We investigated the effect of disrupted VEGF signaling on pulmonary vascular disease in a preclinical model of direct ablative gene manipulation of VEGFR-2. We found that endothelial cell-specific knockout in mice leads to a mild PH phenotype that is aggravated by hypoxia. Moreover we found total vessel occlusion by intimal endothelial cell proliferation and lesions consistent with enMT that resembled the pulmonary arteriopathy of human pulmonary arterial hypertension. We further hypothesized that anti-angiogenic therapies in cancer patients might cause obstructive pulmonary vascular remodeling. Therefore, we studied plasma and lung specimens from patients treated with bevacizumab, a humanized monoclonal antibody directed against VEGF. Cardiovascular side-effects of bevacizumab include thromboembolic events [
22,
41], ischemic events [
10,
41], hypertension [
16,
28,
55], pulmonary embolism [
22,
39,
41] and pulmonary hypertension [
29]. The mechanism of these bevacizumab-related cardiovascular events is not fully understood. Adverse effects of VEGF inhibitors are largely consequences of blocking VEGF function in normal vascular physiology including vascular cell turnover and blood pressure regulation [
21]. Preclinical evidence has shown that VEGF blockade leads to endothelial cell apoptosis in most organ systems [
6]. Interestingly, this effect is reversible, resulting in vessel regrowth and normal vessel density after 1–2 weeks [
31]. Histologic evaluation of lung samples obtained from pulmonary metastasectomy of patients on bevacizumab treatment showed similar vascular alterations as seen in our rodent model. We observed increased media wall thickness, perivascular fibrosis and total vessel occlusions. We assume that intimal hyperplasia may be due to selection of abnormal apoptosis-resistant endothelial cells [
27,
37,
52]. Experimental proliferative pulmonary vasculopathy in a rat model was first described by Taraseviciene-Stewart who applied the VEGF receptor blocker SU5416 in combination with hypoxia [
48]. More recently Ciuclan could replicate this model in mice and also observed histological changes resembling those seen in human disease [
11]. Because Sugen systemically suppresses VEGFR-2 VEGFR-1, platelet-derived growth factor receptor, c-Kit (stem cell factor receptor) and RET (tyrosine kinase receptor) in all cell types and also causes emphysema [
25], we selectively disrupted only VEGFR-2 signaling in endothelial cells, to dissect this pathway in PH and to overcome the pleiotropic effects on different pulmonary cells including alveolar cells type 1 and 2 [
25,
36,
51]. Consequently and in contrast to the Sugen models, we did not observe severe emphysema after
Kdr knockout. Mean linear intercept as a surrogate for alveolar enlargement was not significantly different between
Kdr∆end and controls after Tamoxifen induction. Therefore, we conclude that emphysema as it was observed in Sugen rat models is unlikely to depend on endothelial cell death alone. As expected [
11,
48], we found that mice with disrupted VEGFR-2 signaling develop more extensive PH and RV hypertrophy than wild-type animals exposed to chronic hypoxia. In contrast to Ciuclan, but in line with Taraseviciene-Stewart, we also observed a mild PH phenotype after inhibition of VEGFR signaling without hypoxic exposure. Most importantly, we found proliferative vascular lesions expressing endothelial cell markers and VEGFR-3. There was no systemic response to
Kdr knockout and mean systemic arterial pressure did not change in any of the treatment groups [
11,
46]. However, because Ciuclan reported a left heart failure phenotype in mice following VEGFR blockade, we investigated the effect of
Kdr knockout on LV function utilizing transthoracic echocardiography and cardiac MRT. After hypoxic exposure we observed a significant decrease of CO in all experimental groups [
11], however, without further decrease by
Kdr knockout. We used MRT to assess RV function and found significantly decreased RV ejection fraction as a consequence of
Kdr knockout. Remarkably,
Kdr∆end mice show only modestly increased RV pressures at baseline while RV function was significantly impaired. We hypothesize that mechanisms other than increased RV afterload contribute to altered RV function. Bogaard has shown that isolated RV pressure overload by pulmonary artery banding leads to RV hypertrophy but not failure, whereas angioproliferative pulmonary hypertension results in both hypertrophy and RV failure. Authors hypothesized that structurally altered pulmonary circulation in PAH releases mediators that interfere with adaptive RV responses already maximally challenged to meet the increased mechanical stress [
9]. Therefore, we analyzed both the pulmonary circulation and the ventricles.
