As an alternative and complementary approach to Cas9-based genome editing, Cas12a has not been widely used in mammalian cells largely due to its strict requirement for the TTTV protospacer adjacent motif (PAM) sequence. Here, we report that Mb3Cas12a (Moraxella bovoculi AAX11_00205) can efficiently edit the mouse genome based on the TTV PAM sequence with minimal numbers of large on-target deletions or insertions. When TTTV PAM sequence-targeting CRISPR (cr)RNAs of 23 nt spacers are used, >70% of the founders obtained are edited. Moreover, the use of Mb3Cas12a tagged to monomeric streptavidin (mSA) in conjunction with biotinylated DNA donor template leads to high knock-in efficiency in two-cell mouse embryos, with 40% of founders obtained containing the desired knock-in sequences.
The rapid advancement of CRISPR-Cas-based genome editing technologies has made their clinical applications increasingly promising. However, major obstacles still remain, including safety concerns due to both off-target effects and large on-target deletions or insertions (Adikusuma et al., 2018; Fu et al., 2013; Hsu et al., 2013; Kosicki et al., 2018; Lee and Kim, 2018), and the requirement of proper PAM sequences for efficient and precise cleavage by the commonly used endonucleases, for example, Cas9 and Cas12a/Cpf1 (Komor et al., 2017). The Cas12a endonuclease differs from Cas9 in many aspects and thus, has been considered as an alternative and complementary tool to Cas9. First, the most commonly used Streptococcus pyogenes (Sp)Cas9 and Cas12a [Acidaminococcus sp. BV3L6 (As)Cas12a and Lachnospiraceae bacterium ND2006 (Lb)Cas12a] utilize the NGG and TTTV PAM sequences, respectively, for efficient genome editing (Cong et al., 2013; Jinek et al., 2012; Wang et al., 2013; Zetsche et al., 2015). Second, Cas12a is guided by a single short CRISPR RNA (crRNA), and can efficiently process its own crRNAs, while Cas9 is directed by dual RNAs consisting of a crRNA and a tracrRNA or by a joint large single-guide RNA (sgRNA), and rarely processes its own crRNAs (Cong et al., 2013; Fonfara et al., 2016; Jinek et al., 2012; Zetsche et al., 2017a). Third, Cas12a displays reduced off-target effects compared to SpCas9 due to its irreversible binding to the target region and strong discrimination against off-target sequences (Kim et al., 2017; Kleinstiver et al., 2016; Strohkendl et al., 2018). Finally, Cas9 causes large on-target deletions or insertions (Adikusuma et al., 2018; Kosicki et al., 2018; Lee and Kim, 2018), whereas Cas12a only generates staggered DNA overhangs (Zetsche et al., 2015), which lead to a much lower rate of large on-target deletions or insertions due to the preferred microhomology-mediated end joining (MMEJ) repair mechanism. Despite the advantages, practical applications of Cas12a have been severely hindered due, at least in part, to its strict requirement for the TTTV PAM sequence. Although the PAM sequence for Francisella novicida U112 (Fn)Cas12a has been shown to be YTV (Y stands for C/T and V for A/C/G), the actual editing efficiency of the YTV PAM sequences in mammalian cells remains rather low (Tu et al., 2017; Zetsche et al., 2015). Inspired by a recent report suggesting that Mb3Cas12a (Moraxella bovoculi AAX11_00205) edits HEK-293 cells at a much higher efficiency through the TTV PAM sequences than does AsCas12a and LbCas12a (Zetsche et al., 2017b preprint), we explored whether Mb3Cas12a could be utilized for efficient genome editing to produce knockout (KO) and knock-in (KI) mouse lines.
