Connexin 41.8 mediates the correct temporal induction of haematopoietic stem and progenitor cells

  1. Tim Petzold
  2. Masakatsu Watanabe
  3. Julien Y. Bertrand

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Haematopoietic stem and progenitor cells (HSPCs) derive from a subset of endothelial cells (ECs), known as haemogenic ECs by the process of endothelial-to-haematopoietic transition (EHT). Although many factors involved in EHT have been elucidated, we still have a poor understanding of the temporal regulation of this process. Mitochondrial-derived reactive oxygen species (ROS) have been shown to stabilise hypoxia-inducible factors 1/2α (Hif1/2α), allowing them to positively regulate EHT. Here, we show a developmental delay in EHT and HSPC induction in a gap junction mutant, connexin (cx)41 . 8 (orthologous to mammalian CX40 ), in zebrafish. In mammalian cells, CX40 has been shown to localise to the mitochondria. We demonstrate that Cx41.8 is important for the correct temporal generation of mitochondrial ROS, which stabilise the Hif pathway, allowing for the subsequent specification of the haemogenic endothelium. Taken together, our data indicate that Cx41.8 mediates the correct induction of HSPCs.

Article activity feed

  1. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    1. General Statements

    We would like to thank the 3 reviewers for their comments and suggestions for our manuscript. We believe that the revisions we plan to make, based on the comments by the reviewers, will greatly enhance the quality of our manuscript.

    We would like to respond to some reviewer comments here, since they do not fit into any of the subsequent sections.

    Reviewer #3

    In the Results section that describes the delay in gata2b expression (page 4 and Supp. Fig. 4), the authors show that the mutant embryos start expressing more gata2b at 30 - 36hpf after the decreased expression at earlier time points, with no difference at 48hpf. What could explain that recovery?

    We thank reviewer 3 for this question. The partial functionality of the Cx41.8 channel in *cx41.8tq/tq *mutants may explain why the HSPC program is eventually induced (leading to sufficient mitochondrial ROS production for Hif1/2α stabilisation). However, this could also result from functional redundancy between Cx41.8 and other connexins such as Cx43 or Cx45.6 in the mitochondria, since they are also expressed in zebrafish arterial ECs at 24hpf (Gurung et al, Sci Rep, 2022) and cx43 knockdown has previously been shown to result in an HSPC specification defect in zebrafish (Jiang et al, Fish Physiol Biochem, 2010). Together, these aspects may explain the recovery, although delayed, of gata2b expression in the *cx41.8tq/tq *mutant, as discussed in detail in our manuscript.

    The authors showed that gata2b expression can be rescued by ROS induction in the dose-dependent manner (page 6 and Fig.3 and Supp. Fig. 6). Is this what rescues gata2b expression at 30hpf in the cx41.8 mutants?

    This is exactly right, we hypothesize that in cx41.8tq/tq mutants, it takes longer for mitochondrial ROS production to reach above the threshold required to stabilise Hif1/2α and hence induce gata2b expression, which is supported by the data referred to by this reviewer.

    Are any vascular defects in the mutant embryos?

    Our lab previously reported that cx41.8tq/tq embryos have faster ISV growth rate (Denis et al, Front Physiol, 2019). However, we found no evidence of a link between the ISV growth rate increase and the HSPC specification defect in these embryos. Importantly, we show that aorta specification is normal in *cx41.8tq/tq *mutants, as determined by dll4 expression at 24 (Supp. Fig. 1C) and 28 hpf (Supp. Fig. 1D).

    2. Description of the planned revisions

    Reviewer #1 (Evidence, reproducibility and clarity (Required)):

    Summary

    The manuscript by Petzold et al. explores the functions of connexin 41.8 (cx41.8) (mammalian homologue Connexin 40) in hematopoietic stem cell (HSC) formation in the zebrafish dorsal aorta. The authors use a cx41.8 allele that appears to be hypomorphic, as the phenotype is milder than a previous cx41.8 allele that the same group published (Cacialli et al., 2021). cx41.8tq/tq mutants exhibit delayed onset of hemogenic endothelial specification, as marked by gata2b at 24 hpf, but HSPC development proceeds normally from 48 hpf onwards. A new reporter line for cx41.8, Tg(cx41.8:GFP), was generated and is expressed in the floor of the dorsal aorta, consistent with the location of hemogenic endothelial cells. Lower ROS production in the whole cell and in the mitochondria was reported in the cx41.8tq/tq mutants, and treatment with ROS enhancers, H2O2 and menadione, appeared to rescue the mutant phenotype of reduced HSPCs at 28 hpf. Finally, the authors tested a link between cx41.8 and Hif1α by pharmaceutically (DMOG/CoCl2) or genetically (vhl morpholino) inhibiting Hif inhibitors, and observed a rescue of HSPC formation in cx41.8 mutants.

    I think it would be important for the authors to address the mechanisms of why cx41.8tq/tq and the other cx41.8-/- (leot1/t1) mutant phenotypes are different, with the latter allele showing more severe phenotypes of increased HSPC apoptosis and reduced HSPCs during later development. The authors speculate the cx41.8tq/tq allele encodes a missense mutation in one of the channel domains, and as such, might be a hypomorph. The authors cited the original paper by Watanabe et al. (2006); however, this paper actually noted that the cx41.8tq/tq allele is likely to be a dominant negative - and as such, should have exhibited a stronger phenotype than the leot1/t1 mutant allele. From the paper: "leotw28 and leotq270 heterozygotes have phenotypes different from that of WT; thus, they represent dominant-negative alleles." Importantly, no data are shown to provide evidence that the allele is a hypomorph - at minimum, qPCR data should be provided to show whether there is NMD of the mRNA in cx41.8tq/tq mutants.

    We would like to thank the reviewer for this comment and suggestion. As the reviewer has rightly pointed out, the cx41.8tq/tq mutation is thought to result in a protein with dominant-negative function (Watanabe et al, EMBO Rep, 2006; Watanabe et al, J Biol Chem, 2016).

