Integrin-linked kinase controls retinal angiogenesis and is linked to Wnt signaling and exudative vitreoretinopathy

Familial exudative vitreoretinopathy (FEVR) is a human disease characterized by defective retinal angiogenesis and associated complications that can result in vision loss. Defective Wnt/β-catenin signaling is an established cause of FEVR, whereas other molecular alterations contributing to the disease remain insufficiently understood. Here, we show that integrin-linked kinase (ILK), a mediator of cell-matrix interactions, is indispensable for retinal angiogenesis. Inactivation of the murine Ilk gene in postnatal endothelial cells results in sprouting defects, reduced endothelial proliferation and disruption of the blood-retina barrier, resembling phenotypes seen in established mouse models of FEVR. Retinal vascularization defects are phenocopied by inducible inactivation of the gene for α-parvin (Parva), an interactor of ILK. Screening genomic DNA samples from exudative vitreoretinopathy patients identifies three distinct mutations in human ILK, which compromise the function of the gene product in vitro. Together, our data suggest that defective cell-matrix interactions are linked to Wnt signaling and FEVR.

2. Abstract, page 2, line 44-46: I do not agree to this sentence, as especially the work done on the Norrin/beta-catenin signaling pathway (NDP, FZD4, LRP5 and TSPAN12) has largely increased our insights on the pathophysiological mechanisms underlying FEVR. Please rephrase.
3. One of my major concerns involves the lack of data showing that the addition of tamoxifen in the IlkiECKO mouse model indeed results in a decreased expression or function of Ilk. A Q-PCR, Western blot and/or immunostaining for ILK showing that indeed Ilk transcripts or proteins are reduced would be essential to link the bona fide phenotype in the mice directly to Ilk. Figures 1 and 3 show quite some immunostainings, although from the description, it is sometimes not clear what exactly is stained and, more importantly, what it implicates. A nice example of how it should be explained is on page 7, lines 160-161, where the authors state that "Expression of Pvlap/PV1, which forms diaphragms in endothelial fenestrae….". Please use similar descriptions for all markers used. 5. Page 5, line 127. All of a sudden, the authors switch form a single tamoxifen delivery to three consecutive administrations. Please explain the rationale for this.

Figures 1, 3 and Supplemental
6. Page 7, line 176. The use of the ParvaiECKO model comes out of the blue. Please spend one or two sentences on why this model was selected. 7. Page 7, line 182 and Discussion. Given the importance of heterozygous changes detected in humans, and the presence of a somewhat milder vascular phenotype in heterozygous IlkiECKO/+ mice, I'm totally lacking a discussion of what is observed in constitutive Ilk+/-mice. 8. My other major concern involves the relatively high frequency for the ILK variants described in Exac, especially for p.Arg211Cys and also for p.Leu53Met. Although the authors claim that the occurrence of these variants are not extremely frequent, I think this frequency is too high to underlie a dominant condition such as FEVR. For instance, for p.Arg211Cys, with a frequency of 0.084%, 8 out of 10.000 (appx. 1:1,250 individuals) would have FEVR, only due to this mutation (assuming full penetrance). This is way too high for a rare dominant condition such as FEVR. In addition, the authors do not demonstrate nor discuss any segregation analysis, presence of this variant within healthy or affected relatives, or whether these variants occur de novo or are inherited. Together, despite the fact that the authors clearly demonstrate that the missense variants can affect ILK protein function (Figure 7), the conclusion that these variant underlie FEVR to me is not sufficiently proven.
9. The Discussion section is short, and lacks certain components, some of which are addressed above. Also, the discussion on the Itgb1iECKO model now appears unexpected. Finally, the last paragraph of the Discussion (linking ILK to cardiomyopathy) to me feels inappropriate at the very end of a paper on vitreoretinopathy.
10. Figures 1a and 2a: I propose to expand these figures, by not only putting the tamoxifen but also showing the schematic build-up of the IlkiECKO model (especially in 1a), and indicate an arrow for each tamoxifen delivery (in figure 2a).
Minor issues: 11. Page 5, line 116: the sentence starting with confirming should not be a separate sentence but rather an extension of the previous one. 14. The yellow traces in Figure 6a are hard to see, please use a different coloring.
Reviewer #3 -expert in retinal angiogenesis (Remarks to the Author): This paper by Yamamoto et al. investigates the role of integrin-linked kinase (Ilk) in familial exudative vitreoretinopathy (FEVR). Inducible endothelial cell (EC)-specific Ilk KO mice were generated and analyzed phenotypically, which demonstrated defective retinal vasculature resembling FEVR symptoms. Genetics analysis in FEVR patients identified three loss-of-function mutations in Ilk, which showed impaired biological function in protein binding and vascular endothelial cell spreading and focal adhesion. This is a well conducted and convincing study identifying IlK as a new disease gene for FEVR, which is linked largely with Wnt signaling defects previously. The novel findings are of high interest to researchers in the fields of vascular biology, ophthalmology and eye research. There are just a few concerns from this reviewer, and addressing them may help further strengthen the manuscript.
Major concerns:

