Preprint Multiomic and Spatial Profiling of Colorectal Tissue Reveals Viral Persistence and Immune Dysregulation in Long COVID, 2026, LaFranchi et al.

[Chandelier]

Multiomic and Spatial Profiling of Colorectal Tissue Reveals Viral Persistence and Immune Dysregulation in Long COVID

Spoiler: Authors

LaFranchi, Brian; Maison, David P.; Vinden, Joanna; Rodriguez, Antonio E.; Tout, Ali; Grimbert, Lilian; Velazquez, Emanuel; Vudali, Urmila; Poblano, Belen Altamirano; Dalhuisen, Thomas; Cattle, Javier; Figueroa, Tony; Fudotan, Yuki; Luna, Michael A.; Ryder, Dylan; Deswal, Monika; Abel, Brent S.; Lynch, James; Lipford, Amanda; Razi, Nabeel; Steifman, Carly B.; McCann, Henry N.; Kataria, Nick; Girling, Valerie; Thomas, Reuben; Wang, Chao; Deitchman, Amelia N.; Patel, Shreya; Traglia, Michela; Tseng, Zian H.; Szabo, Gyula; Laszik, Zoltan; Farrow, Alex; Sumimoto, Nanami; Servellita, Venice; Hoh, Rebecca; Fehrman, Emily A.; Kelly, J. Daniel; Martin, Jeffrey N.; Deeks, Steven G.; Chiu, Charles Y.; Somsouk, Ma; Peluso, Michael J.; Henrich, Timothy J.

Abstract
Long COVID (LC) - a chronic condition characterized by persistent, debilitating symptoms following SARS-CoV-2 infection - has emerged as a major public health challenge.
Although many interrelated mechanisms have been proposed as drivers of LC, the root causes have yet to be identified, posing significant challenges for therapeutic development.
While many blood-based studies have been conducted, they have not yielded conclusive mechanistic insights into LC pathogenesis.
Attention has therefore turned toward direct tissue investigation, with the gastrointestinal (GI) tract becoming a major focus due to evidence that virus or viral components can persist at this site for months to years following an episode of COVID-19.

Here, we performed a high-dimensional characterization of colorectal tissue and peripheral blood in a highly characterized cohort of 44 people with LC and 13 recovered controls.
We profiled SARS-CoV-2 persistence, host immune responses, and tissue inflammation using bulk and single-cell RNA sequencing, nCounter RNA probe hybridization, quantitative PCR, metagenomic next-generation sequencing, plasma proteomics, high-dimensional spectral flow cytometry, in situ-hybridization/immunohistochemistry, and single-cell digital spatial omics.

Our results support a model in which LC is driven by long-term immune dysregulation and perturbations of the regulatory gut immune environment which imply ongoing viral persistence, although direct viral detection was only observed in a subset of participants.
Specifically, we identify a tissue-based transcriptional environment in which SARS-CoV-2 activates innate myeloid immune signaling, driving chronic inflammation while simultaneously downregulating pathways responsible for immune-mediated clearance of infected cells, including antigen presentation, phagocytosis, cytotoxic immune cell trafficking, and granzyme production.
Importantly, signatures in peripheral blood are considerably weaker than those observed in tissue.

Together, these findings provide a direct biological rationale for therapeutic strategies in LC aimed at enhancing or redirecting cytotoxic immune function to overcome immune dysregulation and clear persistent viral reservoirs.

Web | DOI | bioRxiv | Open Access

 

[Jonathan Edwards]

An abstract full of interpretation but no data, again.

 

[V.R.T.]

Specifically, we identify a tissue-based transcriptional environment in which SARS-CoV-2 activates innate myeloid immune signaling,

Innate immunity...

Importantly, signatures in peripheral blood are considerably weaker than those observed in tissue

...And signal stronger in tissue than blood
Relevant to some of the thinking on here lately?