    In fact, we agree that the mutant cx41.8tq/tq protein acts as a dominant-negative and although the reviewer is right to point out that the cx41.8t1/t1 mutant may thus exhibit a stronger phenotype which we found not to be the case (runx1 expression was found to be normal in the cx41.8t1/t1 mutant, Cacialli et al, Nature Commun, 2021), we provided our explanation for this in the discussion of the manuscript:

    “The partial functionality of the Cx41.8 channel in *cx41.8tq/tq *mutants [14] may explain why the HSPC program is eventually induced. However, this could also result from functional redundancy between Cx41.8 and other connexins such as Cx43 or Cx45.6 in the mitochondria, since they are also expressed in zebrafish arterial ECs at 24hpf [18] and *cx43 *knockdown has previously been shown to result in an HSPC specification defect in zebrafish [36]. This potential functional redundancy may also provide an explanation as to why HSPCs are specified normally, without any delay, in *cx41.8t1/t1 *embryos [12]. In these null mutants, *cx41.8 *expression is completely absent but may be functionally compensated by other connexins, whereas in *cx41.8tq/tq *mutants, although *cx41.8 *is expressed, its channel function is reduced [14]. Moreover, as Cx41.8 may form heterotypic channels with Cx43 and/or Cx45.6 (and potentially also with others), the function of these chimeric channels would also be altered”

    We believe this addresses the reviewers concern regarding this, especially given the fact that Cx43 and Cx45.6 have been found to be expressed in arterial ECs at 24 hpf, as cited in the manuscript. With regards to the reviewer’s question about whether there is NMD of the cx41.8 transcript, given that the cx41.8tq/tq mutation is missense and does not result in a premature stop codon (usually required for NMD to be induced, Kurosaki et al, J Cell Sci, 2016), we do not believe that there is NMD of the cx41.8 transcript in *cx41.8tq/tq *mutants. We will however verify this by carrying out the experiment suggested by this reviewer, qPCR analysis of *cx41.8 *expression in *cx41.8tq/tq *embryos and wild-type controls.

    The quantification data in this manuscript are not satisfactory. The authors only provide graphs that show embryos with "low", "medium" and "high" numbers of HSPCs, which is incredibly subjective. Considering that the authors already have the cx41.8tq/tq in the Tg(myb:GFP) background (Figure 1E), they could have quantified the precise numbers of Tg(myb:GFP)-positive cells at different timepoints and with the different pharmaceutical rescue experiments. Ideally, this should be combined with other HSPC markers such as Tg(cd41:GFP) or Tg(runx1:GFP) - although this could be limited by the authors' access to the lines or time it takes to cross the mutants to the transgenes.

    We thank reviewer 1 for their concern regarding this. Indeed the reviewer is correct, it would take us too long (at least 6 months) to generate the cx41.8tq/tq cd41:GFP or cx41.8tq/tq runx1:GFP lines, however, as stated, we do already have the cx41.8tq/tq cmyb:GFP zebrafish line. That said, repeating the pharmacological experiments using the cx41.8tq/tq cmyb:GFP zebrafish line would demand months of work and we do not currently have the personnel to perform all of this. However, we will perform the same experiment as performed previously to generate figure 1E but also at earlier timepoints. The cmyb:EGFP transgene marks nascent HSPCs from 28 hpf, and so we will aim to image, and quantify differences in budding HSPCs in cx41.8tq/tq cmyb:EGFP and cmyb:EGFP controls between 28 hpf and 36 hpf. We agree with the reviewer that this will add depth to our study and will provide evidence to back up our conclusions.

    The link between cx41.8 and Hif1α is tenuous. The authors should perform in situ hybridization for the hif1 genes and their downstream effector notch1 which is known to be important for the HSPC specification (Gerri et al., 2018).

    We thank the reviewer for this point. we do not expect hif1/2α expression to be affected in this mutant. Mitochondrial ROS has been shown to stabilise Hif1/2α at the protein level, not the mRNA level. Our data, and that of others (Harris et al, Blood, 2013), suggest that in the absence of mitochondrial ROS, prolyl hydroxylases are not inhibited by mitochondrial ROS, and they target Hif1/2α for ubiquitination and subsequent destruction in a Vhl-dependent manner (as shown in Fig. 4D). We have changed the text in the manuscript to clarify that Hif is stabilised on the protein level (please see the section below).

    Since we do however expect *notch1a *and *notch1b *expression to be altered in our mutant embryos, as they are transcriptionally regulated by Hif1/2α (Gerri, Blood, 2018), we will perform in situ hybridisation and qPCR analysis of these 2 genes at 18-24 hpf in *cx41.8tq/tq *mutants and controls to clarify this point and solidify our model.

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

    Summary

    Petzold et al are here addressing the potential function of the connexin Cx48.1, a protein involved in the structure of gap junctions, in the specification of future hematopoietic stem cells and progenitors (HSPCs). This piece of work is complementing their previous results showing the function of this connexin isoform in HSPC expansion in the transient hematopoietic niche in the caudal tissue of the zebrafish embryo. They explore phenotypes triggered by the expression of a mutant form bearing a single amino-acid substitution in the fourth transmembrane domain of the protein. Using whole mount in situ hybridization (WISH) of the two transcription factors Gata2b and Runx1, a novel transgenic fish line that expresses eGFP under the control of the Cx48.1 promoter region, and a series of drug treatments interfering with, or promoting, the formation of reactive oxygen species (ROS) production and oxidative stress, they propose that Cx48.1 is also involved upstream of HSPC amplification, rather in their specification at the level of the hemogenic endothelium constituting the ventral floor of the dorsal aorta. Mechanistically, they hypothesize that this function relies on mitochondria-derived ROS that would destabilize the VHL protein involved in mediating the degradation of Hif1/2a transcription factors, thereby stabilizing the Hif1/2a-Notch1a/b signaling axis involved in specification of the hemogenic endothelium.

    The WISH and quantitative analyses.

    Most of the quantitative analyses in the work are based on chromogenic WISH, which is not sufficiently accurate because leading to highly variable results, in addition to its lack of linearity. WISH is also subjected to important variations, particularly for transcription factors that are expressed at low levels such as Runx1, and to some extent Gata2b also. One obvious example in the paper is the inconsistency of signals that are observed Fig1C (Gata2b, left, wt, 24hpf) and FigS3B (Gata2b, left, wt, 24hpf) in which the signal is barely visible and is comparable to the signal for the cx41.8tq/tq mutant Fig1C, right.

    In addition, in the timings that are analyzed in FigS3 (Gata2b, 26 and 28hpf) to argue on temporal delay of expression in the cx41.8tq/tq mutant, the Gata2b signal is masked by the strong increase in tissues other than the hemogenic endothelium in the dorsal aorta (including signal in the somites as well as, possibly, increase in background). In this very example, it is legitimate to question the accuracy of the quantification methodology when the signal in the tissue of interest is drowned in the overall signal from surrounding tissues; how can the authors explain the 100% of embryos that have a 'Low' signal in the region of interest (FigS3C, cx41.8tq/tq mutant in comparison to WT)? This point is also valid for the data quantified FigS4 in which the fitting between WISH data and the quantifications appears to be questionable (for all timing points: 30, 32, 36, 48hpf and comparing mutant with the WT.

