mechanistic interpretation was reduced and moved to the Discussion where appropriate.

    Line 72 "has evolved unique strategies" Unique can be easily misconstrued to mean different mechanisms. More likely, it is a subtle balance between promotion of viral replication and suppression of IFN responses.

    Response: We agree and have revised this wording to avoid overstating mechanistic distinctiveness.

    Line 126 - 128 "NECs were derived from three commercially available donor pools". The following text doesn't make it clear that they are the same produce from different lots. The methods clarify somewhat, but should be clarified for transparency.

    Response: We thank the reviewer for noting this lack of clarity. We revised the relevant text in the Results and Methods to make clear that the NECs were derived from the same commercial product obtained across different lots/batches.

    Line 129 "Viral replication kinetics" Need to highlight that this is detection of secreted viral genomes. which is a proxy measure for replication and dissemination in the culture. Direct measurement of the extent of infection is not being made nor can be interpreted.

    Response: We agree and have revised the text and figure legend to clarify that the RT-qPCR measurements represent extracellular viral genome copies released into the apical supernatant and therefore provide a proxy measure of viral RNA output and dissemination within the culture rather than a direct measurement of infection extent. To better contextualise infection dynamics, we have complemented the RT-qPCR data with two additional measures of viral burden. First, we quantified SARS-CoV-2 reads directly from the RNA-seq libraries to estimate intracellular viral RNA abundance in the sequenced samples. Second, we measured infectious virus titres (PFU ml⁻¹) by plaque assay. These complementary analyses were presented in Supplementary Figure 1 and allow us to distinguish extracellular viral RNA release, intracellular viral RNA abundance, and infectious virus production. The revised manuscript explicitly acknowledges that bulk multi-omics measurements reflect mixtures of infected and bystander epithelial cells and therefore capture the integrated epithelial response to infection rather than the exact proportion of infected cells.

    Line 149 "Differentially expressed genes (DEGs)" What is the comparison group? The figure legend/design suggests that IFNa treatment. Is there a matched uninfected control for each timepoint as well? Later experiments specify the comparison group. Text should be clarified here for transparency.

    Response: We thank the reviewer for highlighting that the comparison group was not clearly described in the Results section. Differential expression analysis was performed by comparing each variant-infected condition with mock-infected control samples collected at 24 h. The same mock reference was used for comparisons at both 24 and 72 hpi. IFNα-treated samples were analysed separately and were not used as the reference condition for DEG identification. We have clarified this explicitly in the revised Results and Methods sections.

    Line 224 and Figure 4B - I don't see the value of the "merged NES" values given these are only aggregate of the Pre-Omicron and Omicron species. If you had compared multiple D614G and Delta variants, then there would be utility.

    Response: We agree that the merged NES values provide only a broad visual summary and that the most informative comparisons are at the individual variant level. In the revised manuscript we reduced the emphasis on the merged analysis and clarify in both the Results text and the figure legend that interpretation is primarily based on the variant-specific enrichment profiles, with lineage grouping shown only as a visual summary.

    Line 261 "quantified at 24 hpi" Why this timepoint? Changes were minor and not representative to extensive infection.

    Response: We thank the reviewer for this comment. The amino acid measurements were performed at 24 hpi to capture early metabolic responses to infection, in parallel with the phosphoproteomic analysis performed at the same time point. We agree that at this stage of infection the NEC cultures likely contain mixtures of infected and bystander epithelial cells, and therefore the amino acid measurements reflect the integrated metabolic state of the culture rather than infected cells alone. We clarified this rationale and limitation in the revised Results and Discussion sections.

    Line 268 "rather than variation in cell number." I appreciate the rigor and control of experimentation. And how many of those cells are infected? That is not controlled.

    Response: We thank the reviewer for this important point. We agree that normalisation to viable cell number does not control for infection heterogeneity within the cultures. In the revised manuscript, we revised this sentence to clarify that the amino acid measurements were normalised for cell number, but that, because they were obtained from bulk cultures at 24 hpi, they reflect the integrated metabolic state of infected and bystander cells rather than infected cells alone.

    Line 428-429 "direction of regulation" This seems like an over-interpretation of the data. You have performed pathway analysis based on the quantity of RNA transcription detected in sequencing then imputing an interpretation of regulation. Without pulse labeling of metabolic standards or kinetic analysis of metabolite quantity, it is difficult to assert regulatory direction.

    Response: We agree with the reviewer and have revised this wording accordingly. In the revised manuscript, we avoided describing pathway-level RNA-seq enrichment as direct regulation in a mechanistic sense. Instead, we refered more cautiously to positive or negative pathway enrichment based on transcript abundance patterns, which more accurately reflects the information provided by the enrichment analysis.

    Referee cross-commenting I am in agreement with the comments and suggestions of Reviewer #2 and #3. In particular, the comment of Reviewer #3 to estimate viral replication from the RNASeq data is quite valuable to begin addressing some of the concerns about the extent of viral replication. It does not negate the need to further assess productive viral titer (PFU/mL) or the extent of viral infection (immunofluorescence or flow cytometry). I also agree with Reviewer #3 regarding the extent of mechanistic interpretation that can be drawn from the current study. This concern can largely be addressed through revision of the text and a tempering of the interpretations that are drawn. I also agree and appreciate the detailed analysis of reviewer #2 regarding the inconsistencies between the text and the figures. It is critically important to be consistent in the data and presentation of these complex experiments. Resolving these issues will only strengthen the work.

    Response: We thank the reviewer for these additional comments and for highlighting the useful points raised by Reviewers #2 and #3. In line with these suggestions, we quantified viral reads directly from the RNA-seq libraries to provide an estimate of infection burden in the sequenced samples and included infectious virus titre measurements (PFU/ml) to complement the existing replication analyses. We agree that the current dataset supports a comparative systems-level analysis rather than strong mechanistic conclusions, and we therefore tempered the interpretation throughout the manuscript. Finally, we carefully reviewed and revised the figures, legends, and associated text to ensure consistency and clarity in the presentation of the data.

    Reviewer #1 (Significance (Required)):

    The work detailed in this manuscript is takes a very broad approach to identify differences in the effects of SARS-CoV-2 variant infections. Elements of this work have been published, including transcriptomics, metabolomics, and phosphoproteomics. This work is significant in that multiple variants are evaluated with comparable methods in the very relevant human nasal epithelial cell model. The use of this model, and the direct integration of multiple -omics, sets this work apart from previously published studies. This cross-omic analysis, with the IFN-treated controls, provides a robust foundation of data that can be used to detail the differences in the response to the SARS-CoV-2 variant infections. That said, a significant limitation to the study was the low MOI used to initiate infection and the lack of detailed analysis infection progression of the different variants. Further, there is limited comparison of the IFN-treatment condition in relation to the transcriptional changes, and no inclusion of IFN-controls in the other methods. Both of these limitations undercut the potential significance of the paper and its findings. Audience: This work will have be important to bench researchers interested in further characterizing and comparing the effects of SARS-CoV-2 infection. Potentially, clinicians involved in diagnostics will find utility in the study of changes for potential biomarker analysis for severe COVID19 disease. My expertise is the field of virology, having studying multiple RNA and DNA viruses, including SARS-CoV-2, to understand virus-cell interactions. My focus includes primary cell culture models of infection, proteomic and metabolic analysis of infection induced changes, and monitoring the spread of viral infection through direct and indirect measurements.