To assess the host immune response in LC compared to recovered participants, we performed differential221
expression analyses to a complementary set of sequencing-based, proteomic, and spatial methods to both222
colorectal tissue and paired PBMCs and plasma. In addition to the targeted transcriptomic panel described223
above (Ncounter), we employed (1) unbiased transcriptomic sequencing of colorectal tissue; (2) proteomic224
profiling of plasma samples using the Olink Reveal platform; (3) combined 5’ single-cell (sc)RNAseq with T-225
Cell Receptor (TCR)seq on a subset of 13 colorectal tissue samples; (4) high-dimensional 27-plex spectral226
flow cytometry on cryopreserved colorectal tissue; and (5) single cell spatial targeted transcriptomics227
(Xenium) on a subset of 6 FFPE colorectal tissue sections.

Plasma proteomics. We performed deep proteomic profiling of immune activation and inflammation using291
the Olink Reveal platform32, which simultaneously tests 1,034 plasma proteins, in LC and recovered292
controls. We did not observe significant differences in protein expression between the overall LC and293
recovered cohorts; when examining specific phenotypes, we observed only one individual protein294
(Cathepsin H) upregulated in people with LC experiencing ongoing smell/taste disturbance compared to295
recovered individuals (Fig S3). Interestingly, we observed downregulation of IFNα and IFNγ Hallmark296
signaling pathways in the subset of people with LC experiencing PEM compared to recovered controls (Fig 2f)

Targeted transcriptome analysis. We characterized host transcriptional responses in RNA isolated from237
stabilized colorectal biopsy tissue and paired PBMCs, comparing participants with LC and recovered238
controls using two complementary approaches. First, we applied our custom nCounter assay. Fig 2a,b239
shows volcano plots, heatmaps of significant differentially expressed genes (DEGs) after false discovery240
rate (FDR) adjustments, and gene set enrichment analyses (GSEA) significance scores. Overall, we241
observed 240 significant DEGs in colorectal tissues when comparing participants with LC to recovered242
controls (complete list of DEGs in Table S3), with tight clustering of LC and recovered samples by gene243
expression patterns. One major recovered cluster and two major LC clusters were visualized. Five244
nanoString-defined gene pathways had robust Global Significance Scores (GSS31; GSS of >3) containing 7245
to 54 individual significant genes in each set were identified (Fig 2c). Interestingly, all the top GSS246
pathways were downregulated overall (as defined by the Directed Global Significance Score; DGSS) and247
included major pathways associated with MHC class I antigen processing, mononuclear cell migration,248
lysosome activity, and phagocytosis. CCL8 and CCL13 were the most strongly downregulated individual249
genes, but others such as CD16a/b (NK/nonclassical monocyte markers), CCL4/eotaxin (eosinophil250
recruitment and bowel inflammation), CCL7, TLR3 (mediates detection of viral double-stranded RNAs),251
granzyme A (cytotoxicity) were also markedly downregulated. In contrast, IL1RN and S100A12, which are252
associated with immune activation and bowel inflammation, were the most highly upregulated genes.253
Overall, the DEG and GSE analyses all point towards a decrease in gene transcriptional activity that is254
required for the immune system to recognize and eliminate virus-infected cells, although some255
inflammatory cytokines and NFkB activity were overexpressed in LC.256
Importantly, the same gene expression analysis in matched PBMCs yielded only 3 significant DEGs, all of257
which were upregulated. These DEGs were not observed among those seen in the gut; Fig 2b, with no258
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 10, 2026.;https://doi.org/10.64898/2026.08.07.743616doi:bioRxiv preprint
distinct clustering by disease phenotype. Furthermore, GSE analyses resulted in no robust pathways (GSS259
scores all <2.03). This suggests that biological differences using these analytic approaches may be more260
easily detectable in the tissues than in the peripheral blood.261