    My suggestion would be to complement the WISH data and improve the quantitative analyses using another, more accurate approach such as qRT-PCR for example (on dissected trunk regions and, if necessary because of expression in other surrounding tissues (in the case of Gata2b at later time points), after FACS-sorting using a fish line expressing a fluorescent reporter driven by a vascular promoter, ex: the kdrl:mCherry line used in the work). This is particularly important for the expression of the two transcription factors Runx1 and the more upstream Gata2b, the latter being involved in HSPC specification which is taken as a reference. qRT-PCR experiments should be feasible relatively easily and in a reasonable time frame as the technics is not very time consuming and easily accessible.

    We thank reviewer 2 for their concerns regarding the in situ quantifications used during this study. Although the approach we have used is widely used in the field to quantify gene expression differences, we appreciate that our data could be strengthened by complementing it with another approach. As such we will do the following:

    • We will complement our in situ hybridisation characterisation of delayed hemogenic endothelium formation and HSPC specification with qPCR experiments. For this, we will dissect the trunks of *8tq/tq *embryos and controls and perform qPCR analysis of gata2b expression at the timepoints analysed during development (Supp. Fig. 3 A-D and Supp. Fig. 4 A-D), whilst also using the same approach to compliment the data for gata2b and runx1 expression at 24 hpf (Figure 1C and D). We agree with the reviewer that this is a feasible approach and would add robustness to the data we already show.

    2- Fluorescence imaging and associated interpretation/conclusions.

    The fluorescence images (Fig1E; Fig2B,D; Fig3A) are very difficult to analyze; they lack resolution because they appear to be epifluorescence images and not confocal images. When the signal is low, which is in particular the case for the novel Cx41.8:EGFP fish line, Fig2B (which is confirmed with the FACS GFP signal in comparison to the mCherry of the kdrl:mCherry fish line), it is not possible to provide convincing images on the vascular/aortic expression because of the high background of diffusion (the authors state 'likely to be the aortic floor', indeed it is not possible to validate the fact that the expression is truly in potential hemogenic cells). The double positive population in the FACS (Fig2C, right) does not resolve the issue because if indeed cx41.8 is expressed in endothelial cells (as expected from previous studies), the double positive population could equally be endothelial cells from inter-somitic vessels, for example (not to mention the underlying vein which is very close to the aorta in the trunk)). Fig2D, images are too small and, again, the resolution is not good enough to say that double positive cells are on the aortic floor. It is recommended to convince the reader that the authors try to confirm their statements by using confocal microscopy and increase the magnification of the relevant regions of interest.

    We thank this reviewer for this point. We will address this concern by using, as they suggest, confocal microscopy to try to get higher resolution images. In particular, we will do the following:

    • We will use confocal microscopy to image the 8:EGFP line as was done previously (Fig 2B), in order to obtain higher resolution images of expression of cx41.8 in the floor of the aorta.
    • We will also use confocal microscopy to image the 8:EGFP;kdrl:mCherry line as was done in Fig 2D, in order to gain higher resolution images.
    • We will also increase the magnification of the relevant regions of our confocal microscopy images as suggested by this reviewer.

    There is an inconsistency in the data between Fig1E (40hpf, in vivo imaging using the cmyb:GFP fish line) and FigS2 (48hpf, WISH cmyb); how can we observe 'HSPCs budding from the dorsal aorta' (see legend Fig1, arrowheads) which seems very much decreased in the imaging experiment for the cx41.8tq/tq mutant in comparison to WT, and have no effect on the cmyb signals FigS2B? What are the GFP+ cells that are aligned along the elongated yolk Fig1E and that appeared to be decreased in number in the mutant?

    We agree that this disparity is confusing for the reader. We believe the disparity between these results is due firstly to the fact that the experiment in Supp. Fig 2C was performed 8 hours after that in Fig 1E and secondly due to the time it takes for GFP to fold (in the case of Fig 1E). It is also important to keep in mind that the phenotype is not a complete absence of HSPC budding, but only a delay in the onset of EHT.

    • We will however address this concern by carrying out the experiment described above - we will perform the same experiment as performed previously to generate figure 1E but also at earlier timepoints. The cmyb:EGFP transgene marks nascent HSPCs from 28 hpf, and so we will aim to image, and quantify differences in budding HSPCs in 8tq/tq cmyb:EGFP and cmyb:EGFP controls at numerous timepoints from 28 hpf to 36 hpf. This will add depth to our study by providing evidence to back up our conclusions.
    • We will remove the 40-hpf timepoint (Fig 1E) to avoid confusion regarding the disparity with cmyb expression by WISH in Supp. Fig 2C.
    • Regarding the GFP+ cells aligned along the yolk in 1E, we thank the reviewer for pointing this out. These cells are multiciliated cells, from the kidney tubules (Wang et al, Development 2013). We will determine whether their numbers do indeed differ between 8tq/tq;cmyb:EGFP and cmyb:EGFP controls in our new confocal experiments and will mention this in the manuscript if they do.

    It would be important to investigate/show, at least with qualitative WISH experiments all along the time-window of HSPC specification as stated by the authors (26-54hpf, see main text third paragraph of Results), that Cx41.8 is detected in arterial endothelial cells (and perhaps enriched in the hemogenic endothelium?), in complement to the work they are referring to on transcriptomic data at 24hpf (Ref18 Gurung et al Sci Rep 2022). Ideally, these WISH data should be resolutive enough to provide clear localization in aortic cells versus cells in the aortic floor to bring significant added value to the work that lacks spatial resolution (ex: fluorescent WISH using confocal microscopy, allowing to superpose signal with cell types (either by double fluorescent WISH (vascular marker + Cx41.8) or superposing fluorescence signals with transmitted light)).

    We agree with this reviewer regarding this point. The way we will address this is to use confocal microscopy at different timepoints from 24-40 hpf using the *cx41.8:EGFP; kdrl:mCherry *line to show that expression of cx41.8 is indeed present and enriched in the floor of the dorsal aorta during the timeframe of HSPC specification. We believe that imaging this line using confocal microscopy will be sufficient to clearly show this.

    It would be more informative and secure, Fig2D, to show images of the double transgenics (Cx48.1:eGFP;kdrl:mCherry) at 28-30 hpf (rather than 48 hpf) which is more narrowed down to the specification of the hemogenic endothelium thus preventing any risk to visualize the fluorescence signals coming from recently born HSPCs rather than signals from cells embedded in the aortic floor.

