key: cord-0927590-qm2xpzn9
authors: Triana, Sergio; Metz Zumaran, Camila; Ramirez, Carlos; Kee, Carmon; Doldan, Patricio; Shahraz, Mohammed; Schraivogel, Daniel; Gschwind, Andreas R.; Steinmetz, Lars M.; Herrmann, Carl; Alexandrov, Theodore; Boulant, Steeve; Stanifer, Megan L.
title: Single-cell analyses reveal SARS-CoV-2 interference with intrinsic immune response in the human gut
date: 2020-10-24
journal: bioRxiv
DOI: 10.1101/2020.10.21.348854
sha: 43e356df3cb6ac39098013e6e8a5fbec8f404d31
doc_id: 927590
cord_uid: qm2xpzn9

Objective Exacerbated pro-inflammatory immune response contributes to COVID-19 pathology. Despite the evidence about SARS-CoV-2 infecting the human gut, little is known about the importance of the enteric phase of SARS-CoV-2 for the viral lifecycle and for the development of COVID-19-associated pathologies. Similarly, it remains unknown whether the innate immune response triggered in this organ to combat viral infection is similar or distinct compared to the one triggered in other organs. Design We exploited human ileum- and colon-derived organoids as a non-transformed culture model supporting SARS-CoV-2 infection. We characterized the replication kinetics of SARS-CoV-2 in intestinal epithelial cells and correlated the expression of the viral receptor ACE2 with infection. We performed conventional and targeted single-cell transcriptomics and multiplex single-molecule RNA fluorescence in situ hybridization and used IFN-reporter bioassays to characterize the response of primary human intestinal epithelial cells to SARS-CoV-2 infection. Results We identified a subpopulation of enterocytes as the prime target of SARS-CoV-2. We found the lack of positive correlation between susceptibility to infection and the expression of ACE2 and revealed that SARS-CoV-2 downregulates ACE2 expression upon infection. Infected cells activated strong proinflammatory programs and produced interferon, while expression of interferon-stimulated genes was limited to bystander cells due to SARS-CoV-2 suppressing the autocrine action of interferon in infected cells. Conclusion Our findings reveal that SARS-CoV-2 curtails the immune response in primary human intestinal epithelial cells to promote its replication and spread and this highlights the gut as a proinflammatory reservoir that should be considered to fully understand SARS-CoV-2 pathogenesis. Significance of the study What is already known about this subject? COVID-19 patients have gastrointestinal symptoms which likely correlates with SARS-CoV-2 infection of the intestinal epithelium SARS-CoV-2 replicates in human intestinal epithelial cells. Intestinal organoids are a good model to study SARS-CoV-2 infection of the gastrointestinal tract There is a limited interferon response in human lung epithelial cells upon SARS-CoV-2 infection. What are the new findings? A specific subpopulation of enterocytes are the prime targets of SARS-CoV-2 infection of the human gut. There is a lack of correlation between ACE2 expression and susceptibility to SARS-CoV-2 infection. SARS-CoV-2 downregulates ACE2 expression upon infection. Human intestinal epithelium cells produce interferon upon SARS-CoV-2 infection. Interferon acts in a paracrine manner to induce interferon stimulated genes that control viral infection only in bystander cells. SARS-CoV-2 actively blocks interferon signaling in infected cells. How might it impact on clinical practice in the foreseeable future? The absence of correlation between ACE2 levels and susceptibility suggest that medications influencing ACE2 levels (e.g. high blood pressure drugs) will not make patients more susceptible to SARS-CoV-2 infection. The restricted cell tropism and the distinct immune response mounted by the GI tract, suggests that specific cellular restriction/replication factors and organ specific intrinsic innate immune pathways can represent unique therapeutic targets to treat COVD-19 patients by considering which organ is most infected/impacted by SARS-CoV-2. The strong pro-inflammatory signal mounted by the intestinal epithelium can fuel the systemic inflammation observed in COVID-19 patients and is likely participating in the lung specific pathology.

Coronavirus Disease 2019 (COVID-19) is caused by the severe acute respiratory syndrome 90 coronavirus 2 (SARS-CoV-2). This highly infectious zoonotic virus has caused a global pandemic 91 with almost 30,000,000 people infected worldwide as of September 2020. An exacerbated pro-92 inflammatory immune response generated by the host has been proposed to be responsible for 93 the symptoms observed in patients [1] [2] [3] . Numerous studies have correlated the nature and 94 extent of the immune response with the severity of the disease [4-6]. While many countries have 95 succeeded in curtailing the first wave of infection, there is growing evidence that a second wave 96 of infection is expected to take place and has even already started in some countries. Therefore, 97

it is very urgent that we understand the virus-induced pathogenesis, in particular the immune 98 response generated by the host, to develop prophylactic therapeutics, antiviral approaches and 99 pharmacological strategies to control and revert the pathologies seen in patients. 100 101 SARS-CoV-2 is a member of the betacoronavirus genus, which initiates its lifecycle by exploiting 102 the cellular receptor angiotensin converting enzyme 2 (ACE2) to enter and infect host cells [7] . 103 Virus entry relies not only on ACE2, but also on the cellular proteases furin and the 104 transmembrane serine protease 2 (TMPRSS2) that cleave and activate the SARS-CoV-2 105 envelope spike protein [8] . Following release of the genome into the cytosol, translation of the 106 positive strand RNA genome is initiated and viral proteins quickly induce the formation of cellular 107 membrane-derived compartments for virus replication and de novo assembly of virus particles. 108