Unbiased transcriptomic analysis of colorectal tissue. Building on the targeted transcriptomic analyses262
using nCounter, we performed unbiased transcriptomic sequencing of colorectal biopsy samples to more263
comprehensively define the nature and extent of host gene responses. We identified 4,757 DEGs between264
participants with LC and recovered controls (n=55 as two LC participants did not have data available; Fig265
2d, Fig S1, Fig S2), with clear separation and clustering between groups. Unsupervised hierarchical266
analysis further identified four distinct transcriptional clusters within the LC group defined by distinct267
patterns of DEGs as determined by K-means clustering (Fig 2d). These transcriptional clusters were not268
clearly associated with clinical phenotypes, including the presence of gastrointestinal symptoms (Fig 2d).269
To further understand the unbiased RNAseq data, we again performed GSEA. We generated directional270
ridge plots to identify the top gene pathways and whether the genes in each pathway, taken together, were271
up or downregulated when comparing LC and recovered groups (Fig 2e, complete list of significant272
pathways in Table S4). Top significant upregulated Hallmark pathways in those with LC included MYC273
targets, MTORC1 and PI3K-AKT-mTOR signaling, TNFα signaling, glycolysis, inflammatory response,274
complement and several pathways involving cell cycling, metabolism and protein/hormonal responses.275
Significantly downregulated Hallmark pathways in those with LC included heme metabolism, adipogenesis,276
bile acid metabolism, and oxidative phosphorylation. Applying Kyoto Encyclopedia of Genes and Genomes277
(KEGG) to the GSEA to better understand various disease processes associated with the DEGs, we278
observed increased pathways including mRNA surveillance, TNF signaling, and EBV infection (among279
others as shown in Fig 2e). Interestingly, the acute COVID-19 pathway was significantly downregulated in280
the LC versus recovered groups.281
To further explore the relationship between SARS-CoV-2 detection and these immunologic changes, an282
additional differential expression analysis was performed comparing those with detectable virus (n=12) vs283
those without detectable virus (n=32) within the LC group. There were no significantly differentially284
expressed genes or enriched pathways, implying that changes in host immune response associated with285
LC occur independently of viral detection. A further comparison of LC participants who were hospitalized286
(n=3) vs those who were not hospitalized (n=41) during acute infection also revealed no significant287findings.288nfindings.

Colorectal single-cell RNA transcriptomics. To better understand the contribution of individual cell types299
within the rectosigmoid samples that may be leading to the bulk-transcriptome analyses, we performed300
combined 5’ single-cell (sc)RNAseq with T-Cell Receptor (TCR)seq on a subset of 13 participants including301
those with LC (n=9) and recovered controls (n=4) (Table S5). Among these 13 participants, we observed302
12 major cell clusters as defined by Seurat including T cells, B cells, macrophages, three distinct plasma303
cell clusters, fibroblasts, endothelial cells, enteroendocrine cells, goblet cells, tuft cells, and colonic transit304
amplifying cells (TACs, which are rapidly dividing differentiating cells that are often in high abundance given305
rapid turnover of gut epithelium) (Fig 3a). Overall, there were no observed differences in the PCA cluster306
morphologies or recovered frequencies of cells within each cluster in LC versus recovered tissue groups307
(all p>0.05). Overall, there were only two significant DEGs within two distinct cell clusters that met308
individual FDR-adjusted statistical significance – ACE was significantly upregulated in intestinal goblet cells309
(1.7 Log2 fold change, Q=0.025) and Immunoglobulin J chain (JCHAIN) was significantly downregulated in310
macrophages (Fig S4). We did observe, however, strongly significant gene ontologies as defined by311
Hallmark and KEGG (Fig 3b shows all significant Hallmark pathways and the top 40 significant KEGG312
pathways; all KEGG pathways are shown in Fig S5).313
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 10, 2026.;https://doi.org/10.64898/2026.08.07.743616doi:bioRxiv preprint
We also identified upregulation of similar pathways in macrophages, which we previously demonstrated314
could harbor SARS-CoV-2 in the gut years after initial infection12, and to a lesser extent fibroblasts, which315
appear to express ACE2 in our IHC experiments (Fig 1e). Interestingly, B and T cells showed very little316
differences in any gene pathway and two of the three plasma cell clusters showed significant317
downregulation of pathways such as lymphocyte signaling pathways, coronavirus disease responses, and318
immunoglobin like cell adhesion. Of note, the absolute number of plasma cells and lymphocytes were319
higher than the number of myeloid immune cells and fibroblasts in colorectal tissue (Fig 3a), which may320
explain the bulk nCounter and RNAseq analyses showing downregulation of many adaptive and cytotoxic321
immune response genes as above.322
In addition, we analyzed the single-cell RNAseq data using the clinically validated SURPI mNGS analysis323
pipeline for viral infection33 but identified no reads that aligned to SARS-CoV-2. SURPI applies stringent324
filtering and taxonomic classification criteria to distinguish true viral sequences from spurious alignments325
and other causes of false-positive viral read detection that can confound other viral sequence alignment326
algorithms19,33–35.327
Together, these findings indicate that LC is associated with specific gene expression reprogramming in328
colorectal tissue, particularly within epithelial, macrophage, fibroblast, and plasma cell populations, despite329
minimal changes in cellular composition and no definitive detection of SARS-CoV-2 reads by single-cell330
RNA sequencing