1.
Where is endogenous Ilk localized in the retina? The inducible Ilk KO data presumes vascular endothelial localization but that assumption is not proven with immunohistochemistry. Can Ilk be expressed elsewhere in the retina to exert vascular effects?

2.
The EC specificity of the PDGFb-iCre mice needs to be established, eg. by crossing with a reporter strain to exclude the potential function of Ilk in other retinal cell types (glia and neurons) in causing the vascular abnormalities. Validation of successful IlK knockout in retinal ECs will also be helpful.

3.
Another primary feature of FEVR is persistent hyaloid. Do the Ilk iECKO eyes show delayed hyaloid regression?

4.
Mechanistically, it will be valuable to delineate whether Ilk signaling is somehow linked with Wnt signaling and actin dynamics. Previously mice with defective Wnt signaling and Akt/Girdin (an actin-binding protein) signaling also showed very similar impairment in retinal vasculature. Do IlK knockout retinas/ECs have impaired Wnt signaling or actin cytoskeleton dynamics, eg. altered phosphorylation of b-catenin and any changes in actin/Girdin? Minor concerns: 1. Fig. 5. The vascular phenotype in heterozygous mice is very mild. Do they have reduced body weight, and if so, how can the authors rule out delayed development as a possible cause of secondary vascular developmental delay in hets? It is of interest to note that most other heterozygous Wnt deficient FEVR mice (Fzd4+/-and Lrp5+/-) are phenotypically normal. Please comment. Fig. 7D, define M1, M2, M3.

3.
Define abbreviation the first time of use in the main texts (FEVR, line 61); Line 190, FEVR patients (missing an "F").
We would like to thank all reviewers for their time, effort and valuable suggestions, which are greatly appreciated and have enabled us to improve the manuscript further. While a detailed point-by-point response to all comments and questions is provided below, we would like to emphasize that a substantial amount of new results has been added to the manuscript. Most importantly, we have added new data showing that the loss of ILK leads to alterations in Wnt signaling.
We would also like to apologize for the overly long time taken up by the revision.
This was caused by breeding issues in our Ilk mouse colony, which have substantially delayed all in vivo experiments.

Reviewer #1
In Overall the work is of sound quality, and most of the data are of high quality and well presented.
However, from the mechanistic perspective, the paper could be improved by addressing the following : Question 1. The authors refer extensively to the role of Wnt signaling in FEVR.

ILK has been shown in numerous studies to influence the Wnt signaling pathway, specifically beta-catenin stablization, nuclear localization and beta-catenin mediated transcription. Given the overlap between the Wnt FEVR and ILK FEVR phenotypes, the authors should examine the status of beta-catenin and TCF
activity in the mouse models and in the ILK mutant transfected HUVECs.

Reply:
We agree with the reviewer that this is an important question. In the revised manuscript, we show strongly reduced endothelial beta-catenin immunosignals in Ilk iECKO retina relative to littermate control sections (Fig. 3E).
Strong expression and nuclear localization of LEF1, a member of the TCF/LEF transcription factor family that binds to beta-catenin and mediates canonical Wnt signaling, marks control but not Ilk iECKO mutants veins (Suppl. Fig. 2C). PVLAP expression, which is normally suppressed by beta-catenin, is increased in Ilk iECKO retinal vessel. In vitro, we found that Wnt target genes are downregulated by after siRNA-mediated ILK knockdown both in human umbilical vein endothelial cells (HUVECs) and HRECs (human retinal ECs) ( Fig. 7E and Suppl. Fig. 5B). ILK knockdown in HUVECs also results in a reduction of beta-catenin levels ( Fig. 7A), which is rescued by the re-expression of siRNA-resistant WT ILK but not by any of the 3 ILK variant proteins. Together, these data indicate that ILK is required for canonical Wnt signaling, which provides a mechanistic explanation for the striking phenotypic similarities between Ilk and Wnt pathway loss-of-function mutants. Fig 1G,   Reply: We found a downregulation of phospho-AKT immunosignals in Ilk iECKO retinal endothelial cells relative to littermate control (new data in Fig. 7D). In HUVECs, ILK siRNA treatment reduces Akt Ser473 phosphorylation, which is rescued by re-expression of siRNA-resistant WT ILK but not any of the point mutants (Fig. 7A).