The host cells execute several strategies to counteract viral replication. Cellular pathogen 109 recognition receptors (PRRs) sense viral molecular signatures or pathogen-associated molecular 110 patterns (PAMPs) and induce a signaling cascade leading to the induction of interferons (IFNs)  111 and pro-inflammatory molecules. IFNs represent the first line of defense against viral infection as 112 their autocrine and paracrine signaling leads to the production of hundreds of interferon-113 stimulated genes (ISGs) known to exert broad antiviral functions [9] . 114 SARS-CoV-2 infection is not limited to the respiratory tract and COVID-19 patients show systemic 115 manifestation of the disease in multiple organs [10, 11] . For many of these organs, it is unclear 116

whether the pathology is a side effect of SARS-CoV-2 infection in the lung and its associated pro-117 inflammatory response or whether it is due to a direct SARS-CoV-2 infection of the specific organ. only discrete cells are susceptible to SARS-CoV-2 infection and some evidence suggests that 127 these cells may be enterocytes [16] . However, the precise cell tropism of SARS-CoV-2 within the 128 colon and other parts of the gastrointestinal tract is yet to be fully characterized. Finally, despite 129 the driving role of inflammation in the pathologies observed in COVID-19 patients, we are still 130 lacking important molecular details concerning the inflammatory response generated by SARS-131

CoV-2 infected cells and how the surrounding bystander cells will respond to it. 132 133

Here, we aim to address the outlined gaps by applying single-cell RNA-sequencing to human 134 ileum-and colon-derived organoids infected with SARS-CoV-2. Using differential gene 135 expression analysis and multiplex single-molecule RNA fluorescence in situ hybridization (FISH), 136 we investigated the cell type tropism of SARS-CoV-2 and its link to ACE2 expression levels. While 137

we could show that immature enterocytes represent the primary site of SARS-CoV-2 infection, 138

we did not observe correlation between infectivity and ACE2 expression. Interestingly, we could 139 observe that SARS-CoV-2 infection is associated with a downregulation of ACE2 expression. To identify the cell types present in our organoids, we used the Uniform Manifold Approximation 193

and Projection (UMAP) algorithm for dimensionality reduction of our scRNAseq data. We 194 identified eight and nine clusters of cells in the colon and ileum organoids, respectively (Fig. 1B) . 195

Using both cell-type-specific marker genes ( Interestingly, these cells were infected mostly at 24 hpi, suggesting that they are secondary 252 targets of infection. Taken together, these results suggest that immature enterocytes 2 are the 253 primary target of SARS-CoV-2 infection in hIECs both in colon and ileum. 254 255 SARS-CoV-2 cell tropism and association with expression of ACE2 and TMPRSS2 256

The angiotensin-converting enzyme 2 (ACE2) and the cellular protease type II transmembrane 257 serine protease 2 (TMPRSS2) are known to be major determinants for SARS-CoV-2 infection. 258 ACE2 is the cellular receptor of SARS-CoV-2 mediating viral entry [7] . TMPRSS2 is a cellular 259 protease that processes the SARS-CoV-2 spike (S) protein which is an essential step for viral 260 envelope fusion with the host membrane and release of viral contents in the cytosol of the cells. 261

Combined conventional and targeted scRNAseq enabled us to investigate the link between 262 SARS-CoV-2 genome copy numbers and expression of ACE2 in a cell type-specific manner.

Different to what we have expected, immature enterocytes 2, the main site of SARS-CoV-2 264 infection in both colon and ileum organoids ( Fig. 2A-B) , were not the cells displaying the highest 265 levels of ACE2 ( Fig. S4A-B) . Analysis of ACE2 expression levels in all cell types revealed that 266 cells with relatively high levels of ACE2 (e.g. enterocytes 1) were not susceptible to SARS-CoV-267 2 infection ( Fig. 2B-C) . Similarly, we found that SARS-CoV-2 infection is not associated with the 268 expression of the receptor structural homologue ACE, a candidate receptor for SARS-CoV-2 269 basigin (BSG, also known as CD147), as well as the cellular proteases furin, Cathepsin L1 270 (CTSL), aminopeptidase ANPEP and DPP4 (MERS-CoV receptors) (Fig. 2C ). On the contrary, 271

TMPRSS2 was found to be highly expressed in immature enterocytes 2 (Fig. 2C ). In summary, 272

although ACE2 is a recognized receptor for SARS-CoV-2, we found no association between 273 ACE2 expression levels and susceptibility to infection on the single-cell level or across detected 274 types of hIECs. ACE2 expression was observed in infected cells, progressing from 12 hpi to 24 hpi, as compared 290

to mock-infected cells (Fig. 3B , left panel). Importantly, no significant difference of ACE2 291 expression in the bystander cells was observed (Fig. 3B ). In ileum-derived organoids, ACE2 292 expression was also downregulated in the infected cells. However, in contrast to colon organoids, 293 ACE2 expression in bystander cells of ileum organoids was also downregulated as compared to 294 mock-infected cells (Fig. 3B , right panels). Moreover, ACE2 expression was found to be 295 negatively correlated with the presence of the viral genome ( Fig. 3C-D) . The downregulation of 296 ACE2 expression was not only observed in immature enterocytes 2 which were identified as the 297 primary site of SARS-CoV-2 infection (Fig. 2B ), but it was also observed in most other cell types 298 present in ileum-derived organoids over the course of SARS-CoV-2 infection (Fig. 3E ). 299

Expression levels of the other SARS-CoV-2 putative receptors and of key cellular proteases (e.g. 300

TMPRSS2, furin and CTSL) were also found to be reduced in both infected and bystander cells 301

in ileum derived-organoids as compared to mock-infected cells (Fig. 3B, right panel) . Interestingly, 302 when considering colon-derived organoids, the expression levels of these cellular genes were 303

found slightly increased at 12 hpi and decreased at 24 hpi (Fig. 3B, left panel) . Altogether, these 304 data suggest that ACE2 expression is downregulated in colon-and ileum-derived hIECs upon 305 SARS-CoV-2 infection. To validate this observation, we performed multiplex single-molecule 306 fluorescence in situ hybridization (FISH) on SARS-CoV-2 infected organoids. At 12 and 24 hpi, 307 organoids were fixed and evaluated using transcript-specific probes directed against the SARS-308

CoV-2 genome and ACE2. Fluorescence microscopy analysis confirmed that infected cells indeed 309 display lower expression levels of ACE2 at both 12 hpi and 24 hpi (Fig. 3F , white arrow). 310

Quantification of the relative expression levels of SARS-CoV-2 genome and ACE2 transcripts in 311 the RNA FISH images at the single-cell level again confirmed a negative correlation between 312 SARS-COV-2 and ACE2 (Fig. 3G) . Altogether, our data strongly suggest that the expression 313 levels of ACE2 decrease in both colon and ileum hIECs upon SARS-CoV-2 infection. 314 315

