Phosphorylation of pericyte FAK-Y861 affects tumour cell apoptosis and tumour blood vessel regression

Angiogenesis is a complex biological process that involves multiple cell types, crosstalk between these cells and responses to different stimuli [1‐3]. Interactions between pericytes and endothelial cells play important roles in angiogenesis. Pericyte recruitment to newly forming blood vessels is thought to promote neovessel stabilisation and is an essential step in neovascularisation [4, 5]. Failure to recruit pericytes to blood vessels can affect vascular remodelling, vessel regression and vessel leakage [1, 2, 6]. This has led to pericyte-targeted therapies being developed and a better understanding of the role of pericytes in tumour angiogenesis [7, 8].

It is well documented that focal adhesion kinase (FAK) plays a central role in different aspects of tumour growth and is overexpressed in many types of cancer [9‐11]. One of the many roles of endothelial cell FAK, in the promotion of tumour growth, is in the initiation of angiogenesis [12]. FAK regulates growth, survival, migration and invasion through its dual role as a kinase and as a scaffolding protein. FAK kinase activity results in tyrosine (Y) 397-phosphorylation, which, in turn, allows proteins containing the Src-homology (SH2) domain to bind to FAK, e.g., Src and PI3K. FAK-Src complexing is thought to potentiate further FAK phosphorylation at other FAK domains including FAK-Y861 [13, 14]. Most studies have pointed towards essential roles for phosphorylation of FAK-Y397 but much less in known about the requirement of FAK-Y861. We and others have shown previously that endothelial cell (EC) FAK is required for tumour growth since EC FAK loss leads to a reduction in tumour growth, accompanied by a reduction in tumour vascular density [12]. Endothelial specific loss of FAK inhibits brain tumour formation and leads to normalisation of the vasculature [15]. Global constitutive deletion of exon 5, which encodes FAK-Y397, also leads to an embryonic lethal phenotype with vascular permeability defects [16]. More recently, we have shown that non-phosphorylatable mutations of tyrosines 397 and 861 (Y397 and Y861) in endothelial cells have differential effects on tumour angiogenesis [17] and that endothelial cell FAK regulates angiocrine signalling in the control of doxorubicin sensitivity in malignant cells [18]. Whilst FAK has been studied extensively in endothelial cells, the function of FAK in pericytes during tumour growth and angiogenesis is starting to emerge, with loss of pericyte FAK enhancing tumour angiogenesis [19]. Given that pericytes are an important cell type in the regulation of tumour growth and paracrine signalling [20‐22], here we have examined the role of FAK point mutations in pericytes during pathological angiogenesis. Specifically, we have examined the effect of non-phosphorylatable mutations of tyrosines 397 and 861 (Y397F and Y861F) – FAK residues using pericyte-specific FAK^Y397F/Y397F and FAK^Y861F/Y861F mice.

In this study, we show that in a subcutaneous Lewis Lung Carcinoma (LLC) tumour model, both angiogenesis and tumour growth are reduced only in mice which have the pericyte-specific FAK-Y861F mutation. This correlates with a significant increase in vessel regression. Furthermore, examination of the secretome and protein expression in FAK-Y861F pericytes highlight an altered pericyte signature involving cytokines and protein secretion that are involved in cancer cell apoptosis. Thus, pericyte FAK-Y861 plays a role in the control of tumour growth.

To examine the effects of pericyte FAK-Y397F and Y861F mutations on tumour growth and angiogenesis, PdgfrβCre + ;FAK^WT/WT control mice, PdgfrβCre + ;FAK^Y861F/Y861F and PdgfrβCre + ;FAK^Y397F/Y397F mice were injected subcutaneously with Lewis Lung Carcinoma cells (LLC). In vivo tumour growth was reduced in PdgfrβCre + ;FAK^Y861F/Y861F, but not PdgfrβCre + ;FAK^Y397F/Y397F mice. Furthermore, the PdgfrβCre + ;FAK^Y861F/Y861F but not PdgfrβCre + ;FAK^Y397F/Y397F mice had significantly reduced blood vessel density (as determined by the number of endomucin‐positive vessels per mm² of age‐matched, size‐matched tumours) compared with PdgfrβCre + ;FAK^WT/WT control mice (Fig. 1g, h).

