We first tested if an early strain of SARS-CoV-2 would replicate in wild-type zebrafish larvae after bath exposure. We exposed 4-day post-fertilization (dpf) larvae with inflated swim bladders (ensuring an open gut) as well as 2-dpf dechorionated embryos with suspension of either live or heat-inactivated virus added to water (8 × 10⁴ PFU/ml). Larvae were then incubated at 32°C and observed regularly; no specific signs of distress were noted. After RNA extraction, the amount of polyadenylated SARS-CoV-2 N transcripts was measured by qRT-PCR. Although viral RNA was readily detectable after 6 h of exposure, it then declined and became undetectable after 48 h (Figure 1). Therefore, bath exposure failed to achieve infection.

We then turned to microinjection of larvae with SARS-CoV-2. Using a camera-fitted macroscope under a biosafety hood, a concentrated SARS-CoV-2 suspension was microinjected in various sites of 3 dpf larvae (Figure 2A). Compared with our previous experience of microinjection using the eyepieces of a stereomicroscope, this was significantly harder, notably due to lack of stereovision. These challenging injection conditions resulted in variability during early attempts; this later improved greatly, and although success of intravenous (IV) injections remained difficult to ascertain, others, notably in the coelomic cavity, were achieved reliably and in a reasonable time frame. Injection of the syncytial yolk cell was relatively easy, but leakage was often observed after capillary withdrawal, in which case larvae were discarded. Injected larvae were immediately rinsed and transferred into individual wells of 24-well plates, which were then incubated at 32°C. Larvae were imaged daily; none of the typical disease signs that we noted during other viral infections (e.g., edemas, spine bending, necrotic spots, slow blood flow) (Palha et al., 2013) were observed.

At various time points, individual larvae were euthanized and RNA extracted. Sense transcripts of SARS-CoV-2, genomic or subgenomic, are polyadenylated (Kim et al., 2020) and were measured after reverse transcription with a poly(dT) primer. The initial inoculum, measured in larvae lysed ~30 min postinjection (pi), was readily detectable by qRT-PCR (Table 1). Absolute quantification by qRT-PCR, using certified commercial reagents, revealed an amount of polyadenylated SARS-CoV-2 N transcripts that was ~10⁴-fold higher than the injected number of PFU (Table 1). Therefore, the overwhelming bulk of viral RNA injected in larvae must correspond to non-infectious molecules.

We then measured polyadenylated N copies over time. A decline was observed for all injection sites, with the notable exception of the yolk (Figure 2B). Amounts measured in yolk were highly variable at early time points, more than in other sites, probably due to leakage.

At this point, we wondered if this absence of replication may be due to a loss of infectivity of the virus in our microinjection conditions. We re-titered on Vero cells the leftover inoculum of one of our experiments, which had been left at room temperature for 1.5 h after thawing and addition of phenol red, and refrozen. The titer measured on Vero cells was 4.7 × 10⁷ PFU/ml, i.e., a ~2.5-fold drop from the original titer of 1.13 × 10⁸ PFU/ml. This was consistent with the drop expected from two rounds of freeze-thaw, suggesting that virus infectivity was not significantly decreased by our experimental conditions and confirming that we were injecting a significant amount of infectious particles.

To determine if the relatively high amounts detected in yolk at late time points were due to viral replication, we reanalyzed these RNA samples by performing reverse transcription with a primer that hybridizes to the 5′ leader sequence of negative-strand subgenomic RNAs, a hallmark of active SARS-CoV-2 replication (Kim et al., 2020; Wölfel et al., 2020). Antisense transcripts are expected to be less abundant than sense transcripts in infected cells and absent from virions. Such transcripts were detected in the initial inoculum but in lower amounts than polyadenylated transcripts (median values of 1,042 and 191 copies for coelom-injected larvae with viral suspensions 1 and 2, respectively). In coelom-injected larvae, these antisense transcripts decreased and became undetectable at 48 h postinjection (hpi). By contrast, in yolk-injected larvae, levels were stable (Figure S1A). Therefore, both sense and antisense viral RNA molecules appeared to be protected from degradation in the yolk, and there was no clear evidence for viral replication. Notably, we did not observe yolk opacity in injected animals, a hallmark of yolk cell infection with other viruses such as CHIKV (Palha et al., 2013) and Sindbis virus (SINV) (Figure S2).