    Response: We thank the reviewer for the positive assessment of the significance of the study and for recognising the value of the integrated multi-omics analysis in a physiologically relevant human nasal epithelial cell model. We also appreciate the reviewer's constructive comments regarding infection burden and the interpretation of the IFNα reference condition. As noted above, the reported MOI of 0.01 was a typographical error and was corrected to 0.1 throughout the manuscript. To further address the reviewer's concerns regarding infection extent, we quantified viral reads directly from the RNA-seq libraries and include infectious virus titre measurements (PFU/ml) as an additional measure of productive infection. We also expanded the analysis of IFNα-treated samples to include differential expression and pathway-level comparisons, allowing more direct contextualisation of virus-induced transcriptional responses relative to a canonical interferon-stimulated programme. We believe that these revisions strengthen the interpretability and overall significance of the study.

    Reviewer #2 (Evidence, reproducibility and clarity (Required)):

    The authors of the manuscript entitled "Evolutionary rewiring of host metabolism and interferon signalling by SARS-CoV-2 variants" investigated the diversity of different SARS-CoV-2 isolates regarding gene expression, kinase activity and amino acid profiles in infected primary human nasal epithelial cells. Somji et al. found certain distinct alterations of measured factors after infections compared to mock and differences in cells infected with the mentioned different SARS-CoV-2 isolates. The topic of the manuscript as such is of high importance since understanding virus host interactions in general and virus host coevolution particularly on the level of cellular metabolism and beyond comes with great potential in deeper understanding the infection biology of viral invaders. Nevertheless, the study needs to be enlarged and further defined, the experimental set up has to be improved and the drawn conclusions have to be proven by experiments. The presentation of the obtained data needs to be improved, checked and carefully chosen to allow the reader to follow the article in a much more guided way. At this stage of experimental data depth, presentation and interpretation, there is room for certain overinterpretations of the biological meanings of the presented data.

    Response: We thank Reviewer #2 for the careful evaluation of the manuscript and for recognising the relevance of the topic. We agree that the original submission required clearer presentation, stronger contextualisation of the experimental design, and more cautious interpretation. In response, we revised the figures, legends, and linked text; clarify the number of biological and technical replicates for each experiment; added viral RNA read quantification from the RNA-seq libraries and infectious virus titres (PFU/ml); expanded the IFNα-related analyses; and moderated the conclusions throughout. We believe these revisions directly address the reviewer's core concerns.

    The authors state about virus growth kinetics in Fig.1. To be able to do so in full extend, virus particle counts (PFU/ml) need to be measured and included in this data set.

    Response: We agree and added the infectious virus titre measurements (PFU/ml) to complement the RT-qPCR genome measurements (Supplementary Figure 1).

    From Fig.2 on, the presentation and introduction of the data set is often very hard to follow. Certain panel labeling is not correct e.g. in Figure 2, Figure 2A is not introduced, 72h data are linked to Figure 2C but 2C is a Venn diagram of 24h gene expression downregulation. The Venn diagrams are not mentioned in the text at all. This problem is occurring at different occasion, which makes it hard to impossible to follow the experimental flow of the study. Therefore, a complete revision of the data presentation within the figures and the linked text is needed. Further example, lines 213 and 224, Figure 4B two times mentioned with different data supposed to be shown in Fig. 4B.

    Response: We thank the reviewer for identifying these issues. We comprehensively revised the figure panel labelling, figure legends, and linked text to ensure consistency and readability throughout the manuscript.

    The authors are inconsistent with including statistics in their figures. Please include all statistics in your figures to allow the reader to get this information. Please declare how often and how each experimental set has been done and clarify e.g. in the figure legends. In addition, please improve the figure quality for better allowance of cross comparability of data sets. As example, used the same x-axis scale for all graphs in Fig 4.

    Response: We agree and have revised the figures and legends accordingly. Statistical annotations have been added in the results section, and full values associated with the pathway enrichment analysis are now reported in Supplementary Table S2. For the amino acid measurements, individual biological replicate values are now displayed in the figure panels rather than only summary statistics. Replicate numbers and experimental design (biological replicates, technical replicates, and donor batches where relevant) are now explicitly stated in the figure legends and Methods section.

    To improve comparability across datasets, figure formatting was standardised throughout the manuscript. In particular, the x-axis scales in Figure 4 (below) were harmonised across panels to allow direct comparison of normalised enrichment scores between variants and time points. Additional adjustments were made to improve figure clarity, including consistent axis labelling, colour scales, and panel annotations.

    The authors create claims about metabolic profiles without measuring deeper metabolic circumstances. Why are only amino acids measured and not metabolite concentrations in general. Metabolic gene expressions as measurement of metabolic pathway activities can be strongly misleading since gene expression per definition does not necessarily mean enzyme activity, which of course is finally important for pathway activity as well.

    Response: We agree that amino acid profiling represents a targeted metabolic readout rather than a comprehensive metabolomic analysis, and that transcript abundance does not directly equal enzyme activity or flux. We have revised the manuscript throughout to reflect this limitation more clearly and to expand the Discussion to place our targeted amino acid data and pathway enrichment analyses in the appropriate context.

    The authors need to carefully crosslink the obtained data sets. As an easy example, how much of the found differences in gene expression, pathway activities etc. is due to viral growth differences. With other words, are there regulatory differences or are the differences seen due to different growth kinetics. Are ISG expression level linked to virus growth? These type of questions not be asked and correlations need to done by the authors to guide the reader through all those assays conducted in this study.

    Response: We agree that infection burden is an important variable when interpreting bulk multi-omics datasets obtained from infected epithelial cultures. To address this, we incorporated two additional measures of viral burden into the revised manuscript. First, SARS-CoV-2 reads were quantified directly from the RNA-seq libraries to estimate intracellular viral RNA abundance in the sequenced samples. Second, infectious virus titres (PFU ml⁻¹) were measured by plaque assay. These complementary datasets are presented in Supplementary Figure 1.

    In the revised manuscript, these measures are used to contextualise the transcriptomic and pathway analyses. Intracellular viral RNA reads increased across variants between 24 and 72 hpi and reached broadly comparable levels by 72 hpi, whereas infectious virus production differed between variants, with Delta producing the highest titres at later time points. We therefore revised the Results and Discussion to explicitly acknowledge that bulk transcriptomic, signalling and metabolic signatures may reflect both infection burden and variant-specific regulatory differences. For example, ISG induction at 72 hpi is discussed in the context of similar intracellular viral RNA levels across variants, indicating that differences in interferon-responsive transcription are not explained solely by viral RNA abundance.