In addition to the above, blood vessel perfusion (determined by the percentage of endomucin-positive blood vessels that were also positive for PE-PECAM antibody after ante-mortem perfusion) and pericyte coverage (determined by the percentage of vessels with NG2-positive mural cell association) were both reduced in PdgfrβCre + ;FAK^Y861F/Y861F but not PdgfrβCre + ;FAK^Y397F/Y397F mice (Fig. 2 a, b). Loss of blood vessels can result from partial or persistent regression of vascular endothelial cells. Blood vessel regression is associated with endothelial cell loss and pericyte dropout leaving behind empty collagen IV basement membrane sleeves [23, 24]. Thus, vessel regression can be determined by the presence of endothelial cell-negative, collagen IV-positive basement membrane blood vessel sleeves. Collagen IV / endomucin double-immunostaining of day 14–21 tumours (when the rate of tumour growth in PdgfrβCre + ;FAK^Y861F/Y861F mice slows down) from PdgfrβCre + ;FAK^WT/WT, PdgfrβCre + ;FAK^Y397F/Y397F and PdgfrβCre + ;FAK^Y861F/Y861F mice showed a significant increase in endomucin-negative / collagen IV-positive blood vessel sleeves in tumours from PdgfrβCre + ;FAK^Y861F/Y861F compared with tumours from PdgfrβCre + ;FAK^WT/WT, PdgfrβCre + ;FAK^Y397F/Y397F mice (Fig. 2c). These results correlate with the reduction in tumour growth and suggest that tumour blood vessel regression may be involved in the tumour growth phenotype in PdgfrβCre + ;FAK^Y861F/Y861F mice.

Examination of early stage tumours was undertaken to determine at what stage, during growth of the tumour, angiogenesis was being affected. In early stage tumours (day 12 post tumour cell inoculation) LLC tumour size, blood vessel density and blood vessel perfusion were similar between PdgfrβCre + ;FAK^WT/WT and PdgfrβCre + ;FAK^Y861F/Y861F mice (Fig. 3a-c), suggesting that loss of blood vessels occur at a later stage of tumour growth. Indeed to examine whether FAK-Y861F PCs could directly affect the initial stages of microvessel sprouting, aortic rings from PdgfrβCre + ;FAK^Y861F/Y861F and PdgfrβCre + ;FAK^WT/WT mice were embedded in collagen and stimulated, or not, with VEGF (30 ng/ml). VEGF treatment significantly increased angiogenic sprouting in both genotypes to the same extent suggesting that mutation of FAK-Y861F in pericytes is not sufficient to directly affect angiogenic sprouting in a tumour-free environment (Supplementary Fig. 2).

Examination of early stage tumours, using H&E stained sections, revealed that tumours from PdgfrβCre + ;FAK^Y861F/Y861F mice had a significant increase in tumour necrosis compared with PdgfrβCre + ;FAK^WT/WT control mice (Fig. 3d). These results suggest that tumour necrosis precedes the onset of vessel regression and the reduced late stage tumour blood vessel density and tumour growth in PdgfrβCre + ;FAK^Y861F/Y861F mice.