RESULTS AND DISCUSSION
To determine whether Mb3Cas12a is active in mouse zygotes, we first tried to use one crRNA harboring a 20 nt direct repeat (DR) sequence with a 20 nt spacer recognizing the TTTV PAM sequence of Prps1l1 (Table 1; Fig. S1A,D), a testis-specific gene dispensable for spermatogenesis (Wang et al., 2018). One out of six founders obtained had an indel (insertion or deletion) (16.7%) (Table 1; Fig. S1A,D), suggesting that Mb3Cas12a indeed works in mouse zygotes. Previous studies have shown that FnCas12a, AsCas12a and LbCas12a are capable of processing their own crRNAs (Fonfara et al., 2016; Zetsche et al., 2017a), and that residues H843, K852, K869 and F873 in FnCas12a are required for crRNA processing (Fonfara et al., 2016). Given that these four residues are highly conserved among Mb3Cas12a and its three orthologs (FnCas12a, AsCas12a and LbCas12a) (Fig. S1E), we hypothesized that the Mb3Cas12a could also process its own crRNAs. To test this, we designed crRNAs harboring two 20 nt spacers recognizing two TTTV PAM sequences or two TTV PAM sequences at the Saraf locus (Table 1; Fig. S1B,F). The 20 nt spacers were separated by two 20 nt DRs. One out of three founders was edited by spacer 1 targeting the TTTV PAM sequence (33.3%) from the first crRNA, and one out of four founders was edited with the spacer 2 (25%) targeting the TTV PAM sequence from the second crRNA (Table 1; Fig. S1B,F). Next, we designed one single crRNA containing two spacers recognizing the TTTV and TTV PAM sequences in the Mrvi1 locus, respectively (Table 1, Fig. S1C,G). Out of five founders, one was edited with the spacer recognizing the TTTV PAM sequence (20%), but none for the TTV PAM sequence (Table 1; Fig. S1C,G). These results suggest that Mb3Cas12a can target genomic DNA with either the TTTV PAM or the TTV PAM sequence.
Although Mb3Cas12a can edit the genome in mouse zygotes, the efficiency appeared to be low. We sought to optimize the editing efficiency by adjusting either the crRNA length or microinjection sites. We first tested the effects of crRNA length on genome editing in the DNMT1 locus in HEK-293 cells (Fig. 1A,B). No Mb3Cas12a activities were detected when no spacer or a 17 nt spacer was used, whereas the highest Mb3Cas12a activities, comparable to those for AsCas12a and LbCas12a, were observed in the crRNAs with a 23 nt spacer (Fig. 1B). Although FnCas12a uses 21 nt crRNAs more efficiently (Tu et al., 2017), Mb3Cas12a, similar to AsCas12a and LbCas12a, appears to prefer 23 nt crRNAs. We then examined the potential effects of microinjection sites (cytoplasmic versus pronuclear) on targeting efficiency by injecting Mb3Cas12a mRNA and one Kcnj10-targeting crRNA harboring two spacers (one recognizing the TTTV PAM sequence, and the other specific for the TTV PAM sequence) into either the cytoplasm or both the pronucleus and cytoplasm of the zygotes. Interestingly, the cytoplasmic-only microinjection yielded more edited mice (77%, n=20) than the combined (pronuclear+cytoplasmic) microinjection (28.6%, n=39) (Table 1, Fig. 1C,D). Taken together, our data suggest that the use of crRNAs with a 23 nt spacer in conjunction with cytoplasmic microinjection represents a highly efficient method to generate indels using Mb3Cas12a in murine zygotes.
To explore whether Mb3Cas12a could target multiple loci simultaneously, we used one crRNA harboring three 23 nt spacers recognizing the TTTV PAM sequences in three Prps family genes, including Prps1, Prps2 and Prps1l1. Eleven founders obtained were all edited by both the Prps1 and Prps1l1 crRNAs (100%), whereas one (9.1%) was edited only by the Prps2 crRNA (Table 1; Fig. S2A). These results suggest that Mb3Cas12a can indeed process its own crRNAs and edit multiple loci in murine zygotes. Next, we used a crRNA containing three 23 nt spacers separated by three 20 nt DRs to target miR-10b (Table 1, Fig. S2B,C). Two of the spacers (spacers 2 and 3) were designed to recognize the TTTV PAM sequence, whereas the other targets the TTV PAM sequence (spacer 1) in the miR-10b locus. All five founders were edited by the two spacers targeting the TTTV PAM sequence (100%), and one of them was edited by the one recognizing the TTV PAM sequence (20%) (Table 1; Fig. S2B,C). Moreover, the edited alleles were successfully transmitted to the next generation (Fig. S2D). Similar results were obtained when we used