Midkine is a dual regulator of wound epidermis development and inflammation during the initiation of limb regeneration

Reviewer #1:

[…]

I think some possible revisions that could illuminate this would be-

– Does ectopic expression of midkine result in elevated monocytes or recruit immune cells? This should be done in a regenerating context as well as in intact limbs (injection of the constructs into non-regenerating tissue). Likewise, looking at monocyte numbers in the author's already-generated ectopic expression tissue (Figure 7) might give them some clues.

We thank the reviewer for this suggestion. We performed additional experiments to address this point. We found that ectopic expression of midkine in a regenerating context also led to an elevated density of monocytes. While control tdT-overexpressing limbs exhibited a significant decrease in the number of monocytes, from 6 -16 dpa, mk-overexpressing limbs showed consistently high level of monocytes, indicating aberrantly high levels of mk can induce persistent inflammation. These additional data have now been included in Figure 7K-M. We also performed mk overexpression experiments in intact limbs and saw no difference in proliferation or monocyte density, suggesting mk overexpression in this context is not sufficient to regulate these processes during homeostasis. The data from these experiments are described in the manuscript and presented in Figure 7—figure supplement 4. Altogether, these data provide supportive evidence that mk levels must be tightly regulated for proper resolution of inflammation during regeneration.

– Grafting experiments between mk-^null and wildtype animals. Taking advantage of one of the key aspects of the axolotl system, the authors could transplant epithelial tissue between a mk-^null animal and wildtype host animal (perhaps GFP+). This would inform you if 1) WT epithelia is sufficient to rescue the regeneration delay in mk-^null animals, and 2) if only lacking MK expression in the epithelia can lead to the same phenotypes (limited proliferation, increased apoptosis) observed in the completely null animal. I think this could go a long way toward proving the essential nature of midkine for AEC development.

We thank the reviewers for proposing this elegant experiment and agree that whether the presence of mk in the epithelia alone is sufficient to rescue the phenotype and vice versa is an interesting and important question. On this note, we had originally thought about conducting this exact experiment, however we decided for several reasons that the results from grafting experiments would be confounding. Grafting just the wound epidermis alone is technically difficult and has a very low success rate. To perform this experiment in the most reproducible manner, we would instead graft a cuff of full thickness skin from a mk^null mutant to a wildtype intact limb and vice versa, then amputate through the graft and observe regeneration. Grafting full thickness skin would also include prrx1⁺ dermal fibroblasts that highly express mk and are major contributors to the blastema (Kragl et al., 2009, Gerber et al., 2018). Therefore, if we observed normal wound epidermis expansion when grafting wildtype epithelia onto mutant limbs, we could not conclude that the rescue in phenotype was due to the presence of mk in only the epithelia. While performing this grafting experiment could not directly address the question posed, we did agree that this was an important issue. Therefore, we asked the converse question of whether mk derived from stump tissues is sufficient to rescue the mutant phenotypes. To do this, we exploited the fact that electroporating constructs into the limb epidermis is difficult and only occurs at extremely low efficiency (Fei et al., 2016). We performed a rescue experiment, in which we overexpressed either tdTomato (as a control) and/or mk in mk^null mutants. As expected, the majority of mk was expressed and secreted from regenerating stump tissues as can be seen in Figure 6—figure supplement 3. Overexpression of mk in mk^null mutants was sufficient to rescue the delayed regeneration, AEC developmental defects (thin wound epidermis/high cell death), and high monocyte density. These data altogether suggest 1) that the wound epidermis is responsive to mk secreted from non-epithelial sources and 2) that mk from stump tissues is capable of signaling to the wound epidermis and likely vice versa. These additional data can be found in Figure 6—figure supplements 1 and 2.

– Does ablation of macrophages by chlodronate liposomes also inhibit midkine expression similarly to skin flap suture?

This is an interesting question to explore. In response, we attempted a small pilot experiment in a collaborative effort with Dr. Nicholas Leigh, a postdoctoral fellow in the Whited lab. The original dosage in Godwin et al., 2013, used 12.5 mg/kg and we therefore bracketed this using 2x, 1x, 0.5x mg/kg. Each condition had three animals (1 control, 2 experimental). Unfortunately, in our hands, treatment with clodronate caused severe edema at all three doses which we presumed was due to cytotoxicity associated with the clodronate and/or opportunistic infections in both the regenerating and intact hindlimbs of depleted animals. As a result, many animals had to be euthanized. These cytotoxic effects precluded the interpretation of the results on mk expression (data not shown) and the optimization of this method extends beyond the scope of the current manuscript. However, we do agree this is a possibility and have added this as a future direction in the Discussion (paragraph seven).