Paracrine or juxtacrine signals from endothelial cells, angiocrine signals, have been implicated in the control of tumour growth where endothelial cell derived signals directly affect tumour cells by altering secreted cytokine profiles [18, 25, 26]. The role of pericytes in this type of signalling is also emerging [27, 28]. Thus, we turned our attention to the possible effects of FAK-Y861F pericytes on tumour cell apoptosis. Using R&D protein profiler arrays, lysates from FAK-Y861F pericytes showed a decrease in levels of thrombospondin, MCP-1, proliferin, TIMP-1 and sICAM/CD54 together with an increase in IGFBP-2, endostatin, TNF-alpha, SDF1 and ADAMTS-1 compared with lysates from control WT pericytes. Similar effects on these cytokines were not observed in FAK-Y397F pericytes compared with WT pericytes and SDF-1/CXCL12 was apparently reduced (Fig. 4a). We focused on the role of MCP-1 in the apoptotic phenotype in tumours from PdgfrβCre + ;FAK^Y861F/Y861F mice since, in vitro, MCP-1 promotes mural cell recruitment [29] and in vivo pharmacological inhibition of MCP-1 reduces tumour growth and macrophage recruitment resulting in increased tumour necrosis [30], whilst loss of MCP-1 delays mammary tumourigenesis [31]. Therefore, we treated LLC with MCP-1, after exposure to Y861F conditioned medium (CM), and asked if MCP-1 could be involved in controlling tumour cell apoptosis. CM from FAK-WT and FAK-Y861F pericytes were incubated with cultured LLC cells and apoptosis quantified by TUNEL staining. Y861F CM caused a significant increase in LLC apoptosis compared with LLCs incubated with WT CM. Indeed the pro-apoptotic phenotype of Y861F CM was reduced by the addition of exogenous recombinant MCP-1 to the pericyte CM (Fig. 4b), suggesting that the reduction of MCP-1 in Y861F pericytes is at least partially involved in the pro-apoptotic phenotype in tumours grown in PdgfrβCre + ;FAK^Y861F/Y861F (Y861F) mice. This pro-apoptotic tumour cell phenotype was confirmed, in vivo, where tumours from PdgfrβCre + ;FAK^Y861F/Y861F mice displayed significantly increased levels of apoptosis compared with tumours from either PdgfrβCre + ;FAK^WT/WT or PdgfrβCre + ;FAK^Y397F/Y397F mice (Fig. 4c).

FAK kinase activity leads to phosphorylation of Y397-FAK which, in turn, allows protein containing the Src-homology (SH2) domain to bind to FAK, including Src. FAK-Src complexing are thought to potentiate further FAK phosphorylation at other FAK domains including Y861. Since phosphorylated Src was not reduced in Y861F pericytes, we wanted to determine if the phenotype observed in Y861F pericytes was dependent on Src. To address this, Src levels were depleted in FAK-WT, FAK-Y397F and FAK-Y861F pericytes followed by examination of the effect of CM from these cells on LLC cell death. CM from Src-depleted FAK-WT and FAK-Y397F pericytes significantly increased LLC apoptosis. Conversely, CM from FAK-Y861 Src-depleted pericytes (Supplementary Fig. 3) had no additional effect on LLC apoptosis above LLC exposed to non-treated CM (Fig. 4d).

These results imply a direct regulation of tumour cells through pericyte paracrine signalling—a direct effect on vascular disruption and ultimately tumour growth that is regulated by FAK-Y861 phosphorylation, and independently of Src.

Our data demonstrate that FAK-Y861F pericytes decrease tumour burden, at least in part, by directly affecting tumour cell apoptosis, and call for further consideration of the role of tumour pericytes in the direct control of tumour growth, in addition to their effects on vessel stabilisation. Indeed, we have recently shown that pericytes play a central role in the control of primary tumour growth, through cross-talk with multiple cell types in the tumour microenvironment [19, 20].