one crRNA containing one 23 nt spacer targeting miR-547 with the TTTV PAM sequence and two 23 nt spacers targeting miR-509 with the TTTV PAM and the TTV PAM (Table 1, Fig. 2). All six founders were edited by the two spacers targeting the TTTV PAM sequence (100%), and one of them was edited by the one recognizing the TTV PAM sequence (16.7%) (Table 1, Fig. 2B,C). Although the 23 nt spacers recognizing the TTTV PAM sequence often led to bi- or multi-allelic targeting, and those recognizing the TTV PAM sequence often yielded mono-allelic targeting, the 20 nt spacers appeared to cause mostly mono-allelic mutations (Table 1, Figs 1D, 2C; Figs S1, S2). The mono-allelic targeting may reflect lower Mb3Cas12a editing efficiency when 20 nt spacers or TTV PAM sequences are used. Recent studies have shown that Cas9 with one single gRNA tends to induce large indels in genomic DNA in mouse embryonic stem cells (ESCs), progenitor cells and zygotes (Adikusuma et al., 2018; Kosicki et al., 2018), whereas Cas9 with two gRNAs often causes large deletions within the two flanking gRNA-targeting sites (Wang et al., 2018, 2020). However, Mb3Cas12a-based genome editing appears to generate predominantly indels, because 39 out of the 42 pups derived from Mb3Cas12a-based editing all displayed indels, and only three contained large insertions (7.1%) (Table 1, Figs 1D, 2C; Figs S1, S2). The lack of large on-target insertions or deletions in Mb3Cas12a-based genome editing suggests that the MMEJ mechanism may prefer staggered DNA ends for repair. Interestingly, the large insertions detected in the three founders appeared to be endogenous retroviruses type-L (ERVL) and mammalian apparent LTR retrotransposon (ERVL-MaLR) sequences. A similar phenomenon has been reported in SpCas9-induced double-stranded breaks (DSBs) (Ono et al., 2015), suggesting these retrotransposons may hijack the DSBs induced by either SpCas9 or Mb3Cas12a.
Two-cell homologous recombination (2C-HR)-CRISPR, where Cas9 is tethered with monomeric streptavidin (mSA), which binds to biotinylated DNA donor template, has been shown to have higher KI efficiency (Gu et al., 2018). Given that, we tested whether Mb3Cas12a-mSA could do the same in generating KI mice (Fig. 3A,B). We microinjected two-cell embryos with Mb3Cas12a-mSA mRNA, the Slit2-targeting crRNA and the biotinylated DNA donor template (Fig. 3A). Two out of five founders obtained (40%) contained the KI alleles (Fig. 3B,C). In contrast, microinjection of the non-biotinylated donor together with Mb3Cas12a-mSA mRNA and the Slit2-targeting crRNA generated no positive founders among 36 pups, although T7EI assays verified that Mb3Ca12a worked (Fig. S3A). These results indicate the Mb3Cas12a-mSA indeed can generate KI mice efficiently.
To test whether Mb3Cas12a could induce off-target effects, we predicted the off-targets using Cas-OFFinder (Bae et al., 2014) and performed T7EI assays on those predicted off-targets. T7EI assays revealed that Mb3Cas12a, just like other Cas12a endonucleases (Kleinstiver et al., 2019; Kocak et al., 2019), also has off-target effects (Fig. S3B–D). Further work is needed to improve its specificity, which may be achieved by either modifying the endonuclease (Kleinstiver et al., 2019) or by using hairpin gRNAs (Kocak et al., 2019). For example, a mutation in enhanced (en)AsCas12a (N282A) can turn it to a high-fidelity endonuclease, whereas a mutation in AsCas12a (R1226A) renders it the nickase that cleaves the non-target DNA strand (Kleinstiver et al., 2019; Yamano et al., 2016). Given that these two residues are conserved between AsCas12a and Mb3Cas12a (Fig. S3E), similar modifications in Mb3Cas12a may also enhance its fidelity, and also lead to the Cas12a version of prime editing, which utilizes the combined Cas9 H840A nickase and MMLV reverse transcriptase activities to minimize off-target effects (Anzalone et al., 2019). In summary, our data demonstrate that Mb3Cas12a can edit the murine genome in a manner that is independent of the TTTV PAM sequence and with minimal on-target mutations and high targeting efficiency. Mb3Cas12a-mediated genome editing expands the toolkit for efficient production of mutant mouse lines.
MATERIALS AND METHODS
Animal use and generation of KO and KI mice
The animal protocol for this study was approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Nevada, Reno (protocol #00494).