    We thank the reviewer for this suggestion, which we believe would indeed improve the manuscript. As discussed above, we will indeed use confocal microscopy at different timepoints, including 28-30 hpf using the cx41.8:EGFP;kdrl:mCherry line to show that expression of cx41.8 is indeed present and enriched in the floor of the dorsal aorta during the timeframe of HSPC specification. We believe that imaging this line using confocal microscopy will be sufficient to clearly show this and so thank the reviewer for this excellent suggestion.

    To make the data more convincing on the ROS production in the ventral side of the cord in wild type embryos (which suggests that future hemogenic cells are already ventralized at that stage), it would be important to obtain confocal images of the region of interest and perform reconstitution of Z-stacks with a sagittal view (rather than longitudinal). It would be nice also to obtain comparable images later on, after lumenization and before initiation of HSPC emergence (before 28hpf).

    We thank the reviewer for this suggestion and agree that the suggested approach will solidify our data. As such, we will carry out the proposed experiment, using confocal imaging to gain longitudinal and sagittal images of mitoSOX staining in WT embryos and cx41.8tq/tq mutants at both 16 hpf and 26 hpf.

    3. Description of the revisions that have already been incorporated in the transferred manuscript

    Reviewer #1

    Related to the above point, the authors should test whether the gap junction function of Cx41.8 is intact in the cx41.8tq/tq mutants by assessing calcium waves in the GCamp transgenic line.

    …we have also now found additional published in vivo evidence that Cx41.8 channel function is reduced in the cx41.8tq/tq mutant, which is now also cited in the new version of the manuscript (please see our full response to this point below).

    Please see the section “Description of analyses that authors prefer not to carry out” for additional information regarding the GCamp experiment suggestion.

    The link between cx41.8 and Hif1α is tenuous. The authors should perform in situ hybridization for the hif1 genes…

    We thank reviewer 1 for making this point. To clarify this, we do not expect hif1/2α expression to be affected in this mutant. Mitochondrial ROS has been shown to stabilise Hif1/2α at the protein level, not the mRNA level. Our data, and that of others (Harris et al, Blood, 2013), suggest that in the absence of mitochondrial ROS, prolyl hydroxylases are not inhibited by mitochondrial ROS, and they target Hif1/2α for ubiquitination and subsequent destruction in a Vhl dependent manner (as shown in Fig. 4D).

    To clarify this in the manuscript, we have adjusted the text in three places (including in the abstract) to clarify that Hif1/2α is stabilised at the protein level, as shown below. We believe these changes have made this important point more understandable for the reader:

    1. “… Mitochondrial-derived reactive oxygen species (ROS) have been shown to stabilise the hypoxia-inducible factor 1/2a (Hif1/2a) proteins, allowing them..”
    2. “Recent research has demonstrated that hypoxia and mitochondrial ROS are required for the stabilisation of the transcription factors Hif1/2a at the protein level”
    3. “… as mitochondrial ROS generation may eventually reach the threshold required to sufficiently stabilise the Hif1/2a proteins for downstream”

    Reviewer #2

    Importantly, it appears also that all over the WISH quantifications, the reader cannot appreciate the accuracy of the categories High/Medium/Low, which is not at all developed in the Methods section (paragraph Image processing and WISH phenotypic analyses).

    We have developed the Methods section (paragraph Image processing and WISH phenotypic analyses), which was highlighted as a concern by this reviewer, in order to detail exactly how we performed our image analysis and statistical analyses using this approach. We believe this will satisfy the concerns reviewer 2 has regarding this and appreciate that they have a point that this was indeed underdeveloped in the original submission.

    Finally, there is a confusion in the quantification regarding the number of HSPCs (see the beginning of the second paragraph of Results 'The HSPC specification defect in cx41.8tq/tq mutants is due to a delay in Gata2b expression') and the % of embryos falling into the 3 categories High/Medium/Low FigS2, cmyb 48hpf. The authors use this argument (based on the WISH cmyb signals) to infer that the deficit in the cx41.8tq/tq mutant is not due to controlling HSPC number (no difference in cmyb between WT and mutant) but rather upstream, at the level of the hemogenic endothelium, which is not a thorough argument at that point.

    We thank reviewer 2 for pointing this out to us and agree that the wording we used is a little confusing. We have therefore added to the first sentence of the second paragraph in the results section “'The HSPC specification defect in cx41.8tq/tq mutants is due to a delay in gata2b expression” which now reads:

    “Hence, since HSPC specification is initially reduced, but then recovers in *cx41.8tq/tq *embryos, we suspected a delay in the formation of the haemogenic endothelium in these mutants. To test this hypothesis…”

    We believe this change to the manuscript will satisfy the reviewers concern by making this section more logical for the reader.

    The authors should take care of the fact that at 16hpf, it is an overstatement to speak of an aorta when the cord is starting to lumenize at around 18hpf, Jin et al Development 2005 (see Main text referring to Fig3).

    We thank the reviewer for this clarification. We have changed the relevant text to state “vascular cord” instead of “aorta” and have mentioned that it begins to lumenize around 18hpf for clarification. We have also added the suggested reference.

    Reviewer #3

    As Gata2 has been shown to be a positive autoregulator of itself in mice (Nozawa 2009, Katsumura 2016) and might do so in zebrafish (Dobrzycki 2020), so could gata2b recover itself, in a dose-dependent manner, without the Hif-Nocth1 axis once enough of it is expressed?

    We thank reviewer 3 for this suggestion. We believe that our data show that Cx41.8 is required for mitochondrial ROS production, which stabilises Hif1/2α and switches on downstream gata2b via Notch1a/b (which will be added, see previous section). As such, we believe that the Hif1/2α/Notch1a/b axis is required, at least for the initial induction of* gata2b* expression. However, reviewer 3 makes a very interesting point regarding the potential for* gata2b* to positively autoregulate itself, which may of course occur once gata2b expression has been induced by the Cx41.8-mitoROS-Hif1/2α-Notch1a/b-gata2b pathway. We thank the reviewer again for this interesting proposition and have added this suggestion into our discussion in the following paragraph:

    “GATA2 has been shown to positively autoregulate its own expression in mice (Nozawa et al, Genes to Cells, 2009; Katsumura et al, Cell Reports 2016), and Gata2b may also act in this way in zebrafish (Dobrzycki et al, Commun Biol, 2020). Therefore, it is interesting to speculate that once gata2b expression has been induced by the Cx-mitoROS-Hif1/2α-Notch1a/b-gata2b pathway, it may also further induce its own expression, which would make the induction of the haematopoietic transcriptional program more robust”

    Is Hif1/2a expression affected in the mutant? Is it expressed normally but then degraded faster due to the absence of mitochondrial ROS or is it less Hif1/2a expressed overall?