Question 4. Recently a new interactor of ILK, called LIMD2, has been described:
LIMD2 binds directly to ILK within the kinase domain and this interaction is essential for the pro-migratory role of ILK. Furthermore, in vitro, LIMD2 stimulates ILK activity (Peng H et al, Cancer Res. 74, 1390, 2014. The authors should determine whether the R317Q ILK mutant is defective in LIMD2 binding ( by coimmunoprecipitation). If this were the case, this would provide additional , critical mechanistic data for the requirement of ILK in post-natal vascularization and related pathologies such as FEVR.
Reply: It will be indeed interesting to study the role of LIMD2 or other interaction partners of ILK in retinal angiogenesis, but, in our view, these questions deserve a separate investigation and are beyond the scope of the current manuscript. Reply: This has been corrected. The data is now shown as Suppl. Fig. 5C. Reply: We agree that it is difficult to prove that any of the 3 missense variants is indeed causing exudative vitreoretinopathy. Human pedigree analysis would be helpful in this context, but, as we point out in the manuscript, the mutations were found in spontaneous cases. Despite of these limitations, our data -namely the Ilk loss-of-function phenotype, the similarities to Wnt pathway mutants, the link between ILK and Norrin-induced cellular responses -argue for a role of ILK in exudative vitreoretinopathy.

Comments to manuscript 'Integrin
To address the reviewer's concern, we have modified some parts of the manuscript and changed the title, which now reads "Integrin-linked kinase controls retinal angiogenesis and is linked to exudative vitreoretinopathy". Reply: We agree and did not mean to question that there is a substantial body of research on the important role of Norrin/beta-catenin signaling in the pathophysiological mechanisms underlying FEVR. We have modified the abstract and introduction to avoid any misunderstandings.

Question 3. One of my major concerns involves the lack of data showing that the addition of tamoxifen in the IlkiECKO mouse model indeed results in a decreased expression or function of Ilk. A Q-PCR, Western blot and/or immunostaining for
ILK showing that indeed Ilk transcripts or proteins are reduced would be essential to link the bona fide phenotype in the mice directly to Ilk.
Reply: Agree. We have added new data showing ILK immunostaining in P6 retina after three tamoxifen injections at postnatal day (P) 1-3 (see scheme in While ILK expression in perivascular cells is maintained in Ilk iECKO retinal samples, endothelial signal is lost (Suppl. Fig. 1A). Figures 1 and 3  Reply: In both treatment regimes, which are now shown in diagrams (See Fig. 1A and Fig. 2A), three consecutive tamoxifen injections were used. As mentioned in the manuscript, we switched to the second scheme (tamoxifen at P3, 5 and 7 followed by analysis at P14) to analyze the vascularization of the deeper retina and also circumvent the limited survival observed after early postnatal tamoxifen administration. This is an important experimental refinement limiting adverse effects in the mutant animals. At the same time, we make sure that we are not analyzing indirect, secondary defects caused by poor health of the mutant animals.  Fig. 4A-D) as what we report for EC-specific Ilk iECKO/+ heterozygotes.

