EC-specific deletion of Eng results in retinal AVM accompanied by non-cell autonomous mTORC1 activation

To study effects of Eng deletion on AVM establishment in relation to mTOR signalling we utilised mice allowing for tamoxifen-inducible EC-specific Eng deletion (Cdh5(PAC)-CreERT2+/-:Engflox/flox;R26REYFP, hence forth denoted EngiECKO) [5]. In accordance with previous reports, tamoxifen administration at postnatal day (P) 3 resulted in retinal AVMs at P6, evidenced by confocal microscopy of flat-mounted samples stained for ECs (Fig. 1A, B) [3, 5, 23]. Immunostaining for ENG revealed uniform and abundant expression throughout the vasculature of control mice, while the vast majority ECs of EngiECKO mice had lost ENG expression (Fig. 1C). In agreement with our published data, the few remaining non-recombined ECs in EngiECKO retinas mainly localized to arteries (Fig. 1C) [5]. Note the differences in EC morphologies within arteries (elongated), veins (rounder) and AVMs (disorganized and round), exposed by individual ENG positive ECs.

Previous studies have demonstrated increased activation of the PI3K/AKT/mTORC1 signalling axis following deletion of Alk1 (Acvrl1), Eng or its downstream component SMAD4 [9, 11, 12]. To assess potential regulation of mTORC1 downstream of ENG in vivo, we stained P6 retinas of control and EngiECKO mice for phosphorylated Serines 235/236 of RPS6 (p-RPS6), a downstream effector of mTORC1. Analysis revealed significantly increased phosphorylation of RPS6 in ECs located within AVMs but also in arteries, veins and capillaries in non-AVM regions, suggestive of direct EC-mediated effects of ENG deletion (Fig. 1C, D). At the same time, EngiECKO retinas displayed increased p-RPS6 reactivity in non-ECs surrounding AVMs but not in areas adjacent to the unaffected vasculature, compared to control retinas (Fig. 1E). Immunolabelling exposed RPS6 phosphorylation in mural cells (pericytes, smooth muscle cells, CD13+), a subfraction of immune cells (F4/80+) and ganglion cells (RBPMS+) (Suppl. Figure 1). Quantification of cells in the peri-AVM area exposed a near 2-fold increase in the number of F4/80 and p-RPS6 double-positive cells (from average 0.32 to 0.58 cells/103µm2), but unchanged total F4/80 + area. These findings suggest that the increase in p-RPS6 derives from activation of resident microglia rather than by infiltration of peripheral macrophages, further supported by their less ramified morphology (Suppl. Figure 1 A-C). Also, an increase from 9.4% of mural cells (CD13+) being positive for p-RPS6 in controls, to 36.6% in AVM regions were observed (Suppl. Figure 1D, E). No change in ganglion cell p-RPS6 intensity was recorded (Suppl. Figure 1 F, G). Altogether, this demonstrates secondary or systemic impact on RPS6 activation on subsets of retinal cells.

To investigate potential ENG-mediated EC-autonomous regulation of mTORC1 activation, we analysed retinal vasculatures of EngiECKO mice that displayed mosaic EC-specific deletion of ENG, as a result of random incomplete genetic recombination of ECs (Fig. 2A, B). Detailed investigation of recombined (ENG-, CD31+) and non-recombined ECs (ENG+, CD31+) in close proximity to each other, exposed that ENG levels did not directly dictate RPS6 phosphorylation status, irrespective of subvascular location (AVM, capillary, vein, and artery) (Fig. 2B). Quantification of p-RPS6 intensity in ENG + versus ENG null ECs of mosaic veins, arteries as well as capillaries in non-AVM regions, showed no correlation between ENG and p-RPS6. However, in AVM-associated capillaries, ENG null ECs showed 1.7-fold higher p-RPS6 than ENG expressing ECs (Fig. 2C). This suggests a combination of ENG-mediated cell autonomous, as well as non-cell autonomous regulation of RPS6 phosphorylation exclusively in AVM-associated capillaries. At the same time the general increase in p-RPS6 in AVMs in EngiECKO mice demonstrates secondary non-cell autonomous effects – as a consequence of malformation. We further confirmed that RPS6 serines 240/244, that are specifically activated downstream of mTORC1 (and not mTORC2) [24], showed increased phosphorylation in AVMs of EngiECKO mice (Suppl. Figure 2).