Kdr knockout leads to a loss of microvessels, more in the RV than in the LV, and in the lungs with decreased cross-sectional area of pulmonary vessels and subsequent increase in pulmonary arterial pressure [
19,
34]. We hypothesize that the loss of microvessels predominantely in the RV myocardium is the sequela of the combination of
Kdr blockade and the second ‘hit’ (chronic hypoxia and increased RV afterload) and therefore, LV myocardium remains virtually unaffected. Under hypoxia, major vessel obliterative pulmonary vascular lesions are observed in
Kdr knockout mice that resemble intimal proliferative lesions of severe human PAH [
18,
54]. To understand mechanisms of pulmonary vascular remodeling after
Kdr knockout we examined the impact on apoptosis and proliferation of pulmonary vascular cells. Early after
Kdr knockout we observed a small but significant increase in caspase 3-positive cells that was followed by a similar significant increase in PCNA-positive cells under hypoxia. Furthermore, we observed that the angioproliferative lesions in
KdrΔend mice expressed PCNA, suggesting a proliferative phenotype.
Kdr knockout was associated with a robust pulmonary vascular inflammatory response with accumulation of inflammatory cells in arterioles of
KdrΔend mice. Because perivascular inflammatory infiltrates precede vascular remodeling in the development of PAH [
40], a misguided inflammatory response to vascular injury might contribute to the development of pulmonary vasculopathy [
40,
47]. However, this cellular infiltrate might also be a response to the initial vascular apoptotic processes that are superseded by angioproliferative responses. Therefore, we investigated mRNA levels of C1q, a protein that is crucial for phagocytic removal of apoptotic cells (efferocytosis). We found
C1q mRNA to be significantly upregulated after
Kdr knockout, which may be a signal for efferocytosis deficiency. Because VEGFR-2 has been shown to be important for macrophage–mediated efferocytosis, efferocytosis deficiency might also drive the vasculopathy observed in the present model [
23,
24,
53]. We hypothesize that once efferocytosis is impaired as a consequence of
Kdr knock-out, apoptotic cells persist and trigger inflammation and autoimmunity, leading to vascular occlusion and pulmonary hypertension [
53].
In contrast to Ciuclan we found
Kdr knockout to directly affect BMP signaling. We found both
Bmp2 and
Bmpr2 downregulated after the knockout. Although a direct relationship of VEGF and BMP signaling pathways has not been reported, their interaction seems likely. Reduced expression of
Bmp2 and
Bmpr2 suggests that both pathways act in parallel and underlines the proliferative nature of the disease resulting in a loss of patent pulmonary microvasculature, and eventually, in a loss of endothelial markers. We could identify elevated VEGFa levels as consequence of VEGFR-2 knockout or bevacizumab therapy. These findings may be central to the pathogenesis of pulmonary vasculopathy. If pulmonary hypertension is dependent on multiple injuries or “hits” [
51,
52], one may speculate that we caused an initial hit via VEGF blockade and selected apoptosis resistant cells which then proliferated secondary to VEGF blockade [
51]. In those proliferating cells we found a sustained upregulation of VEGFR-3, which might in part account for the pro-proliferative phenotype. VEGFR-3 shares structural similarities to VEGFR-2 and is capable to bind all members of VEGF ligands (preferentially VEGF-C and VEGF-D), promoting angiogenesis and lymphangiogenesis [
3]. Because VEGFR-3 is more subjected to regulation by Notch than VEGFR-2, it may be able to rescue neoangiogenesis once VEGFR-2 is blocked [
8]. We hypothesize that VEGFR-3 overexpression serves as a mechanistic explanation for the proliferative vasculopathy seen in the present model, which underpins the 2-hits-theory [
51]. Thus, these data are consistent with the hypothesis that sustained VEGFR-2 inhibition in endothelial cells activates a stem cell–related cell proliferation mechanism that includes VEGFR-3 protein expression [
1,
12]. Furthermore, we observed similar VEGFR-3-positive lesions in all cancer patients treated with bevacizumab. Limitations of our work are the lack of a control group for the human studies, the lack transthoracic echocardiograms and the lack of serum samples of metastasectomy patients. Not all proliferating ECs were positive for CD31 (Fig.
3g), and we could not prove monoclonal growth. Our findings in patients after bevacizumab therapy support the concept that VEGF inhibition leads to hyperproliferative endothelial cells that occlude the pulmonary vascular lumen, an observation that has been labeled as “the angiogenesis paradox in pulmonary arterial hypertension” [
51]. Later, these lumenless vessels seem to disappear; however, we have no information on the mechanisms underlying the lack of EC markers in the small vessel compartments of lung and heart. Presumably, vascular changes in patients are not uniform over both lungs, but focally distributed, leading to segmental PH.
We propose that interrupted VEGF signaling leads to a pulmonary arteriopathy in rodents. In humans receiving anti-VEGF treatment, a similar mechanism may be effective. Our findings illustrate the importance of intact VEGF signaling for the maintenance of pulmonary vascular patency.