Question 4. Figures 1, 3 and Supplemental
These results further support the notion that endothelial cells respond very sensitively to changes in ILK function.
Question 8. My other major concern involves the relatively high frequency for the ILK variants described in Exac, especially for p.Arg211Cys and also for p.Leu53Met. Although the authors claim that the occurrence of these variants are not extremely frequent, I think this frequency is too high to underlie a dominant condition such as FEVR. For instance, for p.Arg211Cys, with a frequency of 0.084%, 8 out of 10.000 (appx. 1:1,250 individuals) would have FEVR, only due to this mutation (assuming full penetrance). This is way too high for a rare dominant condition such as FEVR. In addition, the authors do not demonstrate nor discuss any segregation analysis, presence of this variant within healthy or affected relatives, or whether these variants occur de novo or are inherited. Together, despite the fact that the authors clearly demonstrate that the missense variants can affect ILK protein function (Figure 7), the conclusion that these variant underlie FEVR to me is not sufficiently proven.
Reply: First of all, we would like to emphasize that, depending on the gene in question, FEVR is can be inherited as autosomal dominant, X-linked or autosomal recessive. Thus, even though X-linked or autosomal dominant inheritance is frequent, this is not a mandatory feature of the disease. In fact, given the critical roles of ILK in early embryonic development, it would be very surprising to find dominant mutations in human subjects. Furthermore, a fraction of subjects carrying ILK mutations may be actually asymptomatic, as is, for example, also the case for cerebral cavernous malformations (e.g. Velz et al. 2018, Front Neurol. 9: 848). Another interesting aspect concerns the population frequency of the mutants. According to ExAC, p.Leu53Met is 10 times more frequent in the Finnish population than among non-Finnish Europeans. The mutation has not been found in African or East Asian subjects. p.Arg317Glu is actually quite rare, while p.Arg211Cys is most frequent in the Latino population.
We do not dispute that further genetic studies are required to establish whether mutations in the human ILK gene can cause exudative vitreoretinopathy or function as a modifier in the context of Wnt pathway mutations. As we have no evidence for familial inheritance of ILK mutations in humans, we have revised the manuscript accordingly.
Question 9. The Discussion section is short, and lacks certain components, some of which are addressed above. Also, the discussion on the Itgb1iECKO model now appears unexpected. Finally, the last paragraph of the Discussion (linking ILK to cardiomyopathy) to me feels inappropriate at the very end of a paper on vitreoretinopathy.

Reply:
We have revised the Discussion to improve the connection between the different paragraphs. The reason for discussing integrin function in the endothelium is the important connection between integrins and ILK in many different settings and cell types. We also think that it is important to mention other pathophysiological settings related to ILK mutations in humans. Reply: Agree. Schematic diagrams are now shown in Fig. 1A and Fig. 2A.

Minor issues:
Question 11. Page 5, line 116: the sentence starting with confirming should not be a separate sentence but rather an extension of the previous one.

Reply:
We have rephrased this sentence. Question 13. Supplemental Figure 3 is not at all mentioned in the main text.
Reply: Thank you for alerting us to this issue, which has been addressed. Figure 6a are hard to see, please use a different coloring.

Reply:
As requested by the reviewer, we have changed the color in Fig. 6A. Reply: As in many other mouse models with defective vascularization of the retina, regression of hyaloid vessels is compromised in P14 Ilk iECKO mutants (see image for reviewers below). Given that this might reflect a general, secondary defect that may not be directly related to EC-specific Ilk inactivation, we have not included this data in the revised manuscript.

Reply:
We agree with the reviewer and have investigated possible links between ILK and Wnt signaling. In the revised manuscript, we show strongly reduced endothelial beta-catenin immunosignals in Ilk iECKO retina relative to littermate control sections (Fig. 3E). Strong expression and nuclear localization of LEF1, a member of the TCF/LEF transcription factor family that binds to beta-catenin and mediates canonical Wnt signaling, marks control but not Ilk iECKO mutants veins (Suppl. Fig. 2C). PVLAP expression, which is normally suppressed by betacatenin, is increased in Ilk iECKO retinal vessel. In vitro, we found that Wnt target genes are downregulated by after siRNA-mediated ILK knockdown both in human umbilical vein endothelial cells (HUVECs) and HRECs (human retinal ECs) ( Fig.   7E and Suppl. Fig. 5B). ILK knockdown in HUVECs also results in a reduction of beta-catenin levels (Fig. 7A), which is rescued by the re-expression of siRNAresistant WT ILK but not by any of the 3 ILK variant proteins. Together, these data indicate that ILK is required for canonical Wnt signaling, which provides a mechanistic explanation for the striking phenotypic similarities between Ilk and Wnt pathway loss-of-function mutants.
Girdin is known to regulate endothelial cell migration and angiogenesis in the postnatal retina is compromised in knockout mice (Kitamura et al. 2008, Nat Cell Biol. 10:329-37). Given that Girdin is a substrate of Akt/PKB, it is quite likely that its function is altered in the absence ILK. This interesting question, however, deserves a separate investigation and is beyond the scope of the current manuscript.

Minor concerns:
Question 5. Fig. 5 Reply: The body weight of heterozygous mutants is slightly reduced but this effect is not statistically significant. Ilk heterozygotes and littermate controls show a similar size of the eye and normal formation of the superficial vascular plexus, arguing against general developmental delay.