    We thank reviewer 3 for this question, which is similar to a point made by reviewer 1. To clarify, we do not expect hif1/2α expression to be affected in this mutant. Mitochondrial ROS has been shown to stabilise Hif1/2α at the protein level, not the mRNA level. Our data, and that of others (Harris et al, Blood, 2013), suggest that in the absence of mitochondrial ROS, prolyl hydroxylases are not inhibited by mitochondrial ROS, and they target Hif1/2α for ubiquitination and subsequent destruction in a Vhl dependent manner (as shown in Fig. 4D).

    To clarify this in the manuscript, we have adjusted the text in three places (including in the abstract) to clarify that Hif1/2α is stabilised at the protein level, as shown below. We believe these changes have made this important point more understandable for the reader:

    1. “… Mitochondrial-derived reactive oxygen species (ROS) have been shown to stabilise the hypoxia-inducible factor 1/2a (Hif1/2a) proteins, allowing them..”
    2. “Recent research has demonstrated that hypoxia and mitochondrial ROS are required for the stabilisation of the transcription factors Hif1/2a at the protein level”
    3. “… as mitochondrial ROS generation may eventually reach the threshold required to sufficiently stabilise the Hif1/2a proteins for downstream”

    Does MO-mediated knockdown of vhl in the wildtype and mutant (page 7and Fig. ) result in more HSPCs, following the increase in gata2b expression from WT baseline? Does that high expression persist, or does it drop?

    This is an important question. We had already clarified this in the case of cx41.8tq/tq, since we showed that the vhl MO results in more HSPCs (as determined by runx1 expression) at 28 hpf (Supp. Fig. 8A) but we have now added data for the same marker at the same timepoint for WT embryos (Supp. Fig. 8B).

    Although the vhl MO results in an increase in runx1 signal in WT embryos, since the majority of WT embryos injected with the control MO already have “high” runx1 WISH signal at 28 hpf, the difference between injected and control MO injected WT embryos is not significant (Supp. Fig. 8B), as can be expected. This is now explained in the manuscript following the relevant data addition.

    4. Description of analyses that authors prefer not to carry out

    Reviewer #1

    One major missing component is experimental data that distinguish the gap junction/plasma membrane- related and the mitochondrial membrane-related functions of Cx41.8. This is critical, as the role of Connexins in the mitochondria remains poorly understood (and Connexin 43 is the best understood one). Thus, it is a big claim by the authors that Cx41.8 primarily acts through the mitochondria and not the gap junctions. Suggested experiment: The authors should generate a fluorophore-tagged Cx41.8 - under a ubiquitous (ubb or actin) or HSPC-/hemogenic endothelium-specific (gata2b) promoter to monitor the protein localization of Cx41.8. Providing data on whether Cx41.8 protein indeed localizes to the mitochondria is important to support their claim.

    We thank the reviewer for this suggestion, which we agree would be a nice experimental approach to try to investigate whether Cx41.8 does indeed localise to the mitochondria in zebrafish endothelial cells.

    However, EGFP fused full-length cx41.8 has previously been generated and was reported to be nonfunctional, and it was suggested that the amount of localised Cx41.8 is also too small to detect using this approach (Watanabe et al, Pigment Cell Melanoma Res, 2012; Usui et al, BBA Advances, 2021). An EGFP tagged CT-truncated Cx41.8 construct has also been generated and shown to rescue the cx41.8t1/t1 mutant (Usui et al, BBA Advances, 2021), but EGFP expression again could not be detected using this construct in zebrafish.

    As such, since efforts to carry out such an approach have failed in previous attempts and since it has already been demonstrated that CX40 (orthologous to cx41.8) localises to the mitochondria of endothelial cells (Guo et al, Am J Physiol Cell Physiol, 2017), we believe that confirmation of Cx41.8 localisation to the mitochondria in vivo in zebrafish endothelial cells will be very difficult and too time-consuming in the context of this manuscript.

    Related to the above point, the authors should test whether the gap junction function of Cx41.8 is intact in the cx41.8tq/tq mutants by assessing calcium waves in the GCamp transgenic line.

    We agree with the reviewer that this would be a very elegant approach in order to analyse whether Cx41.8 channel function is affected in cx41.8tq/tq mutants. However, we feel that this experiment is definitely beyond the scope of this manuscript. Furthermore, carrying out this experiment would require the acquisition of the GCamp line as well as multiple crosses with the cx41.8tq/tq line which, together, we envisage would take at least 9 months before the experiments can be performed, as so this experiment would also be too time consuming for this manuscript. Finally, we believe there is already strong published evidence that the cx41.8tq/tq mutant results in disrupted channel function (Watanabe et al, EMBO Rep, 2006), as already cited in our manuscript. However, since then, we have also now found additional published in vivo evidence that cx41.8tq/tq channel function is reduced, which is now also cited in the new version of the manuscript.

    The authors might also want to consider performing transcriptomic analysis (bulk RNA sequencing) from purified HSCs in wild types and cx41.8 mutants and assess the downstream pathways affected by the loss of this gene.

    Although this is an interesting proposition, we consider this suggestion to be out of the scope of this manuscript, especially since our model involves changes in gene expression upstream of HSPC induction, and, expression of the key genes thought to be affected (notch1a/b and gata2b) can be checked using a much more cost and time efficient approach, by qPCR, which we will do, as discussed above.

    Are the authors sure of their statement on budding HSPCs when the GFP signal pointed by arrows could in majority be hemogenic cells? (which would be in favor of their hypothesis on Cx41.8 being involved rather in hemogenic endothelium/HSPC specification).

    Since cmyb is a marker of HSPCs and not of the haemogenic endothelium as demonstrated in numerous publications (North et al, Nature, 2007; Bertrand et al, Development, 2008; Bertrand et al, Nature, 2010 and others). Hence, we are confident that this transgene is marking nascent HSPCs and not the haemogenic endothelium.

    As mentioned by the authors in the Discussion, the other connexin Cx43 (Ref 36, Jiang et al 2010) is playing a significant role in HSPC specification in the zebrafish and is expressed in zebrafish arterial cells at 24 hpf. Hence there may be some functional redundancy between Cx43 and Cx48.1, as supported by previous work from the authors showing that a null mutant of Cx48.1 does not exhibit any phenotype in HSPC specification (Ref12, Cacialli et al 2021). This may be problematic for the experiments using drug treatments in the present work, because they are not selective for the different connexins (ex: anti-oxydants (NAC), connexin blockers (heptanol, CBX)), thus blurring interpretations on the specific function of Cx48.1 versus the ones exerted by Cx43 (this should be also valid for the vhl MO treatments).