– Do midkine-disrupted animals form a proper wound epidermis without a basement membrane or is it similar to the full skin flap suture? Deposition of collagens along the wound epidermis was also observed after macrophage depletion (Godwin et al., 2013) and this could link midkine and regeneration phenotype observed.

Picro-mallory staining of the wound epidermis/AEC in both iMDK- and mk^null mutants indicates a lack of a basement membrane similar to a normal wound epidermis (Figure 4D’ and G’, absence of blue collagen staining at the basal surface of the wound epidermis). On the other hand, picro-mallory staining of the full skin flap sutured limbs shows a clear blue collagen thick basement membrane and dermal layer (Figure 1—figure supplement 1B’). Therefore, mk-disrupted animals appear to form a proper wound epidermis with regards to this aspect. Furthermore, the basement membrane and blue collagen thick dermal layer could be seen re-forming at the amputation plane in iMDK-treated limbs much later at 40 dpa (now included in Figure 4—figure supplement 1), providing further evidence for its absence during earlier stages when AEC development normally occurs.

Reviewer #2:

– In order to substantiate that the phenotypes observed following chemical and genetic mk perturbation are due to the specific loss of mk -particularly given the discrepancies of phenotypes observed between mk-/- and iMDK-, a rescue experiment (eg overexpression of mk through electroporation in KO -expected to rescue- or iMDK-treated animals -expected not to rescue if the treatment is sustained-) should be conducted.

We thank the reviewer for this suggestion and agree that this is an important point to address given the difference in phenotypes between iMDK treatment and mk^null mutants. As such, we performed a rescue experiment in which we overexpressed mk in mk^null mutants beginning at 3 dpa (the time point when mk exhibits strong induction). Overexpression of mk, but not tdTomato (as a control), in mk^null mutants was sufficient to rescue the associated phenotypes including delayed regeneration, defective AEC development (thin wound epidermis and higher cell death), as well as high monocyte density associated with persistent inflammation. We believe that these additional data (which can be found in Figure 6—figure supplements 1, 2 and 3) provide strong evidence that mk indeed regulates both AEC development and the resolution of inflammation.

– To discard non-specific effects of CRIPR targeting, data from animals generated with a second gRNA (targeting a different site) should be presented.

While targeting mk with a second gRNA against a different site would provide additional evidence to reinforce our conclusions, we believe that the correlation of phenotypes between two independent loss-of-function methods (chemical and genetic) in our study provides sufficient evidence for our claims. Additionally, our mk rescue experiment in the mutant further supports our conclusions that the phenotypes we observed were indeed due to mk deficiency and not off-target effects of CRISPR. Together, we believe these data are sufficient to support our conclusions that mk is a regulator of AEC development and inflammation.

– The authors show that iMDK treatment does not affect innervation, but does mk expression itself depend on innervation? This should be address by presenting mk analysis in denervated vs innervated limb blastemas.

We thank the reviewer for raising this interesting point. We have taken advantage of a previous published microarray dataset in both innervated and denervated regenerating axolotl limbs to answer this question. The data from this study determined that mk was up-regulated at a similar fold change in both innervated and denervated limbs indicating its upregulation is independent of innervation. We have now included this commentary in the Results (paragraph three in subsection “Mk loss-of-function leads to impaired AEC development and persistent inflammation”) along with referencing the appropriate study (Monaghan et al., 2009).

– The authors focus on the role of MK during the expansion of the wound epidermis into the AEC, however they do not touch on the possible roles of MK itself in the blastema (notably, prrx1⁺ cells display substantial MK expression, and loss of MK leads to increased inflammation in the entire regenerate, which could significantly impact blastema cell behaviour). This should be discussed and supported by data whenever possible. For example, EdU and TUNEL data should be shown for blastema progenitors, not just for the wound epidermis, as mk is also expressed in blastemal prrx1⁺ connective tissue.

We agree with the reviewer on this point and have conducted additional analyses on blastemal cell death and proliferation in iMDK-treated and mk^null regenerating limbs. While we did observe decreased proliferation and persistently higher levels of cell death in blastemal cells of iMDK-treated limbs, we did not observe differences in blastemal cell proliferation or death in mk^null mutants. These experimental data are shown in Figure 5—figure supplement 1 and Figure 7—figure supplement 1. Together, these data suggest that mk is not required for blastemal cell proliferation or cell survival. Alternatively, the discrepancies in the phenotypes could be attributed to off target effects of the inhibitor, which we state and elaborate on in the Discussion (paragraph six).