Pharmacological mTORC1 inhibition reduces retinal AVM establishment and expansion in Eng iECKO mice

To assess the potential impact of mTORC1 activity on AVM establishment and expansion we administered the mTORC1 inhibitor rapamycin to control and EngiECKO mice 1- and 2-days post tamoxifen-induced recombination (see Fig. 3A for experimental outline). Rapamycin at the indicated dosage did not affect body weight, nor radial expansion of the retinal vasculature, but caused a slight reduction in the vascular area of control as well as of EngiECKO retinas (Suppl. Figure 3B, C, D). Importantly, Rapamycin reduced both AVM numbers and thickness in EngiECKO mice (Fig. 3B-E). Rapamycin efficiently inhibited mTORC1 signalling as confirmed by near complete absence of p-RPS6 in all retinal cells, including mural cells, microglia and ECs (Fig. 3F-H). The findings suggest that mTORC1 plays an important role in establishment and expansion of ENG-LOF-mediated AVM, aligning with previous reports on its effect in genetic models of HHT2 [11].

Endothelial mTORC1 activation is recorded only after AVM initialization

Next, we studied the state of mTORC1, inferred by immunoreactivity to p-RPS6, during vascular morphogenesis in control mice and during initialization and progression of AVM in retinas of EngiECKO mice. To this end, tamoxifen was administered at P3 to either control mice or EngiECKO mice, and retinal vasculatures were analysed at different time points thereafter (Fig. 4A, B). In control mice endothelial p-RPS6 remained low and unchanged from P4-P6, with persistent expression of ENG in ECs (Fig. 4C, D). In EngiECKO mice, one day post tamoxifen administration (P4), the vasculature appeared morphologically normal and ENG remained high in most ECs, with only a fraction of ECs displaying complete loss of the ENG protein, evidenced by immunostaining for ENG. Also, the vasculature showed no recordable increase in mTORC1 activity compared to controls (Fig. 4C, D). At P5, two days post induction, ENG levels were drastically reduced, and thin AVMs appeared in the retinal vasculature, but with unchanged phosphorylation of RPS6 (Fig. 4C, D). Three days post induction of Eng deletion (P6), larger AVMs were evident. In agreement with data presented above, AVM-ECs at this stage displayed increased mTORC1 activity compared to all other recorded conditions in control and EngiECKO retinas (Fig. 4D). These results propose that increased EC-mTORC1-mediated S6K activity and thereby phosphorylation of RPS6, at the level detected here, plays a limited role in the process of AVM initiation, downstream of ENG-deletion.

EC-specific inhibition of mTORC1 by Rptor deletion show only tendencies to reduced retinal AVM severity in Eng iECKO mice

To further study the contribution of EC-derived mTORC1 to AVM development we crossed EngiECKO mice with Rptorlxlx mice to generate EC-specific double knockouts (EngiECKO;RptoriECKO). Rptor encodes for regulatory associated protein of MTOR complex 1 (Raptor), a subunit of the mTORC1 complex required for its activity. Deletion of Rptor would hence result in decreased or ablated mTORC1 activity. EngiECKO;RptoriEChet (Cre+, heterozygote Rptor deletion) and EngiECKO;RptoriECKO mice induced by tamoxifen at P3, both showed retinal AVMs at P6 (Fig. 5A-B). However, EngiECKO;RptoriECKO mice (homozygote Rptor deletion) displayed a non-significant tendency to a milder AVM phenotype than EngiECKO;RptoriEChet mice, with respect to AVM incidence (number of retinas displaying at least one AVM), number of AVMs/retina and AVM thickness (Fig. 5B-D). Staining for p-RPS6 exposed reduced levels in the retinal vasculature of EngiECKO;RptoriECKO mice compared to those of EngiECKO;RptoriEChet mice, confirming successful EC-deletion of Rptor and decreased mTORC1 activity (Fig. 5E-F). However, there was no effect on p-RPS6 levels in the non-vascular compartment, indicative of EC restricted genetic manipulation (Fig. 5G). As cell size has been implicated in the biology of AVM, in particular downstream of endoglin LOF, we analysed number of ECs/subvascular area as a proxy for relative EC size. Immunostaining for the EC-restricted transcription factor ERG was used for EC counting in 3D (Suppl. Figure 4 A, B). Results did not expose any statistically significant differences in retinal EC density in arteries, veins, capillaries or sprouting front comparing control, EngiECKO;RptoriEChet and EngiECKO;RptoriECKO mice (Fig. 5H). Nevertheless, taking all groups into consideration, EngiECKO;RptoriEChet mice showed overall statistically lower vascular EC density than both control and EngiECKO;RptoriECKO, whereas there was no difference on this parameter between control and EngiECKO;RptoriECKO mice. These findings suggest that Rptor deletion partially normalises the ENG LOF-mediated increase in EC-size (Fig. 5H). EC-specific deletion of Rptor alone did not cause major patterning effects and did not affect EC density within the sprouting front region (Suppl. Figure 4 C-E).