To evaluate the response of hIECs to SARS-CoV-2 infection, we performed a comparative gene 317 expression analysis between mock-infected and infected organoids. For the infected organoids, 318

we considered separately the infected cells (those with SARS-CoV-2 genome detected) and the 319 bystander cells (those without SARS-CoV-2 genome). In colon organoids, already at 12 hpi hIECs 320 display a strong NFκB and TNF response to infection with this response becoming even more 321 pronounced at 24 hpi ( Fig. S5A-B) . When comparing mock to bystander cells, we noticed that at 322 24 hpi, the response of bystander cells mostly followed an IFN-mediated immune response 323 characterized by the presence of multiple ISGs (Fig. S5C-D) . This observation suggests that 324

infected cells generate a pro-inflammatory response while bystander cells likely respond to the 325 secreted IFN in a paracrine manner. This is supported by the differential gene expression analysis 326 of bystander vs. infected cells showing that infected cells have a stronger NFκB and TNF 327 mediated response compared to bystander cells ( Fig. S5E-F) . Similar results were found in 328 SARS-CoV-2 infected ileum-derived organoids ( Fig. S5G-L) . Interestingly, at 24 hpi, some 329 interferon-stimulated genes (ISGs) (e.g. IFIT1-3, MX1, CXCL10, IRF1) were found to be also 330 upregulated in infected cells but to a much lesser extent compared to bystander cells ( Fig. S5G -331 J). Additionally, while infected cells in ileum-derived organoids were found to generate a similar 332

NFκB/TNF-mediated response compared to the colon-derived organoids ( Fig. S5B and H), ileum-333 derived bystander cells had a stronger IFN-mediated response which can be seen by the overall 334 higher expression of ISGs in ileum organoids compared to colon organoids upon SARS-CoV-2 335 infection ( Fig. S5D and J) . Together, these comparative gene expression analyses revealed that 336 upon SARS-CoV-2 infection of human intestinal epithelial cells, both strong pro-inflammatory and 337

IFN-mediated responses are generated. 338 339

Cell type specific immune response in infected vs. bystander cells 340

Taking into account the differences in the susceptibility of the different hIEC types to SARS-CoV-341 2, with immature enterocytes 2 constituting the main site of SARS-CoV-2 infection (Fig. 2) , we 342 compared the response of each individual cell type to SARS-CoV-2 infection. Similar to the 343 analysis of all cell types taken together (Fig. S5) , differential gene expression analysis of colon-344 derived infected immature enterocytes 2 revealed a strong NFκB/TNF-mediated response while 345 bystander immature enterocytes 2 mostly display an IFN-mediated response (Fig. 4A-C and Fig.  346 S6A). Similarly, in ileum-derived organoids, infected immature enterocytes 2 also showed a strong 347

NFκB/TNF-mediated response (Fig. 4F, 4H and S6B) while bystander cells were characterized 348 by their response to secreted IFNs leading to ISG expression ( Fig. 4G and Fig. S6B ) 349 350

Pathway analysis confirmed that the bystander response was mostly an IFN-related response 351

while the infected cell response was mostly pro-inflammatory (Fig. S7) . Comparison of the 352 transcriptional response to SARS-CoV-2 infection in infected vs. bystander immature enterocytes 353 2 further confirmed that infected cells mount a strong pro-inflammatory response characterized 354 by the upregulation of NFκB and TNF (Fig. 4C-D and 4H-I) . Analysis of the top 30 differentially 355 expressed ISGs in colon-derived organoids clearly shows that at 24 hpi, bystanders cells respond 356

to IFN by upregulating the expression of a large panel of ISGs (Fig. 4E) . Similar findings were 357 observed in immature enterocytes 2 from ileum organoids, although infected cells were also found 358

to express more ISGs compared to their colon-derived counterparts. Importantly, in both colon 359 and ileum organoids, bystanders showed higher levels of expression for all considered ISGs 360 compared to infected cells ( Fig. 4E and 4J ). These results are consistent with the observation that 361

ileum-derived organoids are more immune-responsive compared to colon-derived organoids ( To determine if the characterized NFκB/TNF-high and IFN-low immune response is specific to 368 immature enterocytes 2, we also looked individually at each infected cell type. We found that in 369 all considered types of hIECs, bystander cells display an IFN-mediated response during SARS-370

CoV-2 infection while infected cells are characterized by a strong NFκB/TNF-mediated response 371 (Fig. S8-9 ). Since our targeted scRNAseq analysis revealed the presence of background viral 372 RNA in the samples (Fig. S3) , we asked whether the observed immune response is indeed 373 cascading to bystander cells through type III interferon secreted by infected cells or is caused by 374 the direct action of non-replicating viral particles on bystander cells. To address this, colon and 375 ileum organoids were infected with either live or UV-inactivated SARS-CoV-2. Results revealed 376 that upon infection with live SARS-CoV-2 both IFNs and ISGs were produced (Fig. S10 ). On the 377 contrary, exposure of organoids to UV-inactivated SARS-CoV-2 did not lead to virus replication 378

and cells failed to produce both IFN and ISGs (Fig. S10 ). This demonstrates that active replication 379 is required for the described immune response and allowed us to rule out the exposure to non-380

replicating viral particles as being the cause of this response. Altogether our results show that 381 infected and bystander cells respond differently to SARS-CoV-2 infection where infected cells 382 mount a NFκB/TNF-mediated response while bystander cells mount a IFN-mediated response. 383 384