FAK is upregulated in many cancer types and is currently being targeted as a potential anti-cancer agent because of its essential roles in tumour growth and angiogenesis [12, 32‐34]. Our previous work identified distinct roles for FAK-Y397 and FAK-Y861 phosphorylation in tumour endothelial cells where endothelial cell FAK-Y397 but not FAK-Y861 reduced tumour growth and angiogenesis, and linking these tyrosine domains to specific signaling pathways downstream of FAK and regulation of an inside out signal affecting endothelial cells surface receptor expression [17, 35]. Pericytes are thought to be essential in stabilising the tumour vasculature and thus attention has been given to the potential of targeting pericytes to induce vascular disruption in cancer control [36]. However, recent conflicting studies have emerged challenging this idea [37] and suggest possible additional roles for pericytes in regulating tumour growth. In contrast to the effect of FAK-Y397F mutation in endothelial cells [17], our mouse models show that phosphorylation of FAK-Y397 in pericytes is not a prerequisite for the control of LLC tumour growth or angiogenesis. The reason for this apparent discrepancy likely reflect a different requirement for FAK-Y397-phosphorylation in these two cells types in tumour blood vessels. Additionally, complete loss of pericyte FAK is compensated by an increase in pericyte p-Pyk2, which is sufficient to drive a downstream signaling cascade to promote tumour growth, through effects on both malignant cells and angiogenesis [19]. Importantly, one cannot directly compare pericyte FAK-KO with pericyte FAK- Y397F, since phosphorylation of Src is not affected in pericyte FAK-KO mice, whereas in FAK-Y397F mice, the kinase domain is not affected (as it acts upstream of Y397 in the signaling cascade of FAK activation).

FAK-Y861F mutation in pericytes is associated with reduced tumour growth and angiogenesis that is preceded by an increased tendency for tumour cell necrosis and subsequent blood vessel regression. Since tumour necrosis has been demonstrated to have vascular disrupting effects [38, 39] it is tempting to speculate that the early stage enhanced tumour necrosis may initiate vascular disruption. Additionally, given that the Y861F mutation reduces pericyte association with tumour blood vessels, it is likely that vessel stability and maturation is compromised and this reduced pericyte coverage is also likely to be responsible for the decrease in the numbers of functional blood vessels and reduced tumour growth.

Studies have shown that vascular endothelial cells can control tumour growth via angiocrine signals, including cytokines and chemokines [18, 40]. In our study, the protein signature of FAK-Y861F pericytes is a combination of reduced levels of TSP-2, MCP-1, proliferin, TIMP-1 and sICAM/CD54 together with an increase in IGFBP-2, endostatin, TNF-alpha, SDF1 and ADAMTS-1. This molecular signature associates with pro-apoptotic effects on tumour cells but apparently has little effect on VEGF-stimulated angiogenic sprouting in aortic ring assays in a tumour-free context. The lack of an effect on VEGF-induced aortic ring sprouting suggests that the effect of FAK-Y861F pericyte enhancement on tumour cell apoptosis may be a contributing factor to subsequent vessel regression in a tumour environment. The anti-angiogenic effect of TSP-2 is blocked by VEGF thus providing a possible explanation of why VEGF-stimulated angiogenesis is not affected in aortic ring assays from PdgfrβCre + ;FAK^Y861F/Y861F mice where the decrease in TSP-2 might otherwise have increased angiogenesis [41]. The pro-tumoural effects of MCP-1, proliferin, TIMP-1 and soluble ICAM-1 [42‐45] whilst the anti-cancer effects of IGFBP-2, endostatin and ADAMTS-1 [46‐48] have been published and suggest that this secretome signature may well provide a combination of effects on cancer growth. Indeed MCP-1 addition to conditioned medium from FAK-Y861F pericytes was able to rescue the pro-apoptotic effect of this cocktail on tumour cells.

Furthermore, we have shown here that loss of pericyte FAK-Y861 phosphorylation intrinsically regulates a paracrine signal that controls apoptosis of tumour cells, independent of Src and does not affect phosphorylation of Y416-Src. We and others have previously shown that loss of FAK, or a non-phosphorylatable Y397F mutation in endothelial cells, disrupts active Src-FAK interactions, driving Src degradation, possibly activating other FAK-Src independent signalling pathways [17, 49‐51].

The molecular mechanism by which FAK-Y861 affects this secretome signature is the subject of the future study. Our data support the notion that pericytes are not simply vessel supporting cells, but that via FAK-Y861, can regulate tumour growth via pericyte derived signals directly affecting tumour cell apoptosis.