These results suggest that EC-specific mTORC1 activity is not required for initiation and expansion of Eng-LOF mediated AVMs. Nevertheless, data indicate that the signalling cascade within ECs modestly promote AVM establishment and expansion in HHT1 mouse models. To what extent modulation of cell size, as observed by an increase following Eng deletion and partial normalisation by Rptor deletion, directly relates to the phenotype herein, remains to be investigated.

Forced EC-specific activation of mTORC1 through induced Tsc1 deletion reduces AVM incidence and thickness following BMP9/10 trapping

The biological relevance of the increased phosphorylation of RPS6 in ECs of HHT mouse models has not been investigated. To this end, we first generated mice (Cdh5(PAC)-CreERT2+/-:Tsc1flox/flox;R26REYFP, hence forth denoted Tsc1iECKO) allowing for induced EC-specific deletion of Tsc1, the endogenous inhibitor of mTORC1. Initiation of EC deletion of Tsc1 at P3 had not caused AVMs at P6 (Fig. 6A, B). EC-specific recombination and activation of mTORC1 were confirmed by YFP expression and increased immunoreactivity to p-RPS6 (Fig. 6C, D). To investigate the effect of mTORC1 overactivation on AVM development, we first tried to establish EngiECKO;Tsc1iECKO double knockouts but failed due to the close proximity of their respective loci in the genome. Instead, we applied a previously published model of HHT, based on administration of BMP9 and BMP10 blocking antibodies (bAbs), to provoke AVMs within the developing retinal vasculature [11, 12]. First, tamoxifen was administered at P1 to Cre negative Tsc1lxlx mice serving as controls, and to Tsc1iECKO mice to induce endothelial Tsc1 deletion (see Fig. 7A for administration scheme). Mice were then treated with either isotype IgG control Abs or a mix of BMP9 and BMP10 bAbs on P3, P4 and P5. Analysis at P6 revealed that all anti-BMP9/10 treated control mice had developed AVMs (Fig. 7B, C). AVMs displayed an increase of vascular p-RPS6 levels compared to IgG treated controls (Suppl. 5 A-C). However, Tsc1iECKO anti-BMP9/10 treated mice showed a milder AVM-phenotype, indicated by fewer AVMs/retina and reduced AVM thickness (Fig. 7C, D). Nevertheless, endothelial mTORC1 activity was elevated in anti-BMP9/10 treated Tsc1iECKO mice compared to anti-BMP9/10 treated control mice, inferred by higher immunoreactivity to p-RPS6 (Fig. 7E, F). Although severity of anti-BMP9/10-induced AVMs was reduced by TSC1 deletion, there was only a trend towards reduced non-vascular phosphorylation of RPS6 (Fig. 7G). Near complete tamoxifen-induced EC recombination was confirmed by YFP expression (Fig. 7E). These data show that forced activation of EC mTORC1 – beyond levels that naturally occur following interference with the BMP9-10/ACVRL1/ENG pathway – does not further potentiate the pathology, but instead reduces it.