Signaling activity in infected vs. bystander cells 385

To characterize the signaling that underpins the distinct immune response of infected and 386 bystander cells upon SARS-CoV-2 infection, we inferred the pathway signaling activity from 387 scRNAseq data with PROGENy ( Fig. 5A-B) . For both colon and ileum organoids, infected cells 388

show a strong activation of the MAPKs, NFκB and TNFα pathways. In line with the enrichment 389 analysis (Fig. S7, S8) , these pathways were found to be less activated in bystander cells with 390 higher scores in ileum compared to colon (Fig. 5A-B) . Interestingly and in accordance with our 391 differential gene expression analyses (Fig. 4, S5 and S6) , the JAK-STAT signalling pathway was 392 found to be activated mostly in bystander cells (Fig. 5A-B) . To further elucidate the signalling 393 activity at the single-cell level, we generated diffusion maps of all single cells based on the 394 scRNAseq expression of interferon-related genes (Fig. 5C-D) . In both iluem and colon, we 395 observed a clear bifurcation of all cells into two distinct branches, one branch representing mainly 396 infected cells and another branch representing mainly bystander cells. In addition, we calculated 397 the activities of selected transcription factors (TFs) for all single cells using SCENIC and mapped 398 the inferred activities onto the single-cell diffusion maps (Fig. 5C-D, right insets) . We found that 399 the transcription factors STAT1 and IRF1 were activated mainly in bystander cells (branch along 400 DC1) while JUN was activated in infected cells (branch along DC2) (Fig. 5C-D, left panels) . 401

Extending this analysis to transcription factors whose activity pattern is highly correlated to either 402 DC1 or DC2 revealed that globally, transcription factors that are critical for IFN-mediated signaling 403 (i.e. the ISGF3 complex: STAT1/STAT2/IRF9 and IRF1) are highly active in bystander cells (Fig.  404 5E-F 

To validate that the IFN-mediated response is specific to bystander cells, ileum-derived organoids 413

were infected with SARS-CoV-2. At 12 and 24 hpi, single molecule RNA FISH was performed 414 using probes specific for the SARS-CoV-2 genome and for ISG15 which was found to be highly 415 upregulated upon infection and has the highest -log10 p-value in the differential analysis 416 comparing bystander cells vs mock-infected cells (Fig. 4G) . Microscopy images revealed that 417 bystander cells (non-infected) were indeed positive for ISG15 (Fig. 6A) . Interestingly, SARS-CoV-418 2 infected cells were found to express little to no ISG15 (Fig 6A, white arrows) . Quantification 419 confirmed that cells which displayed the highest expression of ISG15 were negative for SARS-420

CoV-2 genome (Fig. 6B) . These results support the model that bystander cells respond to SARS-421

CoV-2 infection by mounting an IFN-mediated response. On the other hand, as shown by using 422 both scRNAseq (Fig. 4) and RNA FISH (Fig. 6A) the IRF3 KO T84 did not result in the production of ISGs as monitored by q-RT-PCR of ISG15 431 (Fig. 6D ). IRF3 KO T84 cells were mock infected or infected with SARS-CoV-2, at 24 hpi cells 432

were treated with IFN and 12 h post-treatment production of ISG15 was assessed by q-RT-PCR 433 (Fig. 6C) . Results show that mock infected IRF3 KO T84 cells were responsive to IFN 434 demonstrating that genetic depletion of IRF3 does not alter IFN-mediated signaling (Fig. 6D) . 435 Interestingly, in SARS-CoV-2 infected IRF3 KO T84 cells, production of ISG15 upon IFN 436 treatment was significantly downregulated (Fig. 6D ). To confirm that this impaired induction of 437 ISG15 upon IFN treatment was specific to SARS- CoV-2 infection, IRF3 KO T84 cells were  438 infected with astrovirus at an MOI of 3 to achieve full infection (Fig. 6E ). Contrary to SARS-CoV-439 2 infection, infection of IRF3 KO T84 cells by astrovirus did not impair IFN-mediated signaling as 440 a similar upregulation of ISG15 was observed in both mock-infected and astrovirus-infected cells 441 upon IFN treatment (Fig. 6D) . In a second validation approach, to fully demonstrate that only 442 infected cells have an altered IFN-mediated signaling, we developed an assay based on a 443 fluorescent reporter of ISG expression (Fig. 6F ). For this, we generated a T84 cell line transduced 444 with a reporter made of the promoter region of the ISG MX1 driving the expression of the 445 fluorescent protein mCherry. Mx1 is known to be made strictly downstream of the IFN receptor in 446 a JAK-STAT dependent manner but not downstream of IRF3. Upon IFN treatment, about 30-40% 447 of cells expressing this reporter were responsive and became fluorescent (Fig. 6G) . Following 448 infection with SARS-CoV-2 at an multiplicity of infection (MOI) of 3, most of the cells were found 449 to be infected (Fig. 6G ). However when cells were treated 24 hpi with IFN, most infected cells did 450 not respond to IFN and very few became double positive for both virus and MX1 driven mCherry 451 (Fig. 6G, left panel) . To control that non-infected bystander cells could indeed respond to IFN and 452 express mCherry, we repeated this experiment but using an MOI of 1 for SARS-CoV-2 infection. 453

About 40% of the cells were found to be infected. Supplementing IFN affected mainly non-infected 454 cells, as can be seen from the increase of MX1-positive cells and no change in the number of 455 double-positive cells (both infected and MX1-positive) (Fig. 6G, identified that a subpopulation of enterocytes (namely, immature enterocytes 2) is the cell type 473 most susceptible to SARS-CoV-2 infection. Interestingly, other cell types also supported infection 474 by SARS-CoV-2 but to a much lesser extent (Fig. 2B) . The characterized tropism of SARS-CoV-475 2 could be explained by either cell type-specific intrinsic differences rendering some cell type 476 more permissive or due to an overrepresentation of cells of a particular cell type. In our colon-477 derived organoids, there were twice as many immature enterocytes 1 compared to immature 478 enterocytes 2 and in ileum-derived organoids both enterocytes lineages were present in roughly 479 equal numbers. This suggests that the SARS-CoV-2 cell tropism for immature enterocytes 2 is 480 not due to a higher proportion of these cells in our organoids but due to intrinsic differences 481 between immature enterocytes 2 and other epithelial cell lineages. Differential gene expression 482 analysis between immature enterocytes 2 and the most similar other annotated cell type 483

(immature enterocytes 1) does not highlight the presence or absence of obvious 484 restriction/replication factors that could explain the observed tropism. 485 486