All mice were housed and maintained under specific pathogen-free conditions with a temperature- and humidity-controlled animal facility at the University of Nevada, Reno. Generation of KO and KI mice was performed as previously described with modifications (Gu et al., 2018; Wang et al., 2019). Briefly, 4–6-week-old FVB/NJ or C57BL/6J female mice were super-ovulated and mated with C57BL/6J stud males; zygotes and two-cell stage embryos were collected from the oviducts for KO and KI, respectively. For KO, Mb3Cas12a mRNA (200 ng/μl) and crRNA (100 ng/μl) were mixed and injected into the cytoplasm or pronucleus of the zygotes in M2 medium (cat. #MR-051-F, Millipore, Burlington, MA). For KI, Mb3Cas12a-mSA mRNA (75 ng/μl), crRNA (50 ng/μl) and biotinylated donor DNA template (20 ng/μl) were mixed and injected into the cytoplasm or pronucleus of the two-cell stage embryos in M2 medium. After injection, all embryos were cultured for 1 h in KSOM+AA medium (cat. #MR-121-D, Millipore) at 37°C under 5% CO2 in the air before being transferred into 7–10-week-old female CD1 recipients.
To prepare pcDNA3.1-Mb3Cas12a-mSA plasmid, the monomeric streptavidin (mSA) DNA fragment was amplified from PCS2+Cas9-mSA plasmid (Addgene #103882) using Q5® Hot Start High-Fidelity 2X Master Mix (Cat. #M0494S, NEB). The PCR conditions are as follows: initial denaturation at 98°C for 30 s followed by 35 cycles of amplification (denaturation at 98°C for 10 s, annealing at 65°C for 20 s, and elongation at 72°C for 1 min) with the final elongation at 72°C for 2 min. After digestion with BamHI (Cat. #R0136S, NEB, Ipswich, MA) and EcoRI (Cat. #R3101S, NEB) at 37°C for at least 2 h, the digested pY117 plasmid (pcDNA3.1-huMb3Cpf1) (Addgene #92293) and PCR products were purified using Ampure beads and ligated using T4 DNA ligase (Cat. #M0202L, NEB) at 4°C overnight. An aliquot (2 μl) of the ligation products was introduced into 16.7 μl of Mix & Go competent cells (DH5 alpha strain, Cat. #T3007, Zymo Research) for transformation. The transformed cells were spread onto agar plates containing ampicillin (Cat. #BP1760-5, Thermo Fisher Scientific) (100 µg/ml), followed by culture overnight at 37°C. The colonies were picked and genotyped, and the potential positive colonies were cultured overnight in culture medium containing NZCYM broth (Cat. #BP2464-2, Thermo Fisher Scientific) (20 mg/ml) and ampicillin (100 µg/ml). The plasmids were extracted using ZR Plasmid Miniprep (Cat. #D4054, Zymo Research) followed by Sanger sequencing to verify successful cloning. To prepare the pUC-Slit2-BamHI plasmid, two homology arms (∼1 kb) flanking the crRNAs cutting sites of Slit2 locus and pUC empty vector were amplified by Q5® Hot Start High-Fidelity 2X Master Mix (Cat. #M0494S, NEB) from mouse tail genomic DNA and pX330 plasmid (Addgene #42230), respectively. A BamHI restriction site was introduced between the two homology arms through PCR amplification. After purification with Ampure beads, these three DNA fragments were assembled using the NEBuilder® HiFi DNA Assembly Master Mix (Cat. #E2621L, NEB) at 50°C for 1 h. Transformation and verification of ligation were conducted as described above. The primers used for plasmid construction are listed in Table S1.
Generation of Mb3Cas12a and Mb3Cas12a-mSA mRNA, crRNAs and donor DNA template
To synthesize Mb3Cas12a and Mb3Cas12a-mSA mRNAs, the pY117 plasmid (pcDNA3.1-huMb3Cpf1) (Addgene #92293) and pcDNA3.1-Mb3Cas12a-mSA were digested with EcoRI (Cat. #R3101S, NEB) overnight at 37°C, followed by purification with Ampure beads and mRNA synthesis with the HiScribe™ T7 ARCA mRNA kit (Cat. #E2065S, NEB) with addition of a RNase inhibitor (Cat. #M0314L, NEB) at 37°C for at least 2 h. Then the in vitro transcribed mRNAs were treated with DNase I (NEB, Cat. #M0303S) to remove the plasmid DNA template, followed by poly(A) tailing using E. coli poly(A) polymerase (Cat. #M0276S, NEB) at 37°C for 30 min. The poly(A)-tailed mRNAs were purified using the RNA Clean & Concentrator™-5 (Cat. #R1016, Zymo Research, Irvine, CA) and eluted in a Tris-EDTA solution (Cat.#11-01-02-02, IDT, Coralville, IA).