– Furthermore, they fail to explore a possible dynamic interaction between the wound epidermis and blastema cells through MK: this is evident on Figure 8, where MK receptors sdc-1 and ptprz are expressed at different times in either the wep and/or blastema cells. At 7dpa, it seems that ptprz is expressed in blastema cells exclusively (also, sdc-1 is expressed in blastema cells at 14dpa but not at 7dpa) and could represent an example of paracrine interaction where MK secreted by the wound epidermis affects blastema cells. Unfortunately, these issues are neither explored nor discussed, but are important in order to understand MK function.

We apologize if the initial ptprz in situ at 7 dpa was misleading. The strong blue signal shown in the stump tissues was non-specific background staining from the fibrin clot at the amputation plane. We have included new in situs at 7 dpa for ptprz that more clearly demonstrate low expression of ptprz (dark blue puncta) throughout the wound epidermis with minimal background staining in the clot (Figure 8C). This was also validated by single cell RNA-seq expression data from Leigh et al., 2018 (Figure 8—figure supplement 1). That said, we thank the reviewer for bringing this to our attention and agree that a more extensive analysis and discussion of the potential signaling mechanisms of mk was needed to put the findings in this study into a broader regenerative context.

We have now included a more comprehensive analysis of the expression of midkine receptors in our dataset as well as publicly available single cell data (Figure 8 and Figure 8—figure supplement 1). We find overlapping and non-overlapping expression patterns for mk receptors and mk, indicating mk is capable of both autocrine and paracrine signaling. Furthermore, we find the expression of multiple mk receptors in the wound epidermis, indicating mk likely directly binds to and regulates wound epidermis/AEC development. However, by mining the higher resolution single cell data, we were surprised to find that mk receptors were largely detected at low levels or not detected at all in myeloid cell populations, with the exception of integrin β 1. These data bring forth the possibilities that mk may: 1) directly regulate myeloid cell populations through the low expression of these or as of yet unidentified receptors, 2) indirectly regulate inflammation via signaling through other tissues, or 3) modulate inflammation through both. Given the mk receptor expression data, we speculate in the Discussion that mk is likely regulating the resolution of inflammation through both mechanisms and elaborate on how our functional and transcriptional data suggest that the wound epidermis may be an example of one tissue that may indirectly mediate resolution of inflammation.

– In this connection, it would be helpful to add additional timepoints between 1 and 3 dpa in Figure 2 in order to define whether mk expression comes first in the wound epidermis, in connective tissue cells, or in both at the same time.

We agree with the reviewer that a more in-depth analysis of the onset of mk expression in regenerating tissues would provide important clues as to the dynamics of mk signaling. In response, we have now included an additional timepoint, 56 hours post-amputation (hpa), and high magnification images of mk expression in both the wound epidermis and stump tissues at 24, 56, and 72 hpa demonstrating that mk is first expressed in cells within regenerating stump tissues prior to the wound epidermis (Figure 2—figure supplement 2A). These data therefore suggest that mk expression in the wound epidermis may first be induced through paracrine signaling from mk and/or other extracellular molecules in regenerating stump tissues.

Reviewer #3:

1) The main results are that inhibition of midkine results in a loss of regeneration due to a loss of maintenance of proliferation. However in the CRISPR knockout the blastema still forms, it is smaller. Another possibility that is not considered is that cell division still proceeds but is slower. If the CRISPR regeneration are examined after much longer time periods does it complete regeneration?

We agree with the reviewer on this point and now include data indicating that the mk^null mutants do not exhibit defects in blastemal cell proliferation, indicating mk is likely not required for blastemal cell division. The mk^null mutants do regenerate anatomically normal limbs, although delayed and we have now included time-lapse imaging in Figure 6—figure supplement 1 to demonstrate the delay in mutants compared to wildtype and mk-rescued animals.

2) In situ hybridization and immune identifies midkine to be expressed in blastema cells, which are also prxx positive, the interpretation is that 60% of the midkine positive cells are "connective tissue blastema cells". The authors need to look at other markers of cell types that contribute to the blastema like dermal fibroblasts (vimentin), satellite cells(Pax7). Figure 1A shows midkine expression in uninjured cells, what cells are these, maybe satellite cells?

We thank the reviewer for raising this issue and have conducted a more detailed double in situ hybridization analysis to identify the cell types that express mk during homeostasis and regeneration. The analysis shows that the majority of mk-expressing cells are pax7⁺ satellite cells and pecam⁺ endothelial cells, in addition to prrx1⁺ connective tissue cells, in both intact and regenerating tissues. We quantified the breakdown of cell types expressing mk in regenerating stump tissues and included representative images of mk double in situs in both intact and regenerating tissues (3, 7, and 24 dpa). In intact tissues, connective tissue cells, satellite cells, and endothelial cells comprised approximately 43.8%, 44.9% and 7.7% of mk-expressing cells in stump tissues, respectively. During regeneration, the majority of mk-expressing cells were connective tissue cells (~53-54%) and satellite cells (~36-37%) at 3 and 7 dpa. A smaller percentage (~4-5%) of mk-expressing cells were endothelial cells at both time points. These data are now included in the study and can be found in Figure 2H-I and in Figure 2—figure supplement 2B.