The expression levels of the SARS-CoV-2 receptor ACE2 were found to be higher in immature 487 enterocytes 1 while immature enterocytes 2 express more of the key cellular protease TMPRSS2 488 ( Fig. 2C and S4 ). Although expression of ACE2 is mandatory for infection [7], we noticed no 489 correlation between ACE2 expression level and the copy numbers of SARS-CoV-2 genome in 490 the cell (Fig. 2B-C)(Fig. S11) . tissues. Interestingly, we found that TMPRSS2 expression levels were well-associated with the 496 SARS-CoV-2 genome copy numbers in human intestinal epithelial cells (Fig. 2C and S4 ) which 497

is consistent with the observation that TMPRSS2 and the related protease TMPRSS4 are critical 498

for infection of intestinal organoids [18] . As such it is tempting to speculate that TMPRSS2 plays 499 a more important role in the SARS-CoV-2 cell tropism than ACE2, however more studies are 500 required to fully explore this hypothesis. 501 502

Several studies have suggested that ACE2 is an interferon-stimulated gene and is induced upon 503 SARS-CoV-2 infection [25, 26] . This led to a speculation that upon infection and induction of pro-504 inflammatory responses, the ACE2 levels would increase thereby favoring SARS-CoV-2 infection. 505

Our results show that, on the contrary, upon infection ACE2 levels decrease both in infected and 506 bystander hIECs (Fig. 3)(Fig. S11) . Interestingly, differences in the kinetics of ACE2 regulation 507

were observed between colon-and ileum-derived organoids. In colon-derived organoids, a small 508 increase in ACE2 expression was observed at early times post-infection (12 hpi) but at later time 509 points (24 hpi) the overall expression of ACE2 (and other putative SARS-CoV-2 receptors and 510 key cellular proteases) was decreased as compared to mock infected cells (Fig. 3B) . In ileum 511 organoids the expression of ACE2 was decreased over the time course of infection. The observed 512 upregulation of ACE2 upon infection might be tissue-specific and time-dependent. However, 513 recently it has been proposed that ACE2 does not behave as an ISG but instead a novel form of 514 ACE2 (dACE2) is interferon-inducible [38]. dACE2 results from transcription initiation at an 515 internal exon leading to the production of an alternative short version. Within our scRNAseq data 516

we could not distinguish between the two forms and as such the observed temporal increase (in 517 colon-derived organoids) could be due to the downregulation of ACE2 with the concomitant 518 upregulation of dACE2. 519 520

The nature of the PRR responsible for sensing SARS-CoV-2 infection is yet to be determined but 521 from recent work and previous work on SARS-CoV-1 and MERS it could be speculated that TLR3, 522

RLRs and the STING pathways could be involved [39, 40] . In our current work we could show that 523 active virus replication is required to induce an IFN-mediated response as infection by UV-524 inactivated SARS-CoV-2 did not lead to IFN and ISG production (Fig. S10) . Interestingly, when 525 human intestinal epithelial cells are infected by a UV-inactivated reovirus, which is a virus whose 526 genome is a dsRNA, an IFN-mediated response is induced [41] . As SARS-CoV-2 is single-527 stranded RNA virus and dsRNA intermediates will only occur during active replication, it is 528 tempting to speculate that what is being sensed are these dsRNA replication intermediates. 529 530 SARS-CoV-2 infection is characterized by a strong pro-inflammatory response and this has been 531 observed both in tissue-derived samples and in vitro using cell culture models [19] . This pro-532 inflammatory response is characterized by upregulation of the NFκB and TNF pathways. Our 533 scRNAseq approach revealed that this pro-inflammatory response is specific to infected cells and 534 that bystander cells do not show a strong pro-inflammatory response. These differences between 535

infected and bystander cells were earlier observed for other cell types: infection of human 536 bronchial epithelial cells (HBECs) also reveal that the pro-inflammatory response is biased toward 537 infected cells and not bystander cells [31] . 538 539

It was reported that infection of human lung epithelial cells by SARS-CoV-2 is characterized by a 540 low to absent IFN response [19] . On the contrary, in human intestinal epithelium cells an IFN-541 mediated response is readily detectable and is characterized by both the production of IFN and 542

ISGs [16, 17] . Interestingly, upon infection with SARS-CoV-2, we observed the upregulation of 543

IFNλ2-3 but we failed to observe a significant increase in IFNλ1 expression (Fig. S1D) . This 544 absence of IFNλ1 upregulation is not specific to SARS-CoV-2 but a particularity of intestinal 545 organoids, as a similar IFNλ2-3 specific response was observed when intestinal organoids were 546

infected with other enteric viruses [42, 43] . Upregulation of IFN production upon SARS-CoV-2 547 infection of intestinal epithelial cells was found to be low but significant (Fig. 4) and this could 548 raise the question whether this small production of IFN is sufficient to induce the production of 549

ISGs in a paracrine manner. When comparing the immune response generated by organoids derived from different sections 557 of the GI tract, we observed that ileum organoids were more immunoresponsive compared to 558 colon organoids. While the extent of the up-regulation of genes related to pro-inflammatory 559 response was similar between colon and ileum (Fig. 4A, 4F) , we observed that ileum organoids, 560 particularly bystander cells produced significantly more ISGs compared to their colon counterparts 561 (Fig. 4b, 4G ). This compartmentalization of response along the GI tract is consistent with previous 562

reports describing that different sections of the GI tract respond differently to microbial challenges 563 [44] . Most importantly, our results reveal that production of ISGs is mostly restricted to bystander 564 cells, while production of IFN is detected mostly in infected cells (Fig. 5, S5 and S6 ). These 565 observed differences between infected and bystander cells were confirmed using single-molecule 566 RNA FISH showing that production of ISG15 was clearly observed in bystander and, to a much 567 lesser extent, in infected cells (Fig. 6 ). An important finding of this work is that infected cells not 568 only fail to produce ISGs, they also become refractory to IFN (Fig. 6) (Fig. S11) . When SARS-569