crRNAs were designed using Benchling (https://benchling.com/). DNA oligonucleotides for making each crRNA were synthesized by the IDT Inc. and are listed in Table S1. To prepare crRNAs for microinjection, the T7 Top strand primer and antisense oligonucleotides specific for each crRNA were mixed in 1× T4 DNA ligase buffer and heated to 95°C for 5 min, and then allowed to cool down to room temperature on the bench. The annealed oligonucleotides were used as the templates for in vitro transcription (IVT) using the HiScribe™ T7 High Yield RNA Synthesis Kit (Cat. #E2040S, NEB) at 37°C overnight. After IVT, crRNAs were purified using the RNA Clean & Concentrator™-5 (Cat. #R1016, Zymo Research) and eluted in Tris-EDTA solution (Cat.#11-01-02-02, IDT). To prepare crRNAs for transfection of HEK-293 cells, PCR products corresponding to each crRNA were amplified with U6 forward primer and corresponding antisense oligonucleotides (as listed in Table S1) from the pX330 plasmid (Addgene #42230). After digestion with DpnI (Cat. #R0176S, NEB), the PCR products were purified using Ampure beads. The biotinylated donor DNA template was amplified from the pUC-Slit2-BamHI plasmid with biotinylated primers (as listed in Table S1), followed by DpnI digestion to remove the plasmid and purification with Ampure beads.
HEK-293 cell transfection
HEK-293 cells (Cat. #ATCC® CRL-1573™, ATCC) were maintained in DMEM (Cat. #11995, Thermo Fisher Scientific) containing 10% FBS (Cat. #S12450, R&D Systems) and 1× antibiotic-antimycotic (Cat. #15240062, Thermo Fisher Scientific). At 1 day before transfection, the culture medium was changed to DMEM containing 10% FBS. At ∼60% confluence, the HEK-293 cells were co-transfected with 400 ng of pY117 (pcDNA3.1-huMb3Cpf1) (Addgene #92293) and 100 ng of crRNA PCR product using 2.5 μl Lipofectamine 2000 (Cat. #11668, Thermo Fisher Scientific, Waltham, MA) in a 24-well cell culture plate (Cat. #3524, Corning, Corning, NY). The culture medium was changed back to DMEM containing 10% FBS and 1× antibiotic-antimycotic at 12 h after transfection. The cells were collected 48 h after transfection for analyses.
Mouse genotyping and Sanger sequencing
Mouse tail or ear snips were lysed in a lysis buffer (40 mM NaOH, 0.2 mM EDTA, pH 12.0) for 1 h at 95°C, followed by neutralization using the same volume of neutralizing buffer (40 mM Tris-HCl, pH 5.0). PCR was conducted using Platinum™ SuperFi™ Green PCR Master Mix (Cat. #12359010, Thermo Fisher Scientific). The 20 μl PCR mix contained 10 μl of 2× Platinum™ SuperFi™ Green PCR Master Mix, 4 μl of 5× SuperFi™ GC Enhancer, 1 μl of 10 μM forward/reverse primers, 5–50 ng of genomic DNA and nuclease-free water. The PCR conditions were as follows: initial denaturation at 98°C for 30 s, followed by 35 cycles of amplification (denaturation at 98°C for 10 s, annealing at 68°C for 10 s, and elongation 72°C for 1 min) with a final elongation at 72°C for 5 min. Positive samples were subjected to single-A tailing using GoTaq® Green Master Mix (Cat. #M7123, Promega, Madison, WI) for 5 min at 95°C, followed by 15 min at 72°C. The single A-tailed PCR products were then ligated to pGEM®-T Easy Vector using T4 DNA ligase from the pGEM®-T Easy Vector Systems (Cat. #A1360, Promega) at 4°C overnight before transformation. Transformation and verification of ligation were conducted as described in the ‘Plasmid construction’ section. Data were analyzed using Geneious software (Biomatters, Inc.). The primers used for genotyping are listed in Table S1.