    More broadly, we now emphasise that NEC cultures contain mixtures of infected and bystander epithelial cells and that the multi-omics datasets capture integrated epithelial responses rather than cell-intrinsic responses alone. These revisions strengthen the crosslinking between infection dynamics and host-response datasets while avoiding overinterpretation of variant-specific regulatory mechanisms.

    .

    Referee cross-commenting I do fully agree with reviewer 1 and 3 in terms of the importance of much more comprehensive data on virus growth. Measurement of real virus progeny (PFU/ml) and viral protein and RNA expression is needed to state about the importance of altering viral dynamics for interpreting the findings. I do fully agree with reviewer 1 and 3 that data analysis, presentation and interpretation has to be improved. Information such as how often has each experiment been done and how has the experimental set up been constructed has to be clarify e.g. in the figure legends. As reviewer 1 mentioned, direct analysis of metabolite concentrations is needed to be able to judge about metabolic changes driven by the different SARS CoV-2 variants. In line with both, reviewer 1 and 3, conclusions drawn by the authors should be toned down. More data and improved data analysis and presentation are needed to foster the conclusions drawn .

    Response: We thank the reviewer for these additional comments. In response, we added PFU/ml and RNA-seq-derived viral read data, improved experimental detail in the Methods and figure legends, clarified the scope and limitations of the amino acid measurements, and substantially moderated the interpretation throughout the manuscript.

    Reviewer #2 (Significance (Required)):

    While the topic as such is interesting and hoighly relevant, the manuscript has several major flaws, both with regard to paper organisation and content. In the current state it is hard to judge, whether the data are of significance.

    Response: We appreciate this assessment and hope that the extensive revisions in response to the reviewers' comments make the organisation, data presentation, and significance of the study much clearer.

    Reviewer #3 (Evidence, reproducibility and clarity (Required)):

    The manuscript entitled "Evolutionary rewiring of host metabolism and interferon signalling by SARS-CoV-2 variants" by Somji and colleagues sets out to understand SARS-COV-2 variant biology in primary nasal epithelial cells. Understanding this and differences in variant-specific host-virus interactions is essential to understand the molecular basis of replication advantages and enhanced transmission that ultimately lead to variant dominance. The authors employ global transcriptomic, phosphor-proteomic and amino acid metabolism assays with the aim to identify variant-specific changes to cell metabolism and innate immune activation in a comparative systems-level approach. Importantly, this work is performed in primary nasal epithelial cells. It is essential to understand variant biology in the context of relevant primary cell infection models and NECs are a great choice to address the proposed research question. The work is conceptually interesting, but largely descriptive. While this can still be useful for the field, it requires appropriate framing of the interpretations of the data. I agree with the authors that there will be virus- specific signatures that will contribute to variant fitness, but this dataset makes it hard to draw strong conclusions. The main problem with the manuscript and the interpretation are dramatic differences in viral replication. While some of the conclusions are tantalising and would warrant further investigation, I would expect to see some experimental validation to substantiate the interpretation. In the absence of experimental validations and mechanism, the conclusions should be stated as such and contextualised more with previously published work.

    Response: We thank Reviewer #3 for the thoughtful and constructive assessment. We agree that the study is primarily comparative and systems-level in nature and that the original submission overreached in parts of the interpretation. In response, we moderated the conclusions and reframed the manuscript as a comparative analysis of lineage-associated host-response signatures that generates mechanistic hypotheses for future work, rather than claiming definitive causal mechanisms. We also added additional data on viral burden, revised the analysis description, improved figure presentation, and expanded the contextualisation with previously published work.

    A major concern that I have is the analysis of the RNASeq data. Experimental design, analysis and presented data require some clarification: Too little experimental detail for the RNASeq data is given. How many replicates were sequenced/analysed? The figure legend state three independent experiments - but how many individual replicate transwells per condition (and NEC batch) were used? This information needs to be included in the manuscript. Generally, clarification on how many replicates were used per experiment needs to be included in the figure legends for all data panels.

    Response: We thank the reviewer for highlighting the need for clearer reporting of RNA-seq experimental design and replicate structure. We have revised the manuscript to explicitly clarify replicate numbers, experimental batches, and sequencing quality control. RNA-seq experiments were performed using three independent batches of donor-pooled nasal epithelial cultures (MucilAir{trade mark, serif}). For most infection conditions and time points, two to three biological replicate transwells were sequenced per condition derived from independent NEC culture batches. A small number of libraries did not pass sequencing quality control thresholds (e.g. insufficient sequencing depth or technical library failure) and were therefore excluded from downstream analysis, resulting in minor variation in replicate numbers across conditions. To improve transparency, sequencing depth and library quality metrics for all RNA-seq libraries are now provided in Supplementary Table S6. In addition, principal component analysis (PCA) of the RNA-seq dataset has been included as Supplementary Figure S5 to illustrate sample clustering and replicate consistency across conditions. All raw sequencing data, sample metadata, and replicate information are available in the GEO repository under accession number GSE271378. We have also revised the Methods and figure legends throughout the manuscript to explicitly state replicate numbers for each dataset.

    .

    The reported log2 fold changes are beyond what is biologically reasonable. A log2 fold change of 120 or even 30 (Fig.3D, suppl table) indicates issues with the data analysis. It is worth revisiting the analysis and additionally inclusion of some QC data would be helpful (e.g. PCA of the data). Furthermore, viral genome data should be extracted from the RNASeq data to give an indication of infection levels in the relevant samples rather than just relying on a representative graph (Fig.1B).

    Response: Extremely large log₂ fold-change values can arise in RNA-seq analyses when strongly inducible genes are compared to control samples with very low baseline expression. This is common for interferon-stimulated genes (ISGs), which are often undetectable or expressed at near-background levels in mock samples but become highly induced following interferon signalling or viral infection. Similar magnitudes of induction have been reported in transcriptomic studies of interferon responses and SARS-CoV-2 infection of NEC (e.g. Hatton et al., 2021 (PMID: 34876592); Ziegler et al; 2021 (PMID: 34352228); Sharif-Askari et al., (PMID: 36415751) and other.

    To improve clarity, we have revisited the analysis and revised the visualisation of the RNA-seq data. Plotting scales and figure annotations have been adjusted to avoid misleading representation of extreme fold changes. In addition, we have included additional quality-control information for the RNA-seq dataset. Principal component analysis (PCA) of the RNA-seq samples has been added as Supplementary Figure S5 to illustrate sample clustering and replicate consistency, and sequencing depth and quality metrics for all libraries are provided in Supplementary Table S6.

    As suggested by the reviewer, we also quantified viral genome reads directly from the RNA-seq libraries to estimate intracellular viral RNA abundance in the sequenced samples. These data are now presented in Supplementary Figure 1 and discussed in the Results to contextualise infection levels across conditions alongside the extracellular viral RNA measurements shown in Figure 1B.