CoV-2-infected intestinal cells were treated with IFN, only bystander cells upregulated ISG while 570 infected cells did not. This absence of ISG induction in infected cells suggests that SARS-CoV-2 571

has developed mechanisms to shutdown IFN-mediated signaling and the subsequent production 572

of ISGs (Fig. S11) . Preventing IFN-mediated signaling in infected cells would provide a replication 573 advantage to SARS-CoV-2 as secreted IFN will not be able to act in an autocrine manner to 574

induce ISGs which will curtail virus replication and de novo virus production. Although the SARS-575

CoV-2 viral protein responsible for blocking the IFN-mediated signaling is yet to be identified in 576

our system, a recent report suggests that ORF6 could block IFN-mediated signaling by interfering 577

with STAT1 nuclear translocation [45] . 578 579

In conclusion, in this work we identified a subset of immature enterocytes as the primary site of 580 infection of SARS-CoV-2 in ileum-and colon-derived human intestinal epithelial cells. We could 581

show that upon infection, infected cells mount a strong pro-inflammatory response characterized 582 by a strong activation of the NFκB/TNF pathways while bystander cells mount an IFN-mediated 583 response (Fig. S11 ). This differential response between infected and bystander cells is due to an 584 active block of IFN-signaling in infected cells (Fig. S11) The raw sequencing generated during this study is available at the NCBI Gene Expression 940

Omnibus (accession no. GSE156760). The authors declare that all other data supporting the 941 findings of this study are within the manuscript and its supplementary files. 942 943

colon carcinoma cells (ATCC CCL-248) and their knock-out derivative clones were 720 maintained in a 50:50 mixture of Dulbecco's modified Eagle's medium (DMEM) and F12 (GibCo) 721 supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Gibco). Vero E6 722 (ATCC CRL 1586) were maintained in DMEM supplemented with 10% fetal bovine serum and 723 1% penicillin/streptomycin. The mCherry-tagged Mx1

Single clones were derived from this cell line and evaluated for their ability 726 to respond to both type-I and type-III interferons. 727 728 Viruses 729 SARS-CoV-2 (strain BavPat1) was obtained from the European Virology Archive. The virus was

Human organoid cultures and ethic approval 733 Human tissue was received from colon resection or ileum biopsies from the University Hospital 734

This study was carried out in accordance with the recommendations of the University 735

Ethics commission of the University Hospital Heidelberg" 738 under the protocol S-443/2017. Organoids were prepared following the original protocol described 739 by [46]. In brief, stem cells containing crypts were isolated following 2 mM EDTA dissociation of 740 tissue samples for 1 h at 4°C. Crypts were spun and washed in ice cold PBS. Fractions enriched 741 in crypts were filtered with 70 mM filters and the fractions were observed under a light microscope. 742 Fractions containing the highest number of crypts were pooled and spun again. The supernatant 743 was removed, and crypts were re-suspended in Matrigel

Organoids were collected at a ratio of 100 organoids/well. Collected 749 organoids were spun at 450 xg for 5 mins and the supernatant was removed. Organoids were 750 washed 1X with cold PBS and spun at 450 xg for 5 mins. PBS was removed and organoids were 751 digested with 0.5% Trypsin-EDTA (Life technologies) for 5 mins at 37°C. Digestion was stopped 752 by addition of serum containing medium. Organoids were spun at 450 xg for 5 mins and the 753 supernatant was removed and organoids were re-suspended in normal growth media at a ratio of 754 250 μL media/well

Media was removed from cells and 10 6 pfu of SARS-CoV-2 (as determined in Vero cells) was

Cells were washed and permeabilized in 0.5% Triton-X for 15 mins at RT. 773 Mouse monoclonal antibody against SARS-CoV NP (Sino biologicals MM05) and mouse 774 monoclonal against J2 (scions) were diluted in phosphate-buffered saline (PBS) at 1/1000 dilution 775 and incubated for 1h at RT. Cells were washed in 1X PBS three times and incubated with 776 secondary antibodies (conjugated with AF488 (Molecular Probes), or AF568 (Molecular Probes) 777 directed against the animal source) and DAPI for 45 mins at RT. Cells were washed in 1X PBS 778 three times and maintained in PBS

Infections were allowed to proceed for 24h. 24 hours post-785 infection cells were fixed in 2% PFA for 20 mins at RT. PFA was removed and cells were washed 786 twice in 1X PBS and then permeabilized for 10 mins at RT in 0.5% Triton-X. Cells were blocked 787 in a 1:2 dilution of Li-Cor blocking buffer (Li-Cor) for 30 mins at RT. Cells were stained with 1/1000 788 dilution anti-dsRNA (J2) for 1 h at RT. Cells were washed three times with 0.1% Tween in PBS. 789 Secondary antibody (anti-mouse CW800) and DNA dye Draq5 (Abcam) were diluted 1/10,000 in 790 blocking buffer and incubated for 1 h at RT

Organoid dissociation for scRNAseq 794 2D seeded organoids harvested after 0 (mock), 12 and 24 hours post-infection were washed in 795 cold PBS and incubated in TrypLE Express (Gibco) for 25 min at 37°C. When microscopic 796 examination revealed that cells had reached a single cell state

Supernatant was removed and the cell pellet was 798 resuspended in PBS supplemented with 0.04% BSA and passed through a 40 μm cell strainer. 799 Resulting cell suspensions were used directly for single-cell RNAseq

Single-cell RNA-seq library preparation

Single-cell suspensions were loaded onto the 10x Chromium controller using the 10x

Single Cell 3' Library Kit NextGem (10x Genomics) according to the manufacturer's instructions. 804 In summary, cell and bead emulsions were generated, followed by reverse transcription, cDNA 805 amplification (5 μL of of amplified cDNA was set apart for targeted scRNAseq amplification), 806 fragmentation, and ligation with adaptors followed by sample index PCR. Resulting libraries were 807 quality checked by Qubit and Bioanalyzer, pooled and sequenced using HiSeq4000 (Illumina; 808 high-output mode

After washing, cell nuclei were counterstained with DAPI, and samples were mounted using 893