The T7EI assay (Cat. #M0302L, NEB) was used to detect mutations. PCR products from genotyping were resuspended in 1× buffer 2 and then underwent an initial denaturation at 95°C for 5 min. The denatured PCR products were allowed to anneal by decreasing temperature from 95°C to 85°C with a ramp rate of −2°C/s, followed by a second annealing period from 85–25°C with a ramp rate of −0.1°C/s. The annealed PCR products were then treated with 0.5 μl of T7EI at 37°C for 30 min. The reaction was stopped by adding 0.75 μl of 0.25 M ETDA, as well as 1 μl of Proteinase K (Cat. #P8107S, NEB), followed by incubation at 37°C for 30 min. The T7EI-digested samples (10 μl) were mixed with 2 μl of 6× Purple Gel Loading Dye (Cat. #B7024S, NEB), and loaded into each well of the 10% Novex™ TBE Gels (Cat. #EC62755BOX, Thermo Fisher Scientific) in 1× TBE Buffer (Tris-borate EDTA; Cat. #B52, Thermo Fisher Scientific). The gels were run at 180 V for 30 min, then incubated with 1× TBE buffer containing SYBR Gold (Cat. #S11494, Thermo Fisher Scientific) or ethidium bromide (Cat. #X328, Amresco) for 10 min. The gels were then imaged and analyzed based on the gray density by ImageJ (NIH) and the percentage of indels was quantified as described previously (Cong et al., 2013).
MiSeq library construction and sequencing
Genomic fragments of the Prps1, Prps2 and Prps1l1 loci were amplified from DNA isolated from mouse tail or ear snips using Platinum™ SuperFi™ Green PCR Master Mix (Cat. #12359010, Thermo Fisher Scientific). The PCR products were then tagged using Nextera XT DNA Library Preparation Kit (Cat. #15032354, Illumina, San Diego, CA) and indexed using Nextera XT Index Kit (Cat. #15055294, Illumina) according to the manufacturer's instructions. Briefly, 5 μl of the PCR product (total 1 ng) were mixed with 10 μl of Tagment DNA (TD) Buffer and 5 µl of Amplicon Tagment Mix (ATM), followed by incubation at 55°C for 5 min. The tagging reaction was then stopped by adding 5 µl of Neutralize Tagment (NT) buffer and incubation at room temperature for 5 min. For indexing, the reaction was mixed with 15 µl of Nextera PCR Master Mix (NPM), 5 µl Index i7 adapters and 5 µl Index i5 adapters, then incubated at 72°C for 3 min. After initial denaturation at 95°C for 30 s, 12 cycles of amplification were performed with the following conditions: denaturation at 95°C for 10 s, annealing at 55°C for 30 s, and elongation at 72°C for 30 s, ending with a final elongation at 72°C for 5 min. After purification with Ampure beads, the DNA library was sequenced using MiSeq Reagent Kit v2 (500-cycles) (Cat. #MS-102-2003, Illumina). Data were analyzed using Geneious software. The primers used for Prps1, Prps2 and Prps1l1 are listed in Table S1.
Conceptualization: Z.W., W.Y.; Methodology: Z.W., Y.W., S.W., A.J.G., H.M., T.Y., K.C.-G., B.P., H.W., H.Z.; Software: Z.W.; Validation: Z.W., S.W.; Investigation: Z. W., W.Y.; Resources: Y.W., A.J.G., H.Z., W.Y.; Data curation: Z.W., Y.W.; Writing - original draft: Z.W., W.Y.; Writing - review & editing: Z.W., W.Y.; Visualization: Z.W., S.W.; Supervision: W.Y.; Project administration: W.Y.; Funding acquisition: W.Y.
This work was supported by grants from the National Institutes of Health (HD071736, HD085506 and P30GM110767 to W.Y.) and John Templeton Foundation (PID 61174 to W.Y.). Deposited in PMC for release after 12 months.
The datasets generated and/or analyzed during the current study are available in the Sequence Read Archive (SRA) under accession no. PRJNA556550.
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.240705.reviewer-comments.pdf
The authors declare no competing or financial interests.