    Please include virus replication data for all experiments. Only one replication graph is shown (Fig. 1B), but infection level/virus release should be reported for every assay as responses will of course be dependent on how much virus/how many infected cells are present. A difficulty in understanding variant specific host responses in comparative approaches is differences in infection levels. In line with other published work, Fig.1B shows dramatic differences in variant replication. The differences measured at 1hpi indicate issues with input normalisation, this will have a knock-on effect for later replication and ultimately will further increase differences in infected cell counts. L340-342 "These transcriptional shifts occurred despite broadly comparable viral loads across lineages at 24-72 hpi, suggesting that replication level alone does not account for the observed metabolic divergence." - I strongly disagree with this interpretation. The viral loads are clearly not comparable. A 2 log10 difference in virus release is a large difference that will affect the comparison of host response. These replication difference are to be expected and have been previously reported by others. Ancestral variants infect fewer cells compared to Omicron variants. This needs to be acknowledged. In a bulk RNASeq/phopshoproeomic/metabolic measurement the number of infected and uninfected bystander cells across variants will inevitably result in the identification of at least some host responses that correlate with infection levels rather than with specific biology exploited by a variant. The authors must acknowledge this and discuss the contribution of infected vs bystander cells.

    Response: We thank the reviewer for this important point. All downstream analyses in this study (RNA-seq, phosphoproteomics and amino acid profiling) were performed on matched cultures from the same infection experiment; therefore, the replication kinetics shown in Figure 1B represent the infection conditions for all assays. We clarified this explicitly in the Methods and figure legends.

    We agree that differences in viral replication across variants are important when interpreting host responses. The statement suggesting broadly comparable viral loads was removed. We also included PFU/ml measurements and quantify viral reads extracted from the RNA-seq libraries to provide additional estimates of infection burden. Finally, we expanded the Discussion to acknowledge that bulk omics measurements reflect a mixture of infected and bystander cells and that some observed host responses may partly correlate with differences in infection levels across variants.

    Include individual data points to show the spread of the data overall (Fig 6A). Just showing the mean without an indication of how many measurements were taken and the variation in the data makes it hard for the reader to interpret the data.

    Response: We agree that the variability across replicates should be indicated. We added individual data points and error bars to Figure 6A (below) and clarified the replicate structure in the figure legend and Methods. Amino acid measurements were performed using four biological replicates per condition, each processed in duplicate technical measurements that were averaged prior to statistical analysis.

    The choice of 24h for the amino acid abundance analysis needs some further justification. At 24h, some variants will only have infected very few cells. What would this mean for a bulk measurement? Do the authors suggest that there were changes to aa-metabolism in uninfected bystander cells? Would true differences in aa-metabolism in the infected cells be masked by the surrounding uninfected cells?

    Response: We thank the reviewer for this important point. We selected 24 hpi to capture early metabolic responses, in parallel with the phosphoproteomic analyses, before the later-stage transcriptional divergence became dominant. We agree, however, that at this timepoint the amino acid measurements likely reflect the integrated state of both infected and bystander cells within the cultures. We clarified this explicitly in the revised manuscript and discussed this as an important limitation of the bulk metabolite measurements.

    The framing of Alpha and Beta as pre-Omicron is confusing. IC19 and Delta are both equally pre-Omicron variants. Please consider rewording.

    Response: We agree that this terminology is potentially confusing. In the revised manuscript, we used more precise lineage descriptions throughout, distinguishing IC19 as the reference/early strain, Alpha and Beta as earlier VOCs, Delta as a separate later pre-Omicron VOC, and BA.1/BA.5 as Omicron subvariants.

    The Venn diagram in Fig. 2B/C is hard to interpret. How were the percentages calculated? From the total number of DEG across all variants? If so, this would inflate the proportion attributed to the conditions that showed the largest number of DEG genes and shrink the proportion for the conditions with less signal. An UpSet plot might be a better choice to represent the data.

    *Response: *We thank the reviewer for this helpful suggestion. The overlap values in Fig. 2B-C were generated using __InteractiVenn____, __which calculates set intersections and reports them as percentages relative to the total union of differentially expressed genes across all variants at the respective time point. We clarified this explicitly in the figure legend and Methods. We agree that multi-set Venn diagrams can be difficult to interpret when DEG set sizes differ substantially, and we revised the figure legends and associated text to improve clarity of presentation.

    The interpretation of the data as presented requires more mechanistic validation. As it stands, activation of metabolic pathways, or the contribution of the observed phospho changes to variant biology, is not functionally linked to infection outcome. In the absence of more experimental data, the conclusions should be toned down. (For example L330-332 "These patterns suggest that Omicron can replicate despite ongoing cytokine signalling, whereas Delta infection favours stress- and growth-linked pathways to sustain replication.")

    Response: We agree and substantially toned down these statements. The revised manuscript presents these data as comparative host-response signatures rather than mechanistically validated pathways driving infection outcome.

    L440-442 "Similarly, replicate-level variability and confidence intervals for NES values were not plotted, as the scores reflect ranked enrichment rather than absolute expression magnitude." - What do the authors mean by replicate-level variability? I assume the NES was calculated based on fold change which are not replicate-level?

    Response: We thank the reviewer for pointing out this lack of clarity. The previous wording referring to "replicate-level variability" was removed. We now clarify that NES values were calculated from ranked differential expression outputs, with nominal p-values estimated by permutation and FDR-(adjusted p-values) reported in Supplementary Table S2, together with leading-edge genes for each pathway, variant and time point.

    Differences in Oxphos have been reported by others (https://www.sciencedirect.com/science/article/pii/S2589004224012343* ). This study and others should be included in the discussion.*

    Response: We thank the reviewer for highlighting this study. We have now included this in the Discussion to place our findings in the context of previous studies of SARS-CoV-2 infection in nasal epithelial cultures.

    Can the authors speculate whether the innate immune response observed links to the metabolic changes reported?

    Response: While our study does not directly establish a causal link between innate immune activation and metabolic rewiring, interferon signalling is known to influence cellular metabolism during viral infection. In our dataset, IFNα-treated cultures showed strong ISG induction but minimal enrichment of the metabolic pathways analysed here (new Supplementary Figure 4), suggesting that interferon signalling alone does not fully account for the metabolic signatures observed during SARS-CoV-2 infection. These observations support the idea that the metabolic changes detected likely reflect a combination of viral replication demands and host antiviral signalling rather than interferon activation alone. We have added a brief clarification in the Discussion to acknowledge this relationship.

    Overall, in the discussion the data should be contextualised with results from other studies. Particularly work focussing on primary airway epithelial cells and variant infections.

    Response: We agree and expanded the Discussion to better contextualise our results within the existing literature, particularly studies in primary airway epithelial models.

    Please provide more detail on how the merged NES was calculated for Alpha/Beta and BA.1/BA.5. For Fig. 4, either the merged or the unmerged NES data would be sufficient, rather than including both analyses. Enrichment of pathways would benefit from indicating which genes associated have been detected and how they functionally might contribute.