    This comment is strengthened by the fact that the authors do not systematically address, for both WT and mutant embryos (Fig3 E, F; FigS6; FigS8), if expression levels with drugs/H2O2/MO are different for the 2 conditions (if relatively equal, it would indeed indicate that these drugs/conditions possibly act on another connexin, which would help the authors in their analyses and interpretations).

    We thank the reviewer for these comments and we agree with their concerns regarding the possibility of other Connexins being affected by our experiments using drug treatments. However, we do not rule this out in our manuscript and actually discuss it as being a very realistic prospect, as written about in the discussion section.

    Sadly, to the best of our knowledge, no selective Cx41.8 inhibitors have been described for use in zebrafish, otherwise we would of course have used this. Hence, this was the reason for our choice of compounds, many of which we also used in our previous publication (Cacialli et al, Nature Commun, 2021).

    The haemogenic endothelium/HSPC phenotype in cx41.8tq/tq embryos confirms that this connexin plays a role in HSPC specification, whilst we believe disentangling which other connexins are also involved in this process will be interesting to look into in other future studies but is beyond the scope of this one – we believe that together, the data presented in our manuscript, along with the revisions we plan to carry out, will be convincing to demonstrate a role for Cx41.8 in the mechanism we describe.

    The authors may try to rescue the wt phenotype by expressing, in the Cx48.1tq/tq mutants, the mRNA encoding for the wt protein.

    Although we appreciate this suggestion, we do not believe this experiment will add much in terms of value to the conclusions of our manuscript and, as such, we believe this suggestion is surplus to requirements for this manuscript.

  2. Review Commons

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

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    The authors have successfully shown how disruption in connexin (cx)41.8 results in delayed gata2b expression due to Hif1/2a instability in the absence of mitochondrial ROS. The data is presented well, and the paper is written clearly. The paper is well structured, and the data supports the authors' argument. This study provides a valuable contribution to the field.

    Could the authors clarify the following questions:

    1. In the Results section that describes the delay in gata2b expression (page 4 and Supp. Fig. 4), the authors show that the mutant embryos start expressing more gata2b at 30 - 36hpf after the decreased expression at earlier time points, with no difference at 48hpf. What could explain that recovery? The authors showed that gata2b expression can be rescued by ROS induction in the dose-dependent manner (page 6 and Fig.3 and Supp. Fig. 6). Is this what rescues gata2b expression at 30hpf in the cx41.8 mutants? As Gata2 has been shown to be a positive autoregulator of itself in mice (Nozawa 2009, Katsumura 2016) and might do so in zebrafish (Dobrzycki 2020), so could gata2b recover itself, in a dose-dependent manner, without the Hif-Nothc1 axis once enough of it is expressed?
    2. Does MO-mediated knockdown of vhl in the wildtype and mutant (page 7and Fig. 4) result in more HSPCs, following the increase in gata2b expression from WT baseline? Does that high expression persist, or does it drop?
    3. Is Hif1/2a expression affected in the mutant? Is it expressed normally but then degraded faster due to the absence of mitochondrial ROS or is it less Hif1/2a expressed overall? Are any vascular defects in the mutant embryos?

    Significance

    his study provides a valuable contribution to the field of developmental hematopoiesis.

  3. Review Commons

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

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Summary

    Petzold et al are here addressing the potential function of the connexin Cx48.1, a protein involved in the structure of gap junctions, in the specification of future hematopoietic stem cells and progenitors (HSPCs). This piece of work is complementing their previous results showing the function of this connexin isoform in HSPC expansion in the transient hematopoietic niche in the caudal tissue of the zebrafish embryo. They explore phenotypes triggered by the expression of a mutant form bearing a single amino-acid substitution in the fourth transmembrane domain of the protein. Using whole mount in situ hybridization (WISH) of the two transcription factors Gata2b and Runx1, a novel transgenic fish line that expresses eGFP under the control of the Cx48.1 promoter region, and a series of drug treatments interfering with, or promoting, the formation of reactive oxygen species (ROS) production and oxidative stress, they propose that Cx48.1 is also involved upstream of HSPC amplification, rather in their specification at the level of the hemogenic endothelium constituting the ventral floor of the dorsal aorta. Mechanistically, they hypothesize that this function relies on mitochondria-derived ROS that would destabilize the VHL protein involved in mediating the degradation of Hif1/2a transcription factors, thereby stabilizing the Hif1/2a-Notch1a/b signaling axis involved in specification of the hemogenic endothelium.

    Major comments

    My major comments on the work are on the accuracy of the data in regard to the two main experimental approaches used by the authors and their subsequent analysis/quantification.

    1. the WISH and quantitative analyses. Most of the quantitative analyses in the work are based on chromogenic WISH, which is not sufficiently accurate because leading to highly variable results, in addition to its lack of linearity. WISH is also subjected to important variations, particularly for transcription factors that are expressed at low levels such as Runx1, and to some extent Gata2b also. One obvious example in the paper is the inconsistency of signals that are observed Fig1C (Gata2b, left, wt, 24hpf) and FigS3B (Gata2b, left, wt, 24hpf) in which the signal is barely visible and is comparable to the signal for the cx41.8tq/tq mutant Fig1C, right. In addition, in the timings that are analyzed in FigS3 (Gata2b, 26 and 28hpf) to argue on temporal delay of expression in the cx41.8tq/tq mutant, the Gata2b signal is masked by the strong increase in tissues other than the hemogenic endothelium in the dorsal aorta (including signal in the somites as well as, possibly, increase in background). In this very example, it is legitimate to question the accuracy of the quantification methodology when the signal in the tissue of interest is drowned in the overall signal from surrounding tissues; how can the authors explain the 100% of embryos that have a 'Low' signal in the region of interest (FigS3C, cx41.8tq/tq mutant in comparison to WT)? This point is also valid for the data quantified FigS4 in which the fitting between WISH data and the quantifications appears to be questionable (for all timing points: 30, 32, 36, 48hpf and comparing mutant with WT). Importantly, it appears also that all over the WISH quantifications, the reader cannot appreciate the accuracy of the categories High/Medium/Low, which is not at all developed in the Methods section (paragraph Image processing and WISH phenotypic analyses); hence, it is not possible to evaluate the accuracy/validity of statistics in particular in the experiments in which the quantification into these categories is used for CoCl2 and morpholino analyses to address the contribution of the Hif1/2a-Notch1a/b pathway Fig4 (these experiments generating results that are not as 'black and white' than the other ones in the paper, hence requiring more accuracy; for example, are the differences in the quantification (% of embryos) significant between the WT+vhl MO and Cx41.8tq/tq mutant + vhl MO? Comparing the 2 WISH results for those conditions does not appear to be very convincing).