We then tested microinjection of SARS-CoV-2 in the swim bladder (SB), which inflates at 3.5–4 dpf (Parichy et al., 2009). We noticed that when the liquid was injected at the rostral end of the bladder, it was rapidly expelled via the pneumatic duct connecting the SB to the esophagus. By contrast, when liquid accumulated at the caudal end of the SB, if was well retained (Figure 3A). Therefore, injections were performed at 4 dpf by targeting the caudal half of the bladder; larvae with liquid injected at the rostral pole were discarded. As age-matched controls, we also injected 4-dpf larvae in the coelomic cavity, i.e., just next to, but outside of, the SB (Figure 3A).

Remarkably, in SB-infected larvae, after an initial decrease of viral transcripts during the first 24 h, their levels stabilized from 24 to 48 h; in contrast, the decline continued in coelom-injected larvae (Figure 3B). However, no disease signs were observed. We repeated the SB injection several times finding consistent results with a repeated trend of small re-increase from 24 to 48 h; extending the experiment by 1 day yielded comparable results at days 2 and 3 (Figure 3C). We also measured antisense transcripts in these larvae, observing the same trend (Figure S1B).

To perform statistical analysis with reasonable power, we normalized the results of each independent experiment to the mean of the values measured just after inoculation and then pooled the results by injection type. Because the dispersion increased considerably with time, we performed tests that allowed for unequal SDs when comparing time points. This analysis confirmed that after injection in the coelomic cavity, viral RNA amounts decline from 0 to 24 hpi and again from 24 to 48 hpi. By contrast, values measured in yolk were stable. In the SB, a very significant decrease is observed during the first 24 hpi, while from 24 to 48 hpi, a non-significant re-increase of the means is observed (Figure 4A). Comparison between the coelom and the SB showed a significantly higher level of viral RNA in the latter at 48 (but not 24) hpi (Figure 4B). These results clearly establish that while continuous viral RNA degradation occurs in the coelomic cavity, different events take place in the SB, which can be interpreted in several ways. One possibility could be that after 24 h of rapid decay of viral RNAs in the SB, this degradative reaction stops. Alternatively, de novo production could be taking place as a consequence of successful or abortive infection, compensating for ongoing degradation but not at a sufficient level to result in a detectable increase. Since antisense RNA are the first viral RNA species produced during the viral cycle, we analyzed the pooled normalized qPCR measurements of the antisense viral transcripts and found a significant increase of these transcripts from 24 to 48 hpi (Figure 4C). This increase of antisense but stagnation of sense viral RNAs suggests that some cells of the SB were invaded by SARS-CoV-2 but that the viral replication cycle was incomplete, resulting in abortive infection.

To confirm infection by SARS-CoV-2 after SB injection, we used whole-mount immunohistochemistry (WIHC). We tested several commercial Abs against the SARS-CoV-2 nucleoprotein and selected a mouse Mab with minimal non-specific staining of naïve larvae, except for dots in the notochord that we routinely observe and which are due to the secondary antibody only (Levraud et al., 2009). As an anatomical reference, we also labeled glial fibrillary acidic protein (GFAP), to reveal glial cells and main nerves. In most virus-inoculated larvae at 2 dpi, a patchy signal for N could be clearly detected in the SBs which were partially collapsed due to the fixation and staining procedure (Figures 5B–D). 3D reconstruction (Supplementary Video S1) suggests that these signals correspond to a few infected cells in the bladder wall, generally located close to the rear pole. The signal was detected in 5/7 virus-injected larvae, and absent in 3/3 controls, but since this number was too low for statistical analysis, the experiment with repeated (without the GFAP stain) with a larger number of replicates (Figure S3) and at a higher resolution. A positive signal was detected in 11/14 samples, and only 1/14 among negative controls, with a highly significant statistical difference (p = 0.0003, Fisher’s exact test), based on a blind assessment by a naïve observer. A 3D reconstruction of these samples confirmed localization of signal to the collapsed SB walls (Supplementary Video S2), and their relationship with nuclear staining, as visualized on confocal slices, was consistent with N protein in the cytoplasm of at least some cells (Supplementary Video S3). To ensure that the signal was detected inside cells of the SB wall, we counterstained these samples with the lipophilic stain DiI and reimaged them. When the anti-N-labeled structures were parallel to the imaging plane, numerous DiI-stained spheroids were observed to be scattered between the N-containing spots (Figure 5E and Supplementary Video S4). This is reminiscent of the formation of vesicular replication factories in SARS-CoV-2-infected cells (Eymieux et al., 2021), but because of the lower resolution in the Z-axis, overlying and underlying plasma membranes were ambiguous in these cases. However, by imaging bladder walls where they are perpendicular or strongly oblique to the imaging plane, it was possible to determine the relative positions of N-positive patches and plasma membranes. Some of these patches appeared to be affixed to the bladder wall without a membrane separating them from the lumen (Figure 5F and Supplementary Video S4); however, others were clearly surrounded by membranes (Figures 5G–I and Supplementary Videos S4, S5) and were therefore localized within the cytoplasm of swim bladder epithelial cells.