ProLong Gold Antifade Mountant. Imaging was performed with the camera Nikon DS-Qi2 (Nikon 894 Instruments) with the Plan Fluor 20x objective (Nikon Instruments) mounted on the Nikon Ti-E 895 inverted microscope (Nikon Instruments) in bright-field and fluorescence (DAPI, GFP, Cy3 and References 944

Complex Immune Dysregulation in 945 COVID-19 Patients with Severe Respiratory Failure

Heightened Innate Immune Responses in the Respiratory 948 Tract of COVID-19 Patients

COVID-19: consider cytokine storm syndromes and 950 immunosuppression

Clinical and immunological features of severe and moderate 952 coronavirus disease 2019

Deep immune profiling of COVID-19 patients reveals 954 patient heterogeneity and distinct immunotypes with implications for therapeutic 955 interventions. bioRxiv Published Online First: 23

Longitudinal analyses reveal immunological misfiring in 958 severe COVID-19

SARS-CoV-2 Cell Entry Depends on 961 ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

TMPRSS2 and furin are both essential for proteolytic 964 activation of SARS-CoV-2 in human airway cells

Differential Regulation of Type I and Type III 967 Interferon Signaling

Single Virus Targeting Multiple Organs: What We Know and Where 969 We Are Heading?

Extrapulmonary manifestations of COVID-19

Prolonged presence of SARS-CoV-2 viral RNA in faecal 973 samples

Evidence for Gastrointestinal Infection of SARS-CoV-2

Prolonged viral shedding in feces of pediatric patients with 977 coronavirus disease 2019

Virological assessment of hospitalized patients 979 with COVID-2019

SARS-CoV-2 productively infects human gut 981 enterocytes

Critical Role of Type III Interferon in Controlling 983 SARS-CoV-2 Infection in Human Intestinal Epithelial Cells

CoV-2 infection of human small intestinal enterocytes

CoV-2 Drives Development of COVID-19

Impaired type I interferon activity and inflammatory 990 responses in severe COVID-19 patients

Infection of bat and human intestinal organoids by SARS-CoV-2

Intra-and Inter-cellular Rewiring of the 995 Human Colon during Ulcerative Colitis

Single-cell transcriptomics reveals immune 997 response of intestinal cell types to viral infection

Targeted Perturb-seq enables genome-1000 scale genetic screens in single cells

SARS-CoV-2 Receptor ACE2 Is an Interferon-1002

Stimulated Gene in Human Airway Epithelial Cells and Is Detected in Specific Cell Subsets 1003 across Tissues

COVID-19 severity correlates with airway epithelium-1005 immune cell interactions identified by single-cell analysis

ETV7 represses a subset of interferon-stimulated 1008 genes that restrict influenza viruses

Early Growth Response Gene-1 Suppresses Foot-and-Mouth 1010

Disease Virus Replication by Enhancing Type I Interferon Pathway Signal Transduction

Bulk and single-cell gene expression profiling of SARS-1013

CoV-2 infected human cell lines identifies molecular targets for therapeutic intervention

Single cell RNA sequencing of 13 human tissues identify cell 1016 types and receptors of human coronaviruses

Single-cell longitudinal analysis of SARS-CoV-1019 2 infection in human bronchial epithelial cells

Single-cell atlas of a non-human primate reveals new pathogenic 1022 mechanisms of COVID-19

Systemic analysis of tissue cells potentially vulnerable to 1024 SARS-CoV-2 infection by the protein-proofed single-cell RNA profiling of ACE2, TMPRSS2 1025 and Furin proteases

A single-cell RNA expression map of human coronavirus 1027 entry factors

SARS-CoV-2 receptor ACE2 and TMPRSS2 are 1029 primarily expressed in bronchial transient secretory cells

Single-Cell RNA Expression Profiling of ACE2, the 1031 Receptor of SARS-CoV-2

SARS-CoV-2 entry factors are highly expressed in 1033 nasal epithelial cells together with innate immune genes

Interferons and viruses induce a novel 1035 truncated ACE2 isoform and not the full-length SARS-CoV-2 receptor

SARS-CoV-2 infection induces a pro-1038 inflammatory cytokine response through cGAS-STING and NF-κB

Type I and Type III Interferons -Induction, Signaling, Evasion, and 1041 Application to Combat COVID-19

Reovirus intermediate subviral particles constitute 1043 a strategy to infect intestinal epithelial cells by exploiting TGF-β dependent pro-survival 1044 signaling

Type I and Type III Interferons Display 1046 Different Dependency on Mitogen-Activated Protein Kinases to Mount an Antiviral State in 1047 the Human Gut

Asymmetric distribution of TLR3 leads to a 1049 polarized immune response in human intestinal epithelial cells

Location-specific cell identity rather than exposure to 1052 GI microbiota defines many innate immune signalling cascades in the gut epithelium

Activation and evasion of type I interferon responses by SARS-1055 CoV-2

Long-term expansion of epithelial organoids from 1057 human colon, adenoma, adenocarcinoma, and Barrett's epithelium

MAST: a flexible statistical framework for assessing 1060 transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing 1061 data

Controlling the False Discovery Rate: A Practical and Powerful 1063 Approach to Multiple Testing