    My suggestion would be to complement the WISH data and improve the quantitative analyses using another, more accurate approach such as qRT-PCR for example (on dissected trunk regions and, if necessary because of expression in other surrounding tissues (in the case of Gata2b at later time points), after FACS-sorting using a fish line expressing a fluorescent reporter driven by a vascular promoter, ex: the kdrl:mCherry line used in the work). This is particularly important for the expression of the two transcription factors Runx1 and the more upstream Gata2b, the latter being involved in HSPC specification which is taken as a reference. qRT-PCR experiments should be feasible relatively easily and in a reasonable time frame as the technics is not very time consuming and easily accessible.

    Finally, there is a confusion in the quantification regarding the number of HSPCs (see the beginning of the second paragraph of Results 'The HSPC specification defect in cx41.8tq/tq mutants is due to a delay in Gata2b expression') and the % of embryos falling into the 3 categories High/Medium/Low FigS2, cmyb 48hpf. The authors use this argument (based on the WISH cmyb signals) to infer that the deficit in the cx41.8tq/tq mutant is not due to controlling HSPC number (no difference in cmyb between WT and mutant) but rather upstream, at the level of the hemogenic endothelium, which is not a thorough argument at that point.

    1. Fluorescence imaging and associated interpretation/conclusions.

    The fluorescence images (Fig1E; Fig2B,D; Fig3A) are very difficult to analyze; they lack resolution because they appear to be epifluorescence images and not confocal images. When the signal is low, which is in particular the case for the novel Cx41.8:EGFP fish line, Fig2B (which is confirmed with the FACS GFP signal in comparison to the mCherry of the kdrl:mCherry fish line), it is not possible to provide convincing images on the vascular/aortic expression because of the high background of diffusion (the authors state 'likely to be the aortic floor', indeed it is not possible to validate the fact that the expression is truly in potential hemogenic cells). The double positive population in the FACS (Fig2C, right) does not resolve the issue because if indeed cx41.8 is expressed in endothelial cells (as expected from previous studies), the double positive population could equally be endothelial cells from inter-somitic vessels, for example (not to mention the underlying vein which is very close to the aorta in the trunk)). Fig2D, images are too small and, again, the resolution is not good enough to say that double positive cells are on the aortic floor. It is recommended to convince the reader that the authors try to confirm their statements by using confocal microscopy and increase the magnification of the relevant regions of interest.

    There is an inconsistency in the data between Fig1E (40hpf, in vivo imaging using the cmyb:GFP fish line) and FigS2 (48hpf, WISH cmyb); how can we observe 'HSPCs budding from the dorsal aorta' (see legend Fig1, arrowheads) which seems very much decreased in the imaging experiment for the cx41.8tq/tq mutant in comparison to WT, and have no effect on the cmyb signals FigS2B? What are the GFP+ cells that are aligned along the elongated yolk Fig1E and that appeared to be decreased in number in the mutant? Are the authors sure of their statement on budding HSPCs when the GFP signal pointed by arrows could in majority be hemogenic cells? (which would be in favor of their hypothesis on Cx41.8 being involved rather in hemogenic endothelium/HSPC specification).

    Other Major Comments:

    • It would be important to investigate/show, at least with qualitative WISH experiments all along the time-window of HSPC specification as stated by the authors (26-54hpf, see main text third paragraph of Results), that Cx41.8 is detected in arterial endothelial cells (and perhaps enriched in the hemogenic endothelium?), in complement to the work they are referring to on transcriptomic data at 24hpf (Ref18 Gurung et al Sci Rep 2022). Ideally, these WISH data should be resolutive enough to provide clear localization in aortic cells versus cells in the aortic floor to bring significant added value to the work that lacks spatial resolution (ex: fluorescent WISH using confocal microscopy, allowing to superpose signal with cell types (either by double fluorescent WISH (vascular marker + Cx41.8) or superposing fluorescence signals with transmitted light)).
    • As mentioned by the authors in the Discussion, the other connexin Cx43 (Ref 36, Jiang et al 2010) is playing a significant role in HSPC specification in the zebrafish and is expressed in zebrafish arterial cells at 24 hpf. Hence there may be some functional redundancy between Cx43 and Cx48.1, as supported by previous work from the authors showing that a null mutant of Cx48.1 does not exhibit any phenotype in HSPC specification (Ref12, Cacialli et al 2021). This may be problematic for the experiments using drug treatments in the present work, because they are not selective for the different connexins (ex: anti-oxydants (NAC), connexin blockers (heptanol, CBX)), thus blurring interpretations on the specific function of Cx48.1 versus the ones exerted by Cx43 (this should be also valid for the vhl MO treatments). This comment is strengthened by the fact that the authors do not systematically address, for both WT and mutant embryos (Fig3 E, F; FigS6; FigS8), if expression levels with drugs/H2O2/MO are different for the 2 conditions (if relatively equal, it would indeed indicate that these drugs/conditions possibly act on another connexin, which would help the authors in their analyses and interpretations).

    Minor comments

    • The authors should take care of the fact that at 16hpf, it is an overstatement to speak of an aorta when the cord is starting to lumenize at around 18hpf, Jin et al Development 2005 (see Main text referring to Fig3). To make the data more convincing on the ROS production in the ventral side of the cord in wild type embryos (which suggests that future hemogenic cells are already ventralized at that stage), it would be important to obtain confocal images of the region of interest and perform reconstitution of Z-stacks with a sagittal view (rather than longitudinal). It would be nice also to obtain comparable images later on, after lumenization and before initiation of HSPC emergence (before 28hpf).
    • The authors may try to rescue the wt phenotype by expressing, in the Cx48.1tq/tq mutants, the mRNA encoding for the wt protein.
    • It would be more informative and secure, Fig2D, to show images of the double transgenics (Cx48.1:eGFP;kdrl:mCherry) at 28-30 hpf (rather than 48 hpf) which is more narrowed down to the specification of the hemogenic endothelium thus preventing any risk to visualize the fluorescence signals coming from recently born HSPCs rather than signals from cells embedded in the aortic floor.