Flow cytometric analysis of colorectal cells. To characterize immune-cell phenotypes and protein340
expression in colorectal tissue, we performed high-dimensional 27-plex spectral flow cytometry on341
collagenase-digested, thawed cryopreserved tissue from participants with LC and recovered controls,342
including cells isolated from the intraepithelial and laminal propria fractions. Samples were selected based343
on the availability of sufficient bulk cryopreserved tissue, while maximizing overlap with participants who344
were included in the scRNAseq experiments (Table S6). This panel was designed to phenotype γδT cells,345
CD8+ and CD4+ T cell memory subsets (naive, central memory, effector memory, terminally differentiated),346
T cell activation and immune checkpoint [PD-1] markers, IgA- an IgM-producing memory B/plasma cells,347
NK cells (immature cytokine producing, mature cytotoxic), classical and non-classical monocytes, and348
basophils (complete gating strategy outlined in Fig S6-8).349
Participants with LC had a significantly higher percentage of CD14+ classical monocytes in colorectal cells350
than recovered controls (p=0.036). We also observed a non-significant trend towards a higher percentage351
of PD1-expressing CD8+ T cells in the LC group (P=0.076; Fig 3d). We did not observe significant352
differences in the frequencies of memory B/plasma cells expressing IgA or other cell types in the flow353
panel.354

Some of the data from the paper, in case its useful to anyone. Sorry about all the numbers, no spoons to edit them out.

 

[V.R.T.]

Participants with LC had a significantly higher percentage of CD14+ classical monocytes in colorectal cells350
than recovered controls (p=0.036).

@Jonathan Edwards are these kind of T like cells you suggested being involved the other day?

 

[PageofME]

...And signal stronger in tissue than blood
Relevant to some of the thinking on here lately?

Are you referring to the recent discussions on latent virus theory? If you are, keep in mind that SARS persistence must be distinguished from latent virus reactivation, they're two totally different hypotheses investigated by different groups of researchers.

 

[V.R.T.]

Are you referring to the recent discussions on latent virus theory? If you are, keep in mind that SARS persistence must be distinguished from latent virus reactivation, they're two totally different hypotheses investigated by different groups of researchers.

No, I was referring to the discussions about the fact that any immune signalling might be hidden in tissue rather than in the blood, and might involve the innate immune system.

I'm not particularly enthused by covid persistence as a theory. I know you're big on the other kind of latent virus theory - the reactivation. I'm not convinced by that but it's a lot more plausible than covid persistence imo.

 

[PageofME]

No, I was referring to the discussions about the fact that any immune signalling might be hidden in tissue rather than in the blood, and might involve the innate immune system.

I'm not particularly enthused by covid persistence as a theory. I know you're big on the other kind of latent virus theory - the reactivation. I'm not convinced by that but it's a lot more plausible than covid persistence imo.

Got you —and hope you're keeping within your energy limits. Which specific pathomechanism hypothesis were you discussing when that came up? Can you give me a link to that thread?

 

[V.R.T.]

Which specific pathomechanism hypothesis were you discussing when that came up? Can you give me a link to that thread?

Iirc just the general idea of an immune-brain signalling loop. I think it's all over in a few threads.

Hidden in tissue because we can't find much in blood so far, and innate immunity because that might be what BTN2A1/2 is pointing us towards.

 

[PageofME]

Iirc just the general idea of an immune-brain signalling loop. I think it's all over in a few threads.

Hidden in tissue because we can't find much in blood so far, and innate immunity because that might be what BTN2A1/2 is pointing us towards.

Personally, I’m a highly sceptical there—that feels like a pretty speculative model built around a single genetic signal from one team.

 

[fst]

Iirc just the general idea of an immune-brain signalling loop. I think it's all over in a few threads.

Hidden in tissue because we can't find much in blood so far, and innate immunity because that might be what BTN2A1/2 is pointing us towards.

I think this post might be one of them?