We then tested a series of SARS-CoV-2 variants by SB inoculation. We obtained aliquotes from early passages after isolation of clinical strains, which had been titered at 3.10⁷ PFU/ml or more and thus did not require further concentration. We tested the alpha variant (formerly known as UK variant, or B1.1.7), the beta variant (South African variant, B1.351), and the gamma variant (Brazilian variant, P1) as well as a representative of the G-clade which arose early during the pandemic. Non-diluted viral suspensions were injected as described above in the SB of 4-dpf larvae and were then monitored for 2 days; no clinical signs were observed. Viral replication was assessed by qRT-PCR. A global decline of polyadenylated N transcripts over time was observed with all variants (Figure 6). One unique larva injected with the gamma variant was found to contain slightly more N copies than the initial inoculum; therefore, the experiment was repeated for the gamma variant, and again, one larva did not show the same decline as others. Thus, results obtained with the gamma variant were comparable to those obtained with the initial strain, with a fraction of larvae in which some replication appeared to take place. No replication was found with the other strains, which also corresponded to lower inocula according to qPCR results. Overall, we saw no evidence for an increased infectivity of SARS-CoV-2 variants in zebrafish larvae.

Type I interferons (IFNs) are key antiviral cytokines in vertebrates, including teleost fish. We thus tested if SARS-CoV-2 may replicate in larvae with a crippled type I IFN response.

First, we used morpholino-mediated knockdown of the type I IFN receptor chains CRFB1 and CRFB2, known to make zebrafish larvae hypersusceptible to infection with CHIKV or SINV (Palha et al., 2013; Boucontet et al., 2018). After injection of SARS-CoV-2 in the coelom of 3-dpf larvae, decline of N transcripts was found to be similar in IFNR-knocked-down larvae than in controls (Figure 7A).

To ensure a long-lasting suppression of the IFN response, we used a newly generated mutant zebrafish line dubbed “triple ϕ,” in which the three type I IFN genes ifnphi1, ifnphi2, and ifnphi3, tandemly located on chromosome 3, have been inactivated by CRISPR. Heterozygous triple ϕ mutants were viable and fertile; incrossing them yielded homozygous embryos at the expected Mendelian ratio of ~25%. Homozygous triple ϕ mutants could be raised up to juvenile stage, but, unlike their siblings, died in the 2 weeks following genotyping by fin clipping. To validate the phenotype of the mutants, we injected SINV-GFP to 3-dpf larvae from a heterozygous incross. 48 h later, all larvae were alive although some showed strong signs of disease, including loss of reaction to touch, abnormal heartbeat, slow blood flow, edemas, and opacified yolk spots. All larvae were imaged with a fluorescence microscope to measure the extent of infection, then lysed individually and genotyped. The homozygous mutants displayed a considerably higher level of fluorescence (Figure 7B) and were also identified a posteriori as the sickest larvae, confirming that triple ϕ mutants are hypersusceptible to viral infection.

Larvae from triple ϕ heterozygous incrosses were thus injected with SARS-CoV-2, either in the coelomic cavity at 3 dpf or in the SB at 4 dpf. Larvae were lysed at 48 hpi, analyzed by qRT-PCR, and genotyped. Consistent with previous results, a 100-fold decrease of viral RNA was observed in coelom-injected larvae, while a weaker decrease was observed for SB injection, with a bimodal distribution suggesting that infection happened in about one-third of cases. In both situations, viral loads in homozygous triple ϕ mutants were not different from their wild-type siblings (Figure 7C). Thus, our results indicate that type I IFN responses are not responsible for the lack of replication of SARS-CoV-2 observed in wild-type zebrafish larvae.