    Response: We thank the reviewer for this helpful suggestion. In the revised manuscript, we simplified the presentation of the pathway enrichment analysis by focusing primarily on the variant-level NES profiles (Figure 4A), while retaining the grouped lineage visualisation (Figure 4B) only as a simplified overview. The merged NES values were calculated by averaging the normalised enrichment scores of the corresponding variants within each lineage group (Alpha/Beta for pre-Omicron and BA.1/BA.5 for Omicron). To improve interpretability, we now report the leading-edge genes contributing to each enriched pathway in Supplementary Table S2.

    Please include how ISGs were defined for the analysis of Fig. 3.

    Response: In our analysis, interferon-stimulated genes (ISGs) were defined based on the transcriptional response of nasal epithelial cells to IFN-α stimulation, which served as a benchmark condition for interferon-responsive gene expression. Genes significantly up-regulated in IFN-α-treated samples relative to mock controls were used to define the ISG set analysed in Fig. 3. We clarified this definition and the selection criteria in the Methods and figure legend.

    Please clarify for each experiment how many replicates/measurements were taken. This information should be included in the figure legend. If data/or measurements were excluded, this should also be highlighted. From the supplementary data (amino acid data; qPCR vs RNASeq) there seems to be variation in the amount of reported measurements (aa-metabolism: 7 vs 8 measurements; RNASeqvsqPCR: 37 vs 39 measurements).

    Response: We agree and have revised the Methods, figure legends, and supplementary information to clarify replicate numbers and any exclusions.

    • 3D: The colour scaling used for log2FC is unbalanced. Consider using different gradings.*

    Response: We thank the reviewer for this suggestion. The colour scale in Fig. 3D was intentionally asymmetric because several IFNα-responsive genes show extremely large log₂fold changes due to very low baseline expression in mock samples. Using a symmetric colour scale would compress the dynamic range of the virus-infected conditions and obscure biologically meaningful differences between variants. To avoid confusion, we clarified the rationale for the colour scaling in the figure legend and ensure that the scale is clearly labelled.

    In the NES analysis, I would expect an indication of the leading edge in the figure or in the supplementary data.

    Response: We agree. Relevant leading-edge information has been added to the supplementary table S2 and is referenced in the revised Methods and Results.

    Several figures would benefit from inclusion of p-values/indication of significance (Fig. 3D, 5B, 6A/C).

    Response: We agree and we have added statistical information where appropriate including the supplementary material.

    Fig .6D requires some more explanation as to what it is shown in the figure. Statistics should be included to confirm that there are no overall differences between conditions.

    Response: We agree and expanded the explanation of Figure 6D. Amino acid levels were normalised to the total intracellular amino acid pool within each condition to evaluate proportional composition independent of total abundance. We also included statistical analysis of the normalised amino acid proportions using a Friedman test, which detected modest but significant differences across conditions (χ² = 15.33, p = 0.004). These differences reflect small shifts in a limited number of amino acids rather than major changes in overall amino acid composition. The statistical analysis and clarification have been added to the Results, Figure 6 legend, and Supplementary Table S4.

    L266-269 "All amino acid measurements were expressed as nmol per 10⁶ viable, counted cells, and viability at 24 hpi was comparable across conditions, indicating that differences in abundance reflect infection-driven metabolic changes rather than variation in cell number." - Data should be included.

    Response: We thank the reviewer for this comment. In differentiated air-liquid interface nasal epithelial cultures, cells form a structured epithelium attached to the transwell membrane and cannot be routinely counted without dissociation of the insert, which would disrupt the culture and preclude subsequent metabolic analysis. For this reason, individual experimental inserts used for amino acid measurements were not dissociated. Instead, representative inserts were dissociated to verify epithelial cell numbers. Dissociation of one mock control and one IC19-infected insert yielded comparable counts of 0.9-1.10 × 10⁶ epithelial cells per insert, confirming that each transwell contains approximately 10⁶ epithelial cells. Amino acid measurements were therefore normalised to epithelial input and reported as nmol per 10⁶ cell equivalents. The manuscript text was revised to clarify this normalisation and avoid implying routine viable cell counting of each insert.

    L270-273: "The variant-amino acid interaction network (Figure 6B) visualises these differences by linking each variant to its most strongly altered amino acids. Edge width reflects the absolute log₂* fold change, and colour indicates direction (red for increases, blue for decreases relative to mock)." The network figure does not add any additional information that is not already contained in Fig. 6C. Consider removing this panel.*

    Response: We thank the reviewer for this suggestion. While the quantitative differences in amino acid abundance are shown in Fig. 6C, the network representation in Fig. 6B was included to highlight variant-metabolite relationships and to visualise which amino acids show the strongest associations with individual viral lineages. This representation facilitates comparison of the pattern of metabolic alterations across variants rather than only their magnitude. To avoid redundancy, we clarified this purpose in the figure legend and streamline the figure presentation.

    For Fig. 6C: The colour scale for the legend is imbalanced starting at -1 with a mid point at 0 and the max at 0.35.

    Response: We thank the reviewer for noting this point. The colour scale reflects the observed range of log₂ fold changes in the dataset, where decreases in amino acid abundance were larger in magnitude than increases. As a result, the scale is asymmetric. To avoid confusion, we clarified the colour scale in the figure legend and ensure that it is clearly labelled to reflect the underlying data distribution.

    L428-41 "Dot colour indicated the direction of regulation (red, up-regulated; blue, down-regulated), and dot size was proportional to the absolute NES value. Vertical reference lines at NES = 0 were included to indicate neutral enrichment." This does not describe the data that is presented in the figure.

    Response: We thank the reviewer for pointing out that the previous description did not accurately reflect the graphical representation. Figure 4 has been revised to clarify how pathway enrichment is displayed. Dot position now represents the normalised enrichment score (NES) on a common scale across all panels, and dot colour indicates the direction of enrichment (red = positive enrichment, blue = negative enrichment). A shaded central region highlights limited enrichment around NES = 0, and the scale at the bottom indicates thresholds used to categorise moderate and strong enrichment. The figure legend and Methods description have been updated accordingly.

    Reviewer #3 (Significance (Required)):

    My criticisms are in part outlined above. While the central question of the study is important and timely, the data reported is largely incremental and lacks mechanistic insight.

    Response: We thank the reviewer for this candid assessment. We agree that the present study is not a mechanistic dissection of individual pathways, but rather a comparative systems-level analysis of lineage-associated host-response patterns in a physiologically relevant NEC model. We have revised the manuscript to better reflect this scope and to avoid overstatement of mechanistic inference.

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    The manuscript entitled "Evolutionary rewiring of host metabolism and interferon signalling by SARS-CoV-2 variants" by Somji and colleagues sets out to understand SARS-COV-2 variant biology in primary nasal epithelial cells. Understanding this and differences in variant-specific host-virus interactions is essential to understand the molecular basis of replication advantages and enhanced transmission that ultimately lead to variant dominance. The authors employ global transcriptomic, phosphor-proteomic and amino acid metabolism assays with the aim to identify variant-specific changes to cell metabolism and innate immune activation in a comparative systems-level approach. Importantly, this work is performed in primary nasal epithelial cells. It is essential to understand variant biology in the context of relevant primary cell infection models and NECs are a great choice to address the proposed research question.