Post in thread 'The possibility of autoimmunity or auto-reactivity in ME/CFS'

Jan 16, 2021

I still don't understand what sort of longer-range signalling would be involved such that symptoms appear in other parts of the body? I do realise that all of this is speculative.

Yes, this is the problem for pretty much any theory we have so far as far as I can see. Cross out all the known cytokines and then what?

I guess one possibility is that CD8 T cells directly annoy the nerve endings in lymphoid tissues through short-range cytokine action (membrane bound?) or even granzymes. This might be a regular job they do to signal to central control that virus is about.

They might...

• Jonathan Edwards

• 

I think there is more in the BTN2Ax thread, e.g.,

Post in thread 'Genetics: Chromosome 6 BTN2A2 and BTN3A3 (BTN2A1)'

May 22, 2026

Thanks Jonathan this is all really useful. Over recent months I’ve been reading some papers about memory or memory like NK cells, which after exposure to cytokines can hang around for a long time with lower thresholds for activation.

That is something have not caught up with. It would not be surprising. These would not carry memory in the sense of remembering a specific antigen but could be remembering recent danger signals and remaining in a 'primed' state through an epigenetic modification.

And yes, this is very much the area jnmaciuch and I were discussing in terms of which...

• Jonathan Edwards

• 

 

[Evergreen]

In a sense, this is what Chia clamoured for for years in ME/CFS, though if I remember correctly, he was talking about stomach tissue rather than colorectal tissue.

It seems to boil down to this:

A total of 12/44 (27%) of the LC group and 1/13 (7.7%) of the recovered group had evidence of SARS-CoV-2 persistence by any of the above bulk or in-situ tissue methods (p=0.26, Fisher exact test).

I do not see a similar headline sentence about EBV.

What I cannot see is whether there was any blinding. If those taking the biopsies and analysing the biopsies were aware of who had LC and who had recovered, some bias could creep in.

Let me know if I missed a bit on blinding.

This seems reasonable:

Viral reservoirs may be distributed focally, intermittently, or preferentially in anatomical sites that were not sampled. More comprehensive sampling across different regions of the gastrointestinal tract will be required to define the true distribution of persistent viral material.

Of note:

However, we also note that despite anecdotes suggesting improvement in LC with antiviral therapy37–41, randomized trials employing small molecule antivirals or monoclonal antibodies to date have not demonstrated a clear benefit among those with broadly defined LC42–44.

 

[Evergreen]

What do you think of these bits, @Jonathan Edwards? They seem relevant for the various theories that have been considered on here recently:

A key advance of this study is the coexistence of sustained innate immune activation with attenuation of the adaptive immune pathways required for effective elimination of virus-containing cells. Across transcriptional and spatial analyses, macrophages, other myeloid populations, and fibroblasts exhibited programs consistent with ongoing viral sensing and inflammatory signaling, including activation of TLR-, RIG-I- and TNF-associated pathways. At the same time, pathways involved in phagocytosis, antigen processing and presentation, and cytotoxic lymphocyte function were broadly downregulated in bulk tissue analyses. These abnormalities occurred with relatively limited changes in overall tissue cellular composition, indicating an altered cellular state rather than broad changes in immune cell abundance. The resulting picture suggests an uncoordinated immune response in which inflammatory sensing remains active while the mechanisms needed to identify and eliminate the source of that stimulation are impaired.

At the same time, it remains unclear how localized gastrointestinal persistence and immune dysregulation would produce the multisystem manifestations of LC. Potential mechanisms by which this could occur include generation of circulating inflammatory mediators, microbial or viral products released through impaired barrier function47,48, or neural signaling through vagal afferents49,50. Alternatively, gastrointestinal persistence may represent one component of a broader, anatomically distributed reservoir rather than the primary source of systemic disease. Determining whether the gut directly drives distant organ dysfunction, contributes indirectly through systemic immune or neural signaling, or serves as a marker of persistence elsewhere will require longitudinal studies linking tissue findings to circulating mediators, barrier integrity, autonomic signaling and organ-specific outcomes. [bolding added]

 

[fst]

In a sense, this is what Chia clamoured for for years in ME/CFS, though if I remember correctly, he was talking about stomach tissue rather than colorectal tissue.