We then tested if SARS-CoV-2 inoculation in the SB resulted in induction of a type I interferon response or inflammatory cytokines. For this, we performed qPCR on dT17-primed cDNAs from whole larvae. Based on our previous results (Levraud et al., 2019; Kraus et al., 2020), we tested the main type I interferon genes inducible in larvae, namely, ifnphi1 and ifnphi3; the strongly IFN-inducible gene MXA; the classical inflammatory cytokines il1b and tnfa; cytokines that reflect induction of type 2 or type 3 responses, il4 and il17a/f3, respectively; and chemokines ccl19a.1 and ccl20a.3. Although individual experiments suggested significant induction of ifnphi1 at 48 hpi or il17a/f3 at 72 h, this could not be replicated; as shown on Figure 8, in which data from 4 independent experiments have been pooled, no significant change in expression of any of these genes can be observed compared to uninjected control larvae. Similar negative results were obtained with larvae injected at different sites (not shown).

Although these results do not exclude a local response to SARS-CoV-2, they are in striking contrast with those we obtained previously in larvae infected with other pathogens such as SINV or Shigella flexneri, for which many of these genes were induced more than 100-fold (Boucontet et al., 2018). Since these experiments had been performed at 28°C, we verified that zebrafish larvae are also able to mount a strong type I response at 32°C (Figure S4).

Finally, we tested if mosaic overexpression of human ACE2 in zebrafish larvae would increase their infectivity of SARS-CoV-2. We subcloned the hace2 ORF in fusion with mCherryF under the control of the promoter of the ubiquitous ribosomal protein RPS26. In addition, the fragment is flanked by two inverted I-SceI meganuclease sites for higher transgenesis efficiency (Grabher et al., 2004). In order to be sure that the in-frame fusion of hACE2 with mCherry would not interfere with SARS-CoV-2 binding to its receptor and entry in the target cells, another construct was done by inserting a self-cleaving 2A peptide between hACE2 and mCherry ORFs. Both constructs were tested in BHK cells and increased by ~100-fold their infection by SARS-CoV-2 (Figure 9A). We optimized the injected dose of plasmid; 68 pg was the amount yielding the highest mCherry expression without increasing the proportion of misshapen embryos (Figure 9B). In 24-hpf embryos, many mCherry⁺ cells, randomly distributed, were visible in these embryos under the fluorescence microscope. In swimming larvae, mCherry⁺ cells were still clearly visible but in lower amounts (Figure 9C). To get a quantitative assessment of their frequency, we dissociated 4-dpf larvae and analyzed the suspension by flow cytometry, which indicated that ~0.5% of the cells were mCherry⁺ (Figure 9D). Larvae were fixed and processed by immunohistochemistry, which confirmed ACE2 expression at the membrane of mCherry⁺ cells (Figure 9E).

Zebrafish AB eggs were injected with the plasmid, and at 3 dpf, the 25% larvae displaying the highest mCherry expression and good morphology were selected. They were then microinjected with SARS-CoV-2 in the coelom or the brain ventricle and processed as above. qRT-PCR analysis revealed that viral mRNA transcripts decreased just as it did in AB larvae (Figure 10). Thus, this approach did not increase the infectivity of SARS-CoV-2 in zebrafish larvae.

We finally tested if hACE2 overexpression by in vitro cultured fish cells made them susceptible to SARS-CoV-2, using the cyprinid cell line EPC. EPC cells were co-transfected with GFP and hACE2 expression plasmids; transfection efficiency and membrane hACE2 expression was verified by IHC (Figures 11A, B). These transfected cells were incubated with active or heat-killed SARS-CoV-2 at a MOI of 0.1 and then tested for viral replication by qRT-PCR on cell lysates. No difference was observed between GFP-only and GFP+hACE2-expressing cells (Figure 11C); furthermore, the amount of N transcripts fell dramatically from day 0 to day 2 (Figure 11D), showing that hACE2-expressing EPC cells were not able to support SARS-CoV-2 replication.

Animal experiments described in the present study were conducted according to European Union guidelines for handling of laboratory animals (http://ec.europa.eu/environment/chemicals/lab_animals/home_en.htm) and were approved by the Ethics Committee of Institut Pasteur.

Wild-type zebrafish (AB strain), initially obtained from ZIRC (Eugene, OR, USA), were raised in the aquatic facility of Institut Pasteur. After natural spawning, eggs were collected, treated for 5 min with 0.03% bleach, rinsed twice, and incubated at 28°C in Petri dishes in Volvic mineral water supplemented with 0.3 µg/ml methylene blue (Sigma-Aldrich, St. Louis, Missouri, USA). After 24 h, the water was supplemented with 200 µM phenylthiourea (PTU, Sigma-Aldrich) to prevent pigmentation of larvae. After this step, incubation was conducted at 24°C, 28°C, or 32°C depending on the desired developmental speed. Developmental stages given in the text correspond to the 28.5°C reference (Kimmel et al., 1995). Sex of larvae is not yet determined at the time of experiments.