3) Figure 7H claims that overexpression of midkine results in increased proliferation of wound epidermis cells, however in the figure it is clear that the Edu positive cells are in the blastema not the wound epidermis.

We apologize if this was unclear in the figure. The EdU+ cells are indeed in the wound epidermis as outlined in the dotted yellow lines in each panel of Figure 7I. At 8 dpa, a blastema has not formed in both tdT- and mk-overexpressing limbs (this time point is pre-blastema formation). Furthermore, additional quantification of EdU+ cells in the regenerating stump tissues revealed a significant decrease, not increase in proliferation of blastemal cells at 8 and 16 dpa. The additional data quantifying proliferation in blastemal cells is shown in Figure 7J.

4) In what cells is midkine overexpressed? What percentage of the cells in the blastema is it overexpressed?

Overexpression of constructs via electroporation in axolotl limbs predominantly results in expression in stump tissues, as the electroporation into the epithelial tissues is difficult and occurs at exceedingly low efficiencies. The most obvious expression can be seen in myofibers, however other cell types can also uptake constructs through electroporation (Fei et al., 2016). We have included representative images of tdTomato staining in both our mk overexpression experiment during regeneration and our newly added mk rescue experiment in Figure 7—figure supplement 2 and Figure 6—figure supplement 3, respectively. In both experiments, quantification showed that up to 20% of cells constitutively expressed tdTomato or mk. We also show equivalent efficiencies by both cell quantification and mean fluorescence intensity across control and experimental conditions as supportive control data to discount the possibility that the phenotypic effects were due to varying electroporation efficiencies.

5) Can overexpression of midkine rescue the CRISPR knockout phenotype.

We thank the reviewer for raising this important point and have now included data indicating that overexpression of mk beginning at 3 dpa in our mk^null mutants is sufficient to rescue the delayed regeneration, AEC developmental defects (thin wound epidermis and higher cell death), as well as higher monocyte density. These data can be found in Figure 6—figure supplement 1, 2 and 3.

6) Are the limbs of the CRISPR knockout animal the same size as a wild type control and does the embryo develop at the same rate. A figure should be included to show these data.

We have now included representative images of mkWT and mk^null mutants indicating no visible difference in size of the animals as well as anatomically normal limbs. These data can be found in Figure 3—figure supplement 1.

7) The Discussion focuses a lot on the potential interplay between midkine and the immune system, although the RNA seq data suggests this there is very little data in the paper to support this. Does the overexpression of midkine then have an effect on the number of monocytes present or only on cell proliferation? The Discussion should be modified to reflect more the data present in the paper and then put in the broader context of regeneration

We agree that additional evidence is needed to support these claims. As such, we now include data showing that overexpression of mk during regeneration leads to consistently higher monocyte density compared to control limbs, indicative of chronic inflammation (Figure 7K-M). Together with the loss-of-function data, these data show that mk levels need to be tightly maintained to proper modulate the resolution of inflammation. We also present additional data showing that mk overexpression leads to decreased blastemal cell proliferation despite seeing the opposite effects in the wound epidermis (Figure 7J). We have expanded the Discussion section to incorporate all of these data into the broader context of the roles of mk in regeneration.

8) There is data from the zebrafish field on the role of midkine in retina regeneration this should be referred to in the Discussion.

We thank the reviewer for bringing this to our attention and have now included the relevant references in the Discussion (paragraph six) (Gramage et al., 2015, Nagashima et al., 2019).

9) In general the N for a lot of experiments is very low, in some animal experiments the N=4, which means the data are not statistically relevant. The n should be increased to a minimum of 10 for all animal experiments.

While the reviewer brings up a valid and important point, we note that similar statistically significant effects are observed across two different loss-of-function methods and corroborative gain of- function experimental evidence for the role of mk in both wound epidermis development and inflammation. Moreover, the results from the mk rescue experiment in the mutants provide new additional data to support our claims. The statistical methods utilized are included in the legend of each figure as well as the Quantification and Statistical Analysis section of the Materials and methods.

10) The Materials and methods section is incomplete, RNAscope is used for several figure but there is no methods for it included.

We thank the reviewer for pointing this out and have now included a detailed section in the Materials and methods on RNAscope.