It seems to boil down to this:

I do not see a similar headline sentence about EBV.

What I cannot see is whether there was any blinding. If those taking the biopsies and analysing the biopsies were aware of who had LC and who had recovered, some bias could creep in.

Let me know if I missed a bit on blinding.

This seems reasonable:


Of note:

Regardless of the methods used (which I cannot judge), 27% does not seem an awful lot. Maybe this is a proxy for having had a heavier infection?

Also the control group is very small. Might as well have been two or three out of 14 instead of one, just by chance.

 

[Jonathan Edwards]

@Jonathan Edwards are these kind of T like cells you suggested being involved the other day?

No, just bog standard macrophages. They aren't monocytes because they in tissue. The authors might at least know that much.

 

[Jonathan Edwards]

What do you think of these bits, @Jonathan Edwards? They seem relevant for the various theories that have been considered on here recently:

To me it reads 'we didn't find much that made for any coherent story but we thought we would blather a bit all the same to make it sound good.'

It seems that some people still have bits of Covid in their gut but not that many and otherwise the gut has a few more active macrophages - which might be due to chewing Brufen or whatever.

 

[V.R.T.]

To further explore the relationship between SARS-CoV-2 detection and these immunologic changes, an
additional differential expression analysis was performed comparing those with detectable virus (n=12) vs
those without detectable virus (n=32) within the LC group. There were no significantly differentially expressed genes or enriched pathways, implying that changes in host immune response associated with LC occur independently of viral detection.

So I thought this passage was really interesting because the authors seem to be slowly backing away from covid 19 persistence and towards the possibility an immune signaling pathology.

Interestingly, we observed downregulation of IFNα and IFNγ Hallmark
signaling pathways in the subset of people with LC experiencing PEM compared to recovered controls (Fig 2f)

This bit also struck me as potentially useful. it's the opposite of what we might expect to see if some of our IFN theories were correct no?

 

[Silip]

I just got around to reading this paper, and I have to say, honestly, it really confused me. Because to me, the results of these studies—if they mean anything at all—seem to point AGAINST SARS-CoV-2 persistence as a main driver of Long COVID.

Letting aside that the "persistence" they found is no sign for active replication and may be meaningless anyway they didn't find persistent SARS-CoV-2 in 73% of people with Long Covid. Furthermore they state that the immune signature has no correlation to the viral persistence.

"Finally, spatial transcriptomic analyses demonstrated that upregulation of inflammatory signaling in macrophages, epithelial cells, and fibroblasts was distributed throughout the colorectal tissue, rather than localized to the site of viral persistence."

Still their quintessence still seem to be that their results are in support of SARS-CoV-2 persistence being a main driver of Long Covid:

"One possibility is that persistent virus, even if not detected on biopsy, directly drives chronic innate immune activation."

or

"Between these two models, our observation that innate immune activation is sustained months to years after the most recent infection implies the ongoing presence of immune stimuli, i.e. persistent virus, even if it cannot be readily detected."

For me it looks like some of these researchers are still fully locked on to viral persistence and are interpreting every data they get from their studies in favour of this hypothesis.

Or can anybody give me another good reason why not finding persistent viruses in most Long Covid biopsies and that people with and without persistence can't be distinguished by their immune signature should be interpreted in favor of SARS-COV-2 persistence being a main driver of Long Covid?

 

[Chandelier]

New Study Reveals Substantial Tissue-Level Immune Dysfunction Near Persistent Viral Reservoirs in Long COVID - PolyBio Research Foundation

PolyBio-funded UCSF research finds persistent SARS-CoV-2 RNA alongside an altered gut immune response that may impair viral clearance. The findings provide a rationale for treatment strategies that target both persistent viral reservoirs and the tissue immune dysfunction that may allow them to...

polybio.org

Importantly, the researchers caution that the study likely underestimates the prevalence of SARS-CoV-2 persistence.
Persistent viral RNA is present at extremely low abundance relative to human RNA, making it technically difficult to detect.
Moreover, the endoscopic biopsy samples collected by the research team contain only a minute fraction of the ~25-30 foot (7-9 meter) long gastrointestinal tract.
Viral reservoirs could therefore be focal, intermittent or concentrated in nearby or distant anatomical locations that were not sampled, meaning a negative colorectal biopsy does not rule out persistence elsewhere.