Triple type I interferon CRISPR mutants have been generated by the AMAGEN transgenesis platform (Gif-sur-Yvette, France) by co-injection of CAS9 with two sgRNA targeting ifnphi1 (target sequence, GCTCTGCGTCTACTTGCGAAtgg) and ifnphi2 (target sequence, ATGTGCGCGAAAAAGAGTGCtgg) in one-cell eggs from homozygous ifnphi3^ip7/ip7-null mutants of the AB background (Maarifi et al., 2019). After growth to adulthood, a founder was identified that co-transmitted mutations in ifnphi1 and ifnphi2 in addition to the ip7 mutation of ifnphi3. The ip9 allele mutation in ifnphi1 consists in a 7-base pair deletion in the first exon of the secreted isoform (GAATGGC, 75 bases downstream of the start ATG). The ip10 allele in ifnphi2 consists in a 19-bp deletion in the first exon (TGCGTTCTTATGTCCAGCA, 20 bases downstream of the start ATG). This founder was crossed with AB fish, and F1 fish triply heterozygous for mutations ip7, ip9, and ip10 were selected to establish the “triple φ” mutant line. As expected, since ifnphi1, ifnphi2, and ifnphi3 are closely located in tandem on a 35-kb region of zebrafish chromosome 3, the ip7, ip9, and ip10 mutations were always found to co-segregate. Genotyping PCR primers are listed in Table 2.

The main SARS-CoV-2 stock used (BetaCoV/France/IDF0372/2020 strain) was propagated twice in Vero-E6 cells and is a kind gift from the National Reference Centre for Respiratory Viruses at Institut Pasteur, Paris, headed by Dr Sylvie van der Werf; this strain was isolated from a human sample provided by Drs. Xavier Lescure and Yazdan Yazdanpanah from the Bichat Hospital, Paris. To generate concentrated virus, Vero-E6 cells were infected with virus at an MOI of 0.01 PFU/cell in DMEM/2% FBS and incubated for 72 h at 37°C, 5% CO₂. At this point, the cell culture supernatant was harvested, clarified, and concentrated using Amicon Ultra-15 Centrifugal units 30K (Merck Millipore, Burlington, MA, USA). Virus titers were quantified by plaque assay in Vero-E6.

The variant strains used were also supplied by the National Reference Centre for Respiratory Viruses at Institut Pasteur and were used directly without further propagation. The G-clade (BetaCoV/France/GE1973/2020; 3 × 10⁷ PFU/ml), alpha (hCoV-19/France/IDF-IPP11324i/2020; 6.75 × 10⁷ PFU/ml), beta (hCoV-19/France/PDL-IPP01065i/2021; 1.75 × 10⁸ PFU/ml), and gamma (hCoV-19/French Guiana/IPP03772i/2021; 5.53 × 10⁷ PFU/ml) variants were isolated from human samples provided respectively by Dr. Laurent Andreoletti, from Robert Debré Hospital, Reims, France; Dr. Foissaud, HIA Percy, France; Dr. Besson J. from Bioliance Laboratory, France; and Dr. Rousset, Institut Pasteur, Cayenne, French Guiana.

The SINV-GFP virus corresponds to the SINV-eGFP/2A strain described in Boucontet et al. (2018) and was used as a BHK cell supernatant at 2 × 10⁷ PFU/ml. The eGFP sequence is inserted with self-cleaving sequences between the capsid and envelope genes of SINV.

Bath exposures were conducted in a 12-well plate with 4 larvae per well in 2 ml of water plus PTU. 2-dpf embryos were manually dechorionated previously. 10 µl of SARS-CoV-2 suspension 2 (either freshly thawed or heat-inactivated for 5 min at 70°C) was added to each well and gently mixed, then the plates were incubated at 32°C. After a given incubation time, larvae were deeply anesthetized with 0.4 mg/ml tricaine (MS222, Sigma-Aldrich), rinsed twice in 10 ml of water, transferred individually into tubes, and after removal of almost all water, lysed in 320 µl of RLT buffer (Qiagen, Hilden, Germany) supplemented with 1% β-mercaptoethanol (Sigma-Aldrich).

After addition of RLT buffer, larvae were dissociated by 5 up- and-down-pipetting movements. Tubes may then be frozen at -80°C for a few days. Before export from the BSL3 laboratory, RLT lysates were incubated at 70°C for 5 min to ensure complete virus inactivation (preliminary tests confirmed that this had a negligible impact on qRT-PCR results). Total RNA was then extracted with an RNeasy Mini Kit (Qiagen) without the DNase treatment step and a final elution with 30 µl of water.