    The work is conceptually interesting, but largely descriptive. While this can still be useful for the field, it requires appropriate framing of the interpretations of the data. I agree with the authors that there will be virus- specific signatures that will contribute to variant fitness, but this dataset makes it hard to draw strong conclusions. The main problem with the manuscript and the interpretation are dramatic differences in viral replication. While some of the conclusions are tantalising and would warrant further investigation, I would expect to see some experimental validation to substantiate the interpretation. In the absence of experimental validations and mechanism, the conclusions should be stated as such and contextualised more with previously published work.

    Major:

    1. A major concern that I have is the analysis of the RNASeq data. Experimental design, analysis and presented data require some clarification: Too little experimental detail for the RNASeq data is given. How many replicates were sequenced/analysed? The figure legend state three independent experiments - but how many individual replicate transwells per condition (and NEC batch) were used? This information needs to be included in the manuscript. Generally, clarification on how many replicates were used per experiment needs to be included in the figure legends for all data panels.

    2. The reported log2 fold changes are beyond what is biologically reasonable. A log2 fold change of 120 or even 30 (Fig.3D, suppl table) indicates issues with the data analysis. It is worth revisiting the analysis and additionally inclusion of some QC data would be helpful (e.g. PCA of the data). Furthermore, viral genome data should be extracted from the RNASeq data to give an indication of infection levels in the relevant samples rather than just relying on a representative graph (Fig.1B).

    3. Please include virus replication data for all experiments. Only one replication graph is shown (Fig. 1B), but infection level/virus release should be reported for every assay as responses will of course be dependent on how much virus/how many infected cells are present. A difficulty in understanding variant specific host responses in comparative approaches is differences in infection levels. In line with other published work, Fig.1B shows dramatic differences in variant replication. The differences measured at 1hpi indicate issues with input normalisation, this will have a knock-on effect for later replication and ultimately will further increase differences in infected cell counts. L340-342 "These transcriptional shifts occurred despite broadly comparable viral loads across lineages at 24-72 hpi, suggesting that replication level alone does not account for the observed metabolic divergence." - I strongly disagree with this interpretation. The viral loads are clearly not comparable. A 2 log10 difference in virus release is a large difference that will affect the comparison of host response. These replication difference are to be expected and have been previously reported by others. Ancestral variants infect fewer cells compared to Omicron variants. This needs to be acknowledged. In a bulk RNASeq/phopshoproeomic/metabolic measurement the number of infected and uninfected bystander cells across variants will inevitably result in the identification of at least some host responses that correlate with infection levels rather than with specific biology exploited by a variant. The authors must acknowledge this and discuss the contribution of infected vs bystander cells.

    4. Include individual data points to show the spread of the data overall (Fig 6A). Just showing the mean without an indication of how many measurements were taken and the variation in the data makes it hard for the reader to interpret the data.

    5. The choice of 24h for the amino acid abundance analysis needs some further justification. At 24h, some variants will only have infected very few cells. What would this mean for a bulk measurement? Do the authors suggest that there were changes to aa-metabolism in uninfected bystander cells? Would true differences in aa-metabolism in the infected cells be masked by the surrounding uninfected cells?

    6. The framing of Alpha and Beta as pre-Omicron is confusing. IC19 and Delta are both equally pre-Omicron variants. Please consider rewording.

    7. The Venn diagram in Fig. 2B/C is hard to interpret. How were the percentages calculated? From the total number of DEG across all variants? If so, this would inflate the proportion attributed to the conditions that showed the largest number of DEG genes and shrink the proportion for the conditions with less signal. An UpSet plot might be a better choice to represent the data.

    8. The interpretation of the data as presented requires more mechanistic validation. As it stands, activation of metabolic pathways, or the contribution of the observed phospho changes to variant biology, is not functionally linked to infection outcome. In the absence of more experimental data, the conclusions should be toned down. (For example L330-332 "These patterns suggest that Omicron can replicate despite ongoing cytokine signalling, whereas Delta infection favours stress- and growth-linked pathways to sustain replication.")

    9. L440-442 "Similarly, replicate-level variability and confidence intervals for NES values were not plotted, as the scores reflect ranked enrichment rather than absolute expression magnitude." - What do the authors mean by replicate-level variability? I assume the NES was calculated based on fold change which are not replicate-level?

    10. Differences in Oxphos have been reported by others (https://www.sciencedirect.com/science/article/pii/S2589004224012343 ). This study and others should be included in the discussion.

    11. Can the authors speculate whether the innate immune response observed links to the metabolic changes reported?

    Minor:

    1. Overall, in the discussion the data should be contextualised with results from other studies. Particularly work focussing on primary airway epithelial cells and variant infections.

    2. Please provide more detail on how the merged NES was calculated for Alpha/Beta and BA.1/BA.5. For Fig. 4, either the merged or the unmerged NES data would be sufficient, rather than including both analyses. Enrichment of pathways would benefit from indicating which genes associated have been detected and how they functionally might contribute.

    3. Please include how ISGs were defined for the analysis of Fig. 3.

    4. Please clarify for each experiment how many replicates/measurements were taken. This information should be included in the figure legend. If data/or measurements were excluded, this should also be highlighted. From the supplementary data (amino acid data; qPCR vs RNASeq) there seems to be variation in the amount of reported measurements (aa-metabolism: 7 vs 8 measurements; RNASeqvsqPCR: 37 vs 39 measurements).

    5. Fig. 3D: The colour scaling used for log2FC is unbalanced. Consider using different gradings.

    6. In the NES analysis, I would expect an indication of the leading edge in the figure or in the supplementary data.

    7. Several figures would benefit from inclusion of p-values/indication of significance (Fig. 3D, 5B, 6A/C).

    8. Fig .6D requires some more explanation as to what it is shown in the figure. Statistics should be included to confirm that there are no overall differences between conditions.

    9. L266-269 "All amino acid measurements were expressed as nmol per 10⁶ viable, counted cells, and viability at 24 hpi was comparable across conditions, indicating that differences in abundance reflect infection-driven metabolic changes rather than variation in cell number." - Data should be included.

    10. L270-273: "The variant-amino acid interaction network (Figure 6B) visualises these differences by linking each variant to its most strongly altered amino acids. Edge width reflects the absolute log₂ fold change, and colour indicates direction (red for increases, blue for decreases relative to mock)." The network figure does not add any additional information that is not already contained in Fig. 6C. Consider removing this panel.

    11. For Fig. 6C: The colour scale for the legend is imbalanced starting at -1 with a mid point at 0 and the max at 0.35.

    12. L428-41 "Dot colour indicated the direction of regulation (red, up-regulated; blue, down-regulated), and dot size was proportional to the absolute NES value. Vertical reference lines at NES = 0 were included to indicate neutral enrichment." This does not describe the data that is presented in the figure.