 

[Hutan]

I'm struggling to know what to take away from this paper. I suspect it might be one of those ones where they threw everything in, because nothing was really that convincing.

Unsupervised hierarchical analysis further identified four distinct transcriptional clusters within the LC group defined by distinct patterns of DEGs as determined by K-means clustering (Fig 2d). These transcriptional clusters were not clearly associated with clinical phenotypes, including the presence of gastrointestinal symptoms (Fig 2d).

Figure 2 d:

Fig 2. Targeted and bulk transcriptomic analyses of colorectal and peripheral blood mononuclear cell (PBMC) RNA and deep plasma proteomic profiling of Long COVID (LC) cases compared with recovered controls.
A heat map of all 4757 FDR-adjusted DEGs from colorectal bulk tissue whole-transcriptome polyA-RNAseq in people with LC and recovered controls (n=55) is shown in (d). The top horizontal bars represent which participants experienced each LC symptom phenotype at the time of tissue sampling (PEM = post exertional malaise; GI = gastrointestinal, taste = smell and taste disturbances; fatigue; neuro =neurocognitive symptoms; LC>5 = participants experiencing greater than 5 LC symptoms, and LC>1 participants experiencing one or more LC symptom). 8 distinct gene modules were identified by non-hierarchical K-means clustering (row groups separated by white lines). A full list of significant DEGs is listed in Supplementary Data File S1 and a heatmap with only the top 50 DEGs including gene names is shown in Fig S2. Red arrows denote samples in which SARS-CoV-2 was detected by any method.

My interpretation - I might be wrong, there is a lot to read and I have not read it all.

Differential transcriptomics from colorectal samples (comparison between the patients and the controls). The deeper the red or the blue, the bigger the difference between the participant and what? the mean result of all the samples? The comparisons are done for every participant - their codes are along the bottom - LC o or R.

Vertically, you have got the participants divided up into five groups, based on how similar their pattern of differentially expressed genes is. Bear in mind, they are chucked out all the data where the LC and control groups looked similar, and are only working with the expressed genes where the LC and control groups are different. So, most of the recovered participants should look distinctly different, and they do, there in the second column. That in itself is not proof of much, because they selected the data that way.

One question is, do the Long Covid participants, which are a very mixed bag, somehow differentiate by symptom? The symptoms of each participant are given across the top. And really, there is nothing there. Most participants reported fatigue so that doesn't really discriminate. 66% reported PEM, but those people are scattered across the groups.

Even when they did their best to only choose the genes that separated out the controls, some of the controls look more like some of the Long Covid people. Some of the Long Covid people look more like the most of the controls.

Some of the gene groups (those are in the horizontal groups) look a bit interesting, in that they do tend to separate the controls from the LC. So, it will be good to see what genes those are. Bear in mind, many of those differences will be just random variation, because there were so many expressed genes to choose from.

 

[Hutan]

Here's Fig 2e. It shows the gene groups that were differentially expressed between that mixed bag of Long Covid people and the healthy controls. The red blobs indicate more significant differences than the blue blobs. The x axis tells you if the expression was more or less in the LC group compared to the controls.

I don't know, maybe you can find something interesting there. I haven't looked to see how well matched the controls were.

To further understand the unbiased RNAseq data, we again performed GSEA. We generated directional ridge plots to identify the top gene pathways and whether the genes in each pathway, taken together, were up or downregulated when comparing LC and recovered groups (Fig 2e, complete list of significant pathways in Table S4).

Top significant upregulated Hallmark pathways in those with LC included MYC targets, MTORC1 and PI3K-AKT-mTOR signaling, TNFα signaling, glycolysis, inflammatory response, complement and several pathways involving cell cycling, metabolism and protein/hormonal responses. Significantly downregulated Hallmark pathways in those with LC included heme metabolism, adipogenesis, bile acid metabolism, and oxidative phosphorylation.

Applying Kyoto Encyclopedia of Genes and Genomes (KEGG) to the GSEA to better understand various disease processes associated with the DEGs, we observed increased pathways including mRNA surveillance, TNF signaling, and EBV infection (among others as shown in Fig 2e). Interestingly, the acute COVID-19 pathway was significantly downregulated in the LC versus recovered groups.