RT was performed on 6 µl of eluted RNA using MMLV reverse transcriptase (Invitrogen, Carlsbad, CA, USA) with either a dT₁₇ primer (for polyadenylated transcripts) or the SgleadSARSCoV2-F primer (for negative-strand viral transcripts) (Wölfel et al., 2020) (Table 2). cDNA was diluted with water to a final volume of 100 µl, of which 5 µl was used as a template for each qPCR assay.

Real-time qPCR was performed with an ABI7300 (Applied Biosystems, Foster City, CA, USA). Quantitation of sense or antisense viral N transcripts was performed by a TaqMan probe assay, using the primer–probe mix from the 2019-nCoV RUO kit (IDT, Coralville, IA, USA) with the iTaq Universal Probes One-Step kit (Bio-Rad, Hercules, CA, USA). The 2019-nCoV_N_Positive Control plasmid (IDT) was used as a standard for absolute quantification. Quantification of zebrafish transcripts was performed using a SYBR assay using the Takyon SYBR Blue MasterMix (Eurogentec, Seraing, Belgium) with primer pairs listed in Table 2. These primers typically span exon boundaries to avoid amplification of contaminating genomic DNA. For absolute quantification of the housekeeping gene rps11, a standard was produced by PCR using primers to amplify a fragment including the whole open-reading frame, which was gel-purified and quantified by spectrophotometry. Ratios of other transcripts to rps11 were estimated by the 2^ΔCt method.

Morpholino antisense oligonucleotides (Gene Tools, Philomath, OR, USA) were injected (1 nl volume) in the cell or yolk of AB embryos at the one- to two-cell stage as described (Levraud et al., 2008). crfb1 splice morpholino (2 ng, CGCCAAGATCATACCTGTAAAGTAA) was injected together with crfb2 splice morpholino (2 ng, CTATGAA TCCTCACCTAGGGTAAAC), knocking down all type I IFN receptors (Aggad et al., 2009). Control morphants were injected with 4 ng control morpholino, with no known target (GAAAGCATGGCATCTGGAT CATCGA).

Expression plasmids, produced using an endotoxin-free kit (Macherey-Nagel, Düren, Germany), were co-injected with the I-SceI meganuclease (Grabher et al., 2004). Briefly, 12.5 µl of plasmid is mixed with 1.5 µl of CutSmart Buffer and 1 µl of I-SceI (New England Biolabs, Ipswich, MA, USA) and incubated at room temperature for 5 min before being put on ice until injection of 1 nl inside the cell of AB embryos at the one-cell stage.

SINV-GFP-infected or hACE2-mCherry-expressing larvae were imaged with an EVOS FL Auto microscope (Thermo Fisher Scientific, Waltham, MA, USA) using a 2× planachromatic objective (numerical aperture, 0.06), allowing capture of the entire larva in a field. Transmitted light and fluorescence (GFP or Texas Red cube) images were taken. They were further processed (superposition of channels, rotation, crop, and fluorescence intensity measure) using Fiji. Mean background fluorescence of uninjected control animals was subtracted from the measured signal to obtain the specific fluorescence.

Pools of 10 larvae were dissociated by a combination of mechanical trituration (repeated pipetting) and enzymatic treatment at 30°C, first with 200 µl of 0.25% Trypsin-EDTA (Gibco, Grand Island, NY, USA) for 10 min, and 10 more min after addition of 10% sheep serum, CaCl₂ to 2 µM, and 1 µl of 5 mM collagenase (C9891, Sigma). Cell suspensions were then washed with PBS 1×, pelleted, resuspended in PBS, and filtered on a 40-µm mesh. Dead cells were labeled with SYTOX AADvanced (Thermo Fisher). Cell suspensions were acquired on an Attune NxT flow cytometer (Thermo Fisher) with blue and yellow lasers, and data analyzed with FlowJo (Ashland, OR, USA).

The Epithelioma papulosum cyprini cell line (EPC) was maintained in Leibovitz’s 15 media (L-15, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco), 100 µg/l penicillin, and 100 µg/ml streptomycin. EPC cells were cultured at 32 °C without CO₂.

BHK-21 cells were maintained in DMEM (Gibco) supplemented with 5% FBS and 2 mM L-glutamine and maintained at 37°C with 5% CO₂.