    Significance

    Petzold et al propose a potentially appealing function of connexin Cx48.1 expressed in the zebrafish in the specification of the vascular aortic subtype of cells that will ultimately lead to the formation of hematopoietic stem cell precursors, ie the hemogenic endothelium. They build the work on a possible translation of the function of connexin Cx40 in mammals that is described to localize to mitochondrial membranes in endothelial cells and promote the production of ROS in mitochondria. They propose a function of mitochondria-derived ROS in stabilizing the Hif1/2-Notch1 pathway that is essential for HSPC precursor specification and that may be extended to developmental hematopoiesis in mammals (the putative ortholog of zebrafish Cx48.1 in mammals (Cx40) is highly expressed in the hemogenic endothelium of mouse and human species (see the Discussion paragraph)).

    The proposed model is potentially of high significance for the field of hematopoiesis and more generally for translation of knowledge to regenerative medicine aimed at producing hematopoietic stem cells endowed with long term regenerative potential. However, the current work remains preliminary, suffering from lack of resolution in the main experimental axes that are undertaken (WISH analyses and their low accuracy quantifications; low resolution of in situ live imaging; apparent weaknesses of methodologies that are difficult to fully appreciate since poorly detailed in the Method section, in particular regarding WISH quantification and, hence, statistical significance). My recommendation is that the authors should put some efforts in completing the work with other, more quantitative, methodologies (ex: qRT-PCR) and improving the quality/resolution of imaging (by providing confocal images to alleviate any ambiguity on what is visualized and strengthen the results); these are technical approaches that are relatively standard in the field and the authors have extensively used qRT-PCR and FACS-sorting in their previously published work. Also, the endogenous expression of Cx48.1 in the hemogenic endothelium, during the time-window of its specification (20-28hpf), should be addressed; this would be essential to complement the imaging performed with the new transgenic line that expresses eGFP under the control of the Cx48.1 promoter and which provides weak fluorescence signals).

  4. Review Commons

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

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary

    The manuscript by Petzold et al. explores the functions of connexin 41.8 (cx41.8) (mammalian homologue Connexin 40) in hematopoietic stem cell (HSC) formation in the zebrafish dorsal aorta. The authors use a cx41.8 allele that appears to be hypomorphic, as the phenotype is milder than a previous cx41.8 allele that the same group published (Cacialli et al., 2021). cx41.8tq/tq mutants exhibit delayed onset of hemogenic endothelial specification, as marked by gata2b at 24 hpf, but HSPC development proceeds normally from 48 hpf onwards. A new reporter line for cx41.8, Tg(cx41.8:GFP), was generated and is expressed in the floor of the dorsal aorta, consistent with the location of hemogenic endothelial cells. Lower ROS production in the whole cell and in the mitochondria was reported in the cx41.8tq/tq mutants, and treatment with ROS enhancers, H2O2 and menadione, appeared to rescue the mutant phenotype of reduced HSPCs at 28 hpf. Finally, the authors tested a link between cx41.8 and Hif1α by pharmaceutically (DMOG/CoCl2) or genetically (vhl morpholino) inhibiting Hif inhibitors, and observed a rescue of HSPC formation in cx41.8 mutants.

    Major comments

    • I think it would be important for the authors to address the mechanisms of why cx41.8tq/tq and the other cx41.8-/- (leot1/t1) mutant phenotypes are different, with the latter allele showing more severe phenotypes of increased HSPC apoptosis and reduced HSPCs during later development. The authors speculate the cx41.8tq/tq allele encodes a missense mutation in one of the channel domains, and as such, might be a hypomorph. The authors cited the original paper by Watanabe et al. (2006); however, this paper actually noted that the cx41.8tq/tq allele is likely to be a dominant negative - and as such, should have exhibited a stronger phenotype than the leot1/t1 mutant allele. From the paper: "leotw28 and leotq270 heterozygotes have phenotypes different from that of WT; thus, they represent dominant-negative alleles." Importantly, no data are shown to provide evidence that the allele is a hypomorph - at minimum, qPCR data should be provided to show whether there is NMD of the mRNA in cx41.8tq/tq mutants.
    • One major missing component is experimental data that distinguish the gap junction/plasma membrane- related and the mitochondrial membrane-related functions of Cx41.8. This is critical, as the role of Connexins in the mitochondria remains poorly understood (and Connexin 43 is the best understood one). Thus, it is a big claim by the authors that Cx41.8 primarily acts through the mitochondria and not the gap junctions. Suggested experiment: The authors should generate a fluorophore-tagged Cx41.8 - under a ubiquitous (ubb or actin) or HSPC-/hemogenic endothelium-specific (gata2b) promoter to monitor the protein localization of Cx41.8. Providing data on whether Cx41.8 protein indeed localizes to the mitochondria is important to support their claim.
    • Related to the above point, the authors should test whether the gap junction function of Cx41.8 is intact in the cx41.8tq/tq mutants by assessing calcium waves in the GCamp transgenic line.
    • The quantification data in this manuscript are not satisfactory. The authors only provide graphs that show embryos with "low", "medium" and "high" numbers of HSPCs, which is incredibly subjective. Considering that the authors already have the cx41.8tq/tq in the Tg(myb:GFP) background (Figure 1E), they could have quantified the precise numbers of Tg(myb:GFP)-positive cells at different timepoints and with the different pharmaceutical rescue experiments. Ideally, this should be combined with other HSPC markers such as Tg(cd41:GFP) or Tg(runx1:GFP) - although this could be limited by the authors' access to the lines or time it takes to cross the mutants to the transgenes.
    • The link between cx41.8 and Hif1α is tenuous. The authors should perform in situ hybridization for the hif1 genes and their downstream effector notch1 which is known to be important for the HSPC specification (Gerri et al., 2018). The authors might also want to consider performing transcriptomic analysis (bulk RNA sequencing) from purified HSCs in wild types and cx41.8 mutants and assess the downstream pathways affected by the loss of this gene.

    Significance

    Overall, this study presents another piece of evidence that Connexin 41.8 regulates HSC formation. It provides a potential link between Connexin 41.8, mitochondrial ROS regulation and Hif/hypoxia-sensitive pathways in promoting endothelial-to-hematopoietic transition. The role of the mitochondrial ROS in particular is quite interesting and might provide a new angle into the role of connexins in regulating hemato-vascular development; however, the authors would need to strengthen the link between Cx41.8 and mitochondrial respiration.

    It is important to note that the quantitative data in this manuscript need to be strengthened and refined to strengthen the conclusions. The study is not very deeply mechanistic and appears to be more at an observational/correlational level. The manuscript might be of interest for people in the hematopoietic field but does not shed much more insight into the cellular and molecular mechanisms that govern HSC formation, particularly in light of the paper on Cx41.8 role by the same group (Cacialli et al., 2021).

  5. Version published to 10.1101/2023.07.27.550806v2 on bioRxiv
  6. Version published to 10.1101/2023.07.27.550806v1 on bioRxiv