 Targeted single-cell RNA-sequencing  811 For targeted scRNAseq, outer and inner primers for targeted amplification were designed using 812 an R package for primer design described in [24] and available through Bioconductor 813 (http://bioconductor.org/packages/release/bioc/html/TAPseq.html). Primers were ordered 814 desalted as ssDNA oligonucleotides and pooled in an equimolar amount, except for the primer 815targeting SARS-CoV-2 mRNA which was added in 8-fold excess to the outer and inner panel. All 816primer sequences are described in Supplementary Table 1 . Targeted scRNA-seq was performed 817 as previously described [24] , except for using amplified cDNA from the 10X Genomics 3' scRNA-818 seq protocol as input material. In short, 10 ng of amplified cDNA were used as input for the outer 819primer PCR and amplified with 10 PCR cycles. A second seminested PCR using 10 ng of Ampure 820 purified outer PCR as input was performed with inner primer mix and 7 cycles of PCR. Then, a 821 third PCR was done adding Illumina adapters. Resulting libraries were quality checked by Qubit 822and Bioanalyzer, pooled and sequenced using HiSeq4000 (Illumina; high-output mode, paired-823end 26 x 75 bp). 824 825Pre-processing and quality control of scRNAseq data 826Raw sequencing data was processed using the CellRanger software (version 3.1.0). Reads were 827 aligned to a custom reference genome created with the reference human genome (GRCh38) and 828 SARS-CoV-2 reference genome (NC_045512.2). The resulting unique molecular identifier (UMI) 829 count matrices were imported into R (version 3.6.2) and processed with the R package Seurat 830 (version 3.1.3). Low-quality cells were removed, based on the following criteria. All cells with 831 mitochondrial reads > 30% were excluded. Second, we limited the acceptable numbers of 832 detected genes. For both types of samples, cells with <1500 or >9000 detected genes were 833 discarded. The remaining data were further processed using Seurat. To account for differences 834 in sequencing depth across cells, UMI counts were normalized and scaled using regularized 835 negative binomial regression as part of the package sctransform. Afterward, ileum and colon 836 organoids samples were integrated independently to minimize the batch and experimental 837 variability effect. Integration was performed using the canonical correlation analysis and mutual 838 nearest neighbor analysis. The resulting corrected counts were used for visualization and 839clustering downstream analysis and non-integrated counts for any quantitative comparison. 840 841Pre-processing of targeted scRNAseq 842Targeted scRNA-seq pre-processing was done as described in [24] . In summary, following 843 demultiplexing by sample, sequencing data were processed following the workflow provided by 844Drop-seq tools (v. observations were converted to transcript counts. Cell-containing droplets were extracted using 852 the filtered cell barcodes from the scRNASeq data. Other detected cell barcodes droplets were 853 categorized as empty droplets. The infection status for every cell was extracted from the targeted 854 32 gene expression data by thresholding the SARS-CoV-2 counts using the media expression of the 855 cell containing droplets ( Figure S2 ). Furthermore, Pearson correlation of each targeted gene to 856 its WTA equivalent was calculated. 857 858Clustering and identification of cell type markers 859 We performed principal component analysis (PCA) using 3000 highly variable genes (based on 860 average expression and dispersion for each gene). The top 30 principal components were used 861to construct a shared nearest neighbor (SNN) graph and modularity-based clustering using the 862Louvain algorithm was performed. Finally, Uniform manifold approximation and projection 863 (UMAP) visualization was calculated using 30 neighboring points for the local approximation of 864 the manifold structure. Marker genes for every cell type were identified by comparing the 865 expression of each gene in a given cluster against the rest of the cells using the receiver operating 866 characteristic (ROC) test. To evaluate which genes classify a cell type, the markers were selected 867 as those with the highest classification power defined by the AUC (area under the ROC curve). 868These markers along with canonical markers for intestinal and colonic cells were used to annotate 869 each of the clusters of the ileum and colon samples. 870 871Differential expression analysis 872To identify the changes in expression across conditions. Differential expression tests were 873 performed using MAST [47] . To reduce the size of the inference problem and avoid cell proportion 874 bias, separate models were fit for each cell lineage and comparisons between mock, 12 hours, 875and 24 hours post-infection were performed. False discovery rate (FDR) was calculated by the 876Benjamini-Hochberg method [48] and significant genes were set as those with FDR of less than 877 0.05. Subsequently, genes whose mRNAs were found to be differentially expressed were 878 subjected to a gene set overrepresentation analysis using the EnrichR package in R. Furthermore 879 signalling pathways enrichment was calculated using PROGENy. 880 881Multiplex FISH 882Organoids were seeded in expansion medium on glass coverslips. At 24 hours post-seeding, the 883 expansion medium was replaced by differentiation medium and organoids were left to differentiate 884 for 4 days. After validation that organoids were differentiated and contained all expected cell 885 types, organoids were infected, harvested after 12 and 24 hpi, and fixed in 4% PFA for 30 minutes. 886HiPlex (RNAscope) was performed following the manufacturer's instructions. Briefly, fixed 887 samples were dehydrated with 50%, 70%, 100% ethanol, then treated with protease. All the 888HiPlex probes were hybridized and amplified together. Probes were designed for genes identified 889 as cell type markers and/or corroborated by literature. The probes used were: 890 891Gene name Role of the marker Catalog number 892 Cy5 channels). The microscope was controlled using the Nikon NIS Elements software. After 897 each round, fluorophores were cleaved and samples moved on to the next round of the 898 fluorophore detection procedures. All images from all rounds of staining were then registered with 899 each other to generate images using HiPlex image registration software (ACD Bio). Further 900 brightness and contrast adjustments were performed using Fiji. 901 902The HiPlex probe fluorescent signal was used to determine the ACE2 and ISG15 RNA expression 903 levels, as well as the SARS-CoV-2 infection levels. To obtain a resolution at a single cell level, 904first nuclei segmentation and classification was done on raw DAPI images using the Pixel 905Classification + Object Classification workflow from ilastik 1.2.0. The resulting Object Prediction 906 masks represented all nuclei as individual objects in a 2D plane and were saved as 16bit Tagged  907 Image File Format. To measure the single cell fluorescent intensity for the ACE2, ISG15 and 908 SARS-CoV-2 probes, a pipeline using CellProfiler 3.1.9 was developed. Briefly, first the raw 909grayscale images corresponding to the ACE2, ISG15 and SARS-CoV-2 probe fluorescent signals 910were uploaded on the pipeline. These images were specified as images to be measured. The 911corresponding Object Prediction masks previously generated by ilastik were then uploaded, 912converted into binary nuclei masks and used to define the objects to be measured. Finally, with a 913MeasureObjectIntesity module the fluorescence intensity features, the cell number and the single 914 cell localization were measured for the identified objects from the binary nuclei mask. The 915outcome was exported to a spreadsheet and contained the localization as well as the mean 916 intensity units rescaled from 0 to 1 of ACE2, ISG15 and SARS-CoV-2 fluorescent signals for each 917 single cell. 918 919To determine the infection status for every cell a threshold was calculated using the SARS-CoV-920 2 mean fluorescent intensity signal of mock treated versus representative infected cells.