The hACE2 ORF was amplified from clone IOH80645 (Thermo Fisher, GenBank NM_021804.2) using primers hACE2NotStart3 and hACE2EndNot3 (Table 2). The amplified PCR fragment was digested by NotI and inserted in the NotI site of the Tol2S263C:mC-F vector between the promoter of the zebrafish ubiquitous ribosomal protein RPS26 encoded by chromosome 3, and the mCherry ORF. In this construct, the RPS26 promoter drives the expression of a hACE2 protein fused at its C-term with farnesylated mCherry. In order for the ORF to drive the expression of two separated proteins (hACE2 and mCherry-F), primers hACE2NotStart3 and hAce2.2ANot were used to amplify the hACE2 ORF from the IOH80645 clone. The amplified fragment was digested by NotI and cloned in the NotI site of Tol2S263C:mC-F between the promoter of the zebrafish ubiquitous ribosomal protein RPS26 encoded by chromosome 3, and the mCherry ORF. Maps and sequences of plasmids are available at https://doi.org/10.5281/zenodo.4672028.

For in vitro transfection of EPC cells, plasmid pcDNA3.1-hACE2 (Addgene, Watertown, MA, USA, #1786) was directly used along plasmid pmEGFP-N1 (Chen and Reich, 2010).

EPC cells were electroporated with the Neon transfection system (Invitrogen). Briefly, EPC cells were trypsinized and resuspended in L15 media supplemented with FBS and antibiotics. Cells were counted and centrifuged at 2,000 rpm for 5 min. 0.8 × 10⁶ cells per condition were resuspended in 80 µl of L15 without phenol red with 2.4 µg of each plasmid. Cells were electroporated using 10-µl neon tips with 1 pulse of 1,700 V during 20 ms. Electroporated cells were plated in a 6-well plate in L15 + FBS + antibiotics and incubated 3 days at 32°C before experiment.

BHK21 were transfected with Lipofectamine 2000 (Invitrogen). Briefly, cells at 80% confluence in 12-well plates were incubated with 750 ng of plasmid and 1 µl of Lipofectamine in Opti-MEM (Gibco). Transfection efficiency was checked at 1 day post-transfection.

Transfected EPC cells were transferred to BSL3 laboratory for infection with SARS-CoV-2. Cells were rinsed with L15 media + FBS + antibiotics and incubated 5 min at 32°C. Cells were infected at MOI 0.1 with virus diluted in L15 media + FBS + antibiotics and incubated at 32°C during 1 h with agitation. After incubation, L15 media + 10% FBS + antibiotics was added and cells were incubated 2 days at 32°C or processed directly for RNA extraction.

Transfected BHK21 cells were transferred to BSL3 laboratory for infection with SARS-CoV-2. Cells were rinsed with DMEM + 5% FBS and incubated for 5 min at 37°C + 5% CO₂. Cells were infected with virus at MOI 0.1 diluted in DMEM + 5% FBS and incubated at 37°C + 5% CO₂ during 1 h with agitation. After incubation, DMEM + 5% FBS was added and cells were incubated 2 days at 32°C or processed directly for RNA extraction.

Before RNA extraction, culture medium was removed and cells were rinsed once with PBS. Extraction of total RNA was performed using TRI Reagent (Sigma) following the manufacturer’s recommendations. Total RNA was resuspended in 100 µl of RNase-free water.

Reverse transcription was performed on 5 µl of RNA suspension using the QuantiTect Reverse Transcription Kit (Qiagen) with either the Qiagen RT Primer Mix or the SgleadSARSCoV2-F primer (for negative-strand viral transcripts) (Wölfel et al., 2020). cDNA was diluted with water to a final volume of 50 µl, of which 2.5 µl was used as a template for each qPCR assay.

Real-time qPCR was performed with a RealPlex 2 (Eppendorf). Quantitation of viral N transcripts was performed by a TaqMan probe assay, using the primer–probe mix from the 2019-nCoV RUO kit (IDT) with the iTaq Universal Probes kit (Bio-Rad). Quantitation of actin transcripts was performed by a SYBR Green assay, using primers specific for fathead minnow β-actin (Table 2) with iTaq Universal SYBR Green Supermix (Bio-Rad).