    Significance

    My criticisms are in part outlined above. While the central question of the study is important and timely, the data reported is largely incremental and lacks mechanistic insight.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    The authors of the manuscript entitled "Evolutionary rewiring of host metabolism and interferon signalling by SARS-CoV-2 variants" investigated the diversity of different SARS-CoV-2 isolates regarding gene expression, kinase activity and amino acid profiles in infected primary human nasal epithelial cells. Somji et al. found certain distinct alterations of measured factors after infections compared to mock and differences in cells infected with the mentioned different SARS-CoV-2 isolates.

    The topic of the manuscript as such is of high importance since understanding virus host interactions in general and virus host coevolution particularly on the level of cellular metabolism and beyond comes with great potential in deeper understanding the infection biology of viral invaders.

    Nevertheless, the study needs to be enlarged and further defined, the experimental set up has to be improved and the drawn conclusions have to be proven by experiments. The presentation of the obtained data needs to be improved, checked and carefully chosen to allow the reader to follow the article in a much more guided way. At this stage of experimental data depth, presentation and interpretation, there is room for certain overinterpretations of the biological meanings of the presented data.

    Please find a detailed list of comments for the consideration of the authors below.

    1. The authors state about virus growth kinetics in Fig.1. To be able to do so in full extend, virus particle counts (PFU/ml) need to be measured and included in this data set.

    2. From Fig.2 on, the presentation and introduction of the data set is often very hard to follow. Certain panel labeling is not correct e.g. in Figure 2, Figure 2A is not introduced, 72h data are linked to Figure 2C but 2C is a Venn diagram of 24h gene expression downregulation. The Venn diagrams are not mentioned in the text at all. This problem is occurring at different occasion, which makes it hard to impossible to follow the experimental flow of the study. Therefore, a complete revision of the data presentation within the figures and the linked text is needed. Further example, lines 213 and 224, Figure 4B two times mentioned with different data supposed to be shown in Fig. 4B.

    3. The authors are inconsistent with including statistics in their figures. Please include all statistics in your figures to allow the reader to get this information. Please declare how often and how each experimental set has been done and clarify e.g. in the figure legends. In addition, please improve the figure quality for better allowance of cross comparability of data sets. As example, used the same x-axis scale for all graphs in Fig 4.

    4. The authors create claims about metabolic profiles without measuring deeper metabolic circumstances. Why are only amino acids measured and not metabolite concentrations in general. Metabolic gene expressions as measurement of metabolic pathway activities can be strongly misleading since gene expression per definition does not necessarily mean enzyme activity, which of course is finally important for pathway activity as well.

    5. The authors need to carefully crosslink the obtained data sets. As an easy example, how much of the found differences in gene expression, pathway activities etc. is due to viral growth differences. With other words, are there regulatory differences or are the differences seen due to different growth kinetics. Are ISG expression level linked to virus growth? These type of questions not be asked and correlations need to done by the authors to guide the reader through all those assays conducted in this study.

    Referee cross-commenting

    I do fully agree with reviewer 1 and 3 in terms of the importance of much more comprehensive data on virus growth. Measurement of real virus progeny (PFU/ml) and viral protein and RNA expression is needed to state about the importance of altering viral dynamics for interpreting the findings.

    I do fully agree with reviewer 1 and 3 that data analysis, presentation and interpretation has to be improved. Information such as how often has each experiment been done and how has the experimental set up been constructed has to be clarify e.g. in the figure legends.

    As reviewer 1 mentioned, direct analysis of metabolite concentrations is needed to be able to judge about metabolic changes driven by the different SARS CoV-2 variants.

    In line with both, reviewer 1 and 3, conclusions drawn by the authors should be toned down. More data and improved data analysis and presentation are needed to foster the conclusions drawn .

    Significance

    While the topic as such is interesting and hoighly relevant, the manuscript has several major flaws, both with regard to paper organisation and content. In the current state it is hard to judge, whether the data are of significance.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary:

    In the submitted work, authors seek to understand the transcriptional and metabolic changes induced by different variants of SARS-CoV-2 infection. They employ a model of pooled, primary nasal epithelial cells (NEC) differentiated within an air-liquid-interface. Subsequently, cultures are infected with isolates representing key variants of SAR-CoV-2 from initial D614G, Alpha, Beta, Delta, and Omicron. Following initial characterization, authors compare transcriptional changes at 24 and 72 hours post infection. Analysis focuses on differentially expressed genes, upregulated Interferon Stimulated genes, and transcripts within known metabolic pathways. Subsequently, infected cultures are also analyzed by phosphoproteomic analysis to identify changes in cell signaling and measured for amino acid content. Throughout, changes in each profile are correlated with different variants of SARS-CoV-2, with Delta and Omicron revealing particular diametrically opposed changes. There are reasonable controls applied, including the use of IFNalpha treatment to "benchmark" ISG production. Overall, authors create a picture that Omicron infections do not suppress IFN signaling as efficiently as Delta variants and further exhibit limited hallmarks of cell stress and metabolic dysregulation.

    This is a remarkable study that attempts to cross-correlate multiple -omics analyses of cell responses to characterize differences in infection. It is very well written and the data is exemplary. I do have some concerns related to the placement and emphasis of interpretation in the results section that need to be revised. Beyond my stylistic concern, the interpretation of the experimental changes between variants are compromised by the failure to analyze the extent of infection within the NEC model. Using an MOI of 0.01 will produce a dramatically heterogeneous extent of infection at both 24 and 72 hours post infection that will also depend on the extent of viral transmission within the culture. The limited analysis of secreted E-gene detection is insufficient to overcome the inherent unequal comparison of cell responses between variants. There are ways to assuage, but not eliminate, this problem when it comes to comparing and interpreting experimental results. My concerns and suggestions are detailed in the concerns below.

    Major Concerns:

    1. Heterogeneous extent of infection. The MOI of 0.01 used to initiate infection is extraordinarily low for the types of analysis that is employed with the NEC culture. The interpretation of the data does not take into account that there will be infected and uninfected cells, of varying extents, making up the changes observed. Further, the variants likely have differing abilities to spread through the NEC culture, complicating both interpretation of changes and comparison between variants. At a minimum, authors need to evaluate the extent of SARS-CoV-2 infection through either flow cytometry or immunofluorescence analysis against viral protein(s). It is possible that Omicron, while secreted well, has more limited transmission allowing for more cells to mount an IFN response. Delta is a prolifically spreading virus that likely has more extensive infection at 72 hpi than the other variants. These statements are conjecture and highlight how such differences could alter the interpretation of the subsequent experiments.

    2. Further evaluation of IFNalpha treated cells. The paper emphasizes the ISG analysis, but the IFN treated cells should be included in the DEG and metabolic pathway analysis. IFN treatment is known to alter metabolic changes in cells, and it would be valuable to see those changes reflected in your analysis. Consider the evidence presented in the following:

    Fritsch SD, Weichhart T. Effects of Interferons and Viruses on Metabolism. Front Immunol. 2016 Dec 21;7:630.