Whole-mount immunohistochemistry of larvae was performed essentially as described in Palha et al. (2013) and Santos et al. (2018). For COV2-N detection, additional treatment with glycine 0.3 M in PBST (30 min at RT) and heat-induced antigen retrieval (HIER) were performed. HIER treatment was performed in 150 mM Tris–HCl, pH 9.0, at 70°C for 15 min. Primary Ab antibodies used for this labeling were mouse anti-SARS-CoV-2 nucleoprotein (Sino Biological, Beijing, China, 40143-MM05, 1:100) and rabbit anti-GFAP (GeneTex, Irvine, CA, USA, GTX128741, 1:100). As secondary Ab antibodies used were goat anti-mouse F(ab)′2 Alexa Fluor 488 (Molecular Probes, Eugene, OR, USA, A11017, 1:300) and goat anti-rabbit Cy3 (Jackson Laboratories, Bar Harbor, ME, USA, 111-166-003, 1:300). Furthermore, to label the nuclei NucRed Live 647 (Thermo Fisher, R37106, 4 drops for mL for 45 min) was used. For hACE2 detection, staining was performed sequentially since both the primary Ab for ACE2 and the secondary Ab for mCherry were from goat. Primary staining for ACE2 (goat anti-ACE2, AF933, R&D Systems, Minneapolis, MN, USA, 4 µg/ml) was performed first, followed by its secondary staining (donkey anti-goat Ig Alexa 488, A100555, Invitrogen, 1:300), then primary staining for mCherry (rabbit anti-DsRed, 632393, Clontech, Mountain View, CA, USA, 1:300) and secondary staining (goat anti-rabbit Ig Cy3, 111-166-003, 1:300). Nuclei were labeled with 2 µg/ml Hoechst 33342 (Invitrogen). To label membranes, we used DilC18(3) (Thermo Fisher Scientific, D282) staining at 1 µM working solution for 24 h at 4°C, followed by 6 × 45-min washes in PBST.

Afterward, IHC larvae were conserved in 80% glycerol until acquisition. For acquisition of N-CoV-2, the larvae were mounted in 2% agarose in 80% glycerol singularly in a glass-bottom 8-well slide (ibidi, Gräfelfing, Germany, 80827). Images were acquired using an inverted confocal microscope Leica SP8 using a ×10 objective zoomed 1.25× (PL FLUOTAR ×10/0.30 DRY) and ×20 immersion objective (HC PL APO CS2 ×20/0.75 multi-IMM). For both magnification, the bidirectional resonant scanning method was used and images were deconvolved using Leica Lightning Plug-in. For acquisition of hACE2, images were acquired on an upright Leica SPE confocal microscope using a ×40 oil objective (numerical aperture, 1.15).

For IHC of in vitro transfected cells, EPC were cultured in a 6-well plate containing sterilized coverslips. At 3 days post-transfection, culture media were removed, and cells were rinsed with PBS once. Cells were fixed overnight at 4°C with 4% methanol-free formaldehyde (Sigma) in PBS. Formaldehyde was removed, and cells were rinsed twice with PBS and kept at 4°C in PBS + 0.05% sodium azide. Fixed cells were rinsed 3 times in PBS. Cells were then permeabilized and blocked with PBS + 0.3% Triton X-100 + 10% horse serum during 45 min at RT. Cells were stained for 1 h at RT with a goat polyclonal anti-human ACE2 (AF933, R&D Systems) diluted at 3 µg/ml in PBS + 0.3% Triton X-100 + 1% horse serum + 1% BSA + 0.01% sodium azide. Cells were then rinsed and stained during 1 h at RT with Alexa 647 anti-goat diluted at 1/500 in PBS + 0.3% Triton X-100 + 1% horse serum + 1% BSA + 0.01% sodium azide. After 3 rinses with PBS, cells were incubated for 1 h at RT with DAPI diluted at 2.5 µg/ml in PBS. After 3 rinses in PBS, coverslips were mounted on slides with Fluoromount-G (Thermo Fisher Scientific).

Transfection efficiency was checked at 3 days post-transfection using a Zeiss Axio Observer Z1 widefield microscope with a ×10/NA 0.25 objective. The phase and GFP channel were acquired on a field of view of 5. Confocal acquisition of immunostained EPC cells was performed on a Leica SP8 upright microscope using a ×25/NA 0.95 coverslip-corrected objective. Endogenous GFP and Alexa 647 were excited with 488 and 638 nm, respectively, and detected with PMT. Fiji was used to adjust the brightness and contrast of confocal images of immunostained EPC cells. Transfection efficiency was quantified using Fiji by manually counting total cells and GFP-expressing cells, respectively.

Analyses were performed with GraphPad Prism. Methods used are indicated in Figure legends. Normality/log-normality tests of data distribution were performed to decide the most appropriate assays.