    Heer CD, Sanderson DJ, Voth LS, Alhammad YMO, Schmidt MS, Trammell SAJ, Perlman S, Cohen MS, Fehr AR, Brenner C. Coronavirus infection and PARP expression dysregulate the NAD metabolome: An actionable component of innate immunity. J Biol Chem. Elsevier BV; 2020 Dec 25;295(52):17986-17996.

    Palmer CS. Innate metabolic responses against viral infections. Nat Metab. 2022 Oct;4(10):1245-1259

    Further, It is possible that the changes attributed to Omicron are quite similar to the effects of the IFN treatment, given the extensive ISG detection. The same is true for the phosphor-proteomic analysis and amino acid content. I also have concerns that using a treatment of IFNalpha that impacts all cells as a benchmark for heterogeneous infection is not truly comparable. How was the concentration of IFN chosen? What was the extent of IFN activation in the culture?

    1. Further correlation of transcriptional changes with metabolic changes - While many published works emphasize transcriptional changes as a proxy for metabolic changes, there are robust methods that can be applied to directly analyze metabolite content and changes in the context of viral infection. In particular these studies should be assessed and compared for the interpretation of the presented results:

    Kramaric, T., Thein, O.S., Parekh, D. et al. SARS-CoV2 variants differentially impact on the plasma metabolome. Metabolomics 21, 50 (2025).

    Loveday EK, Welhaven H, Erdogan AE, Hain KS, Domanico LF, Chang CB, June RK, Taylor MP. Starve a cold or feed a fever? Identifying cellular metabolic changes following infection and exposure to SARS-CoV-2. PLoS One 2025 Feb 12;20(2):e0305065.

    Irún P, Gracia R, Piazuelo E, Pardo J, Morte E, Paño JR, Boza J, Carrera-Lasfuentes P, Higuera GA, Lanas A. Serum lipid mediator profiles in COVID-19 patients and lung disease severity: a pilot study. Sci Rep. 2023 Apr 20;13(1):6497.

    Luke Whiley, Nathan G. Lawler, Annie Xu Zeng, Alex Lee, Sung-Tong Chin, Maider Bizkarguenaga, Chiara Bruzzone, Nieves Embade, Julien Wist, Elaine Holmes, Oscar Millet, Jeremy K. Nicholson, and Nicola Gray, "Cross-Validation of Metabolic Phenotypes in SARS-CoV-2 Infected Subpopulations Using Targeted Liquid Chromatography-Mass Spectrometry (LC-MS)", Journal of Proteome Research 2024 23 (4), 1313-1327

    1. Editing to limit interpretation within experimental results. I appreciate that this is a stylistic concern and it is an issue in the paper. Statements in the results are often over-reaching. Some examples include: Line 156 -"suggesting attenuated or delayed early sensing" - The Low MOI and time leaves these results open to various explanations. Better to just state and move on.

    Line 157 "Delta drove the most extensive" - drove has a lot of assumption. "produced" "resulted in " or something more passive is more appropriate

    Line 179 "pointing to sustained suppression of interferon responses." - sustained is a leading interpretation. Effective? Comprehensive? again, the MOI is complicating interpretations of global transcript changes.

    Line 186 "suggesting a weaker activation of interferon signaling" Too much leading interpretation here. You detect fewer ISGs that are differentially regulated. Could be for many reasons.

    Minor Concerns:

    1. Line 72 "has evolved unique strategies" Unique can be easily misconstrued to mean different mechanisms. More likely, it is a subtle balance between promotion of viral replication and suppression of IFN responses.

    2. Line 126 - 128 "NECs were derived from three commercially available donor pools". The following text doesn't make it clear that they are the same produce from different lots. The methods clarify somewhat, but should be clarified for transparency.

    3. Line 129 "Viral replication kinetics" Need to highlight that this is detection of secreted viral genomes. which is a proxy measure for replication and dissemination in the culture. Direct measurement of the extent of infection is not being made nor can be interpreted.

    4. Line 149 "Differentially expressed genes (DEGs)" What is the comparison group? The figure legend/design suggests that IFNa treatment. Is there a matched uninfected control for each timepoint as well? Later experiments specify the comparison group. Text should be clarified here for transparency.

    5. Line 224 and Figure 4B - I don't see the value of the "merged NES" values given these are only aggregate of the Pre-Omicron and Omicron species. If you had compared multiple D614G and Delta variants, then there would be utility.

    6. Line 261 "quantified at 24 hpi" Why this timepoint? Changes were minor and not representative to extensive infection.

    7. Line 268 "rather than variation in cell number." I appreciate the rigor and control of experimentation. And how many of those cells are infected? That is not controlled.

    8. Line 428-429 "direction of regulation" This seems like an over-interpretation of the data. You have performed pathway analysis based on the quantity of RNA transcription detected in sequencing then imputing an interpretation of regulation. Without pulse labeling of metabolic standards or kinetic analysis of metabolite quantity, it is difficult to assert regulatory direction.

    Referee cross-commenting

    I am in agreement with the comments and suggestions of Reviewer #2 and #3. In particular, the comment of Reviewer #3 to estimate viral replication from the RNASeq data is quite valuable to begin addressing some of the concerns about the extent of viral replication. It does not negate the need to further assess productive viral titer (PFU/mL) or the extent of viral infection (immunofluorescence or flow cytometry).

    I also agree with Reviewer #3 regarding the extent of mechanistic interpretation that can be drawn from the current study. This concern can largely be addressed through revision of the text and a tempering of the interpretations that are drawn.

    I also agree and appreciate the detailed analysis of reviewer #2 regarding the inconsistencies between the text and the figures. It is critically important to be consistent in the data and presentation of these complex experiments. Resolving these issues will only strengthen the work.

    Significance

    • The work detailed in this manuscript is takes a very broad approach to identify differences in the effects of SARS-CoV-2 variant infections. Elements of this work have been published, including transcriptomics, metabolomics, and phosphoproteomics. This work is significant in that multiple variants are evaluated with comparable methods in the very relevant human nasal epithelial cell model. The use of this model, and the direct integration of multiple -omics, sets this work apart from previously published studies. This cross-omic analysis, with the IFN-treated controls, provides a robust foundation of data that can be used to detail the differences in the response to the SARS-CoV-2 variant infections.

    • That said, a significant limitation to the study was the low MOI used to initiate infection and the lack of detailed analysis infection progression of the different variants. Further, there is limited comparison of the IFN-treatment condition in relation to the transcriptional changes, and no inclusion of IFN-controls in the other methods. Both of these limitations undercut the potential significance of the paper and its findings.

    • Audience: This work will have be important to bench researchers interested in further characterizing and comparing the effects of SARS-CoV-2 infection. Potentially, clinicians involved in diagnostics will find utility in the study of changes for potential biomarker analysis for severe COVID19 disease.

    • My expertise is the field of virology, having studying multiple RNA and DNA viruses, including SARS-CoV-2, to understand virus-cell interactions. My focus includes primary cell culture models of infection, proteomic and metabolic analysis of infection induced changes, and monitoring the spread of viral infection through direct and indirect measurements.

  5. Version published to 10.1101/2025.11.15.688605 on bioRxiv