Microcirculation surrounding end-stage human chronic skin wounds is associated with endoglin/CD146/ALK-1 expression, endothelial cell proliferation and an absence of p16

Angiogenesis is an essential part of normal skin healing, re-establishing blood flow in developing granulation tissue. Non-healing skin wounds are associated with impaired angiogenesis and although the role of re-establishing macroscopic blood flow to limbs to prevent wound chronicity is well investigated, less is known about vascular alter-ations at the microcirculatory level. We hypothesised that significant phenotypic changes would be evident in blood vessels surrounding chronic skin wounds. Wound edge tissue, proximal to wound (2 cm from

in sub-populations of keratinocytes. We conclude that the endoglin-ALK-1-endothelial proliferation axis is active in the vasculature at the edge of chronic skin wounds and is not associated with p16 Ink4a mediated senescence. This information could be further used to guide treatment of chronic skin wounds and optimise debridement protocols.

| INTRODUCTION
Skin wound chronicity remain a significant clinical problem associated with diabetes, vascular insufficiency and ageing. 1 Classified clinically as pressure, vascular and diabetic ulcers, the associated impaired skin healing has a major impact on patient quality of life and mortality. 2 The normal skin healing process in humans is associated with a robust neovascularisation response, which re-establishes blood flow to the granulation tissue forming in the wound bed, facilitating granulation tissue formation and maturation, re-epithelialisation and ultimately restoration of barrier function. 3,4 The initial response of skin to injury involves rapid and robust infiltration of the developing granulation tissue by capillaries, with total vessel number being up to 10 times higher than in quiescent healthy skin. [4][5][6] However, many of these capillaries are never perfused, 7 with regression of sprouting capillaries occurring temporally with healing time, with some studies positively correlating increased capillary density and stability with scar formation. 8,9 In contrast to over-healing situations, inadequate or dysfunctional angiogenesis is considered a major contributor to development of skin wound chronicity, where a lack of nutrient and oxygen supply leads to disruption of the normal healing response. [10][11][12] Many factors are considered to result in both macro-and microvascular dysfunction that can manifest in the formation of chronic skin wounds. [12][13][14][15][16] Peripheral arterial disease due to arteriosclerosis is common in patients with impaired skin healing in lower extremities, 17,18 particularly in those with diabetes. 19,20 Irrespective of aetiology, angiogenesis within the granulation tissue of chronic wounds is aberrant or absent which can result in localised hypoxia. 21,22 While macrovascular examination and intervention is standard of care in patients with chronic skin wounds, assessment and quantification of the microvasculature in the lower extremities remains problematic primarily due to the technologies available to clinicians. 22 Healthy micro-vasculature and angiogenesis are associated with the expression of certain molecules, although their presence in chronic wound tissue has not been well validated. Platelet/endothelial cell adhesion molecule-1 (PECAM-1, CD31) is normally expressed on endothelial cells, where it maintains the integrity of endothelial cellcell junctions and provides protection in the presence of apoptotic stimuli. 23 CD146 is expressed at all levels of the vascular system, including in endothelial cells, smooth muscle and pericytes. 24 CD146 is implicated in maintenance of vessel homeostasis, endothelial cellpericyte interactions 25 and critically, in angiogenesis, defined largely through cancer models. 26 CD105, also termed endoglin, is a type I transmembrane glycoprotein which acts as an ancillary receptor for transforming growth factor superfamily receptor-mediated signalling. [27][28][29] Endoglin, like CD146, is considered a reliable marker of angiogenesis 30 ; association of endoglin with activin receptor-like kinase 1 (ALK-1) induces endothelial proliferation in response to transforming growth factor β (TGF-β). 28 Based on their defined roles in vascular health and angiogenesis, by quantifying the localisation of CD31, CD146 and endoglin in chronic skin wound tissue, an assessment of endothelial health and angiogenic signalling could be derived.
By analysing the activation of signalling molecules and localisation of capillaries, arterioles and venules in tissue surrounding the wound area, it could have importance for debridement protocols, wound management and design of more appropriate treatment strategies.
In previous studies, we noted a qualitative increase in vascular density in the dermal tissue on the edge of human chronic wounds compared to non-involved tissue. 31  Twenty patients were enrolled with demographics shown in Table 1.
Vascular status of the affected limbs is shown in Table 2. Skin samples were collected from the wound edge, 2 cm proximal to the wound as well as from a non-involved region of the limb at least 10 cm away ( Figure S1). All samples were immediately immersed in 4% paraformaldehyde (Sigma Aldrich, St. Louis, Missouri) upon extraction from the limb.

| Microvasculature quantification
Nineteen out of 20 human wounds were assessed for the presence of microvasculature based on the different molecular markers, CD31, endoglin, and CD146. Blood vessel quantification was performed in an unblinded manner using ImageJ software according to previous studies. 33,34 Briefly, blood vessels were manually highlighted in each section and thresholded for measuring the size and number of the vessels in a box of 1 mm 2 . The data were then imported into Prism (Graphpad Software, La Jolla, California) and three comparisons made: total % area of vessels per mm 2 of tissue, average size of each vessel and number of vessels per mm 2 were then calculated in wound edge, proximal and non-involved tissue in each patient.

| Co-localisation and quantification of ALK signalling with endoglin
Dual immunofluorescence images were opened in ImageJ and overlayed with a 1 mm by 1 mm grid. Five grid squares each were selected from under the epithelial tongue, proximal to the wound on the same slide, and non-involved tissue for analysis. The colour channels were separated with set thresholds, Endoglin-positive vessel structures were outlined, and outlines were overlayed on the red channel image to assess co-localisation with ALK-1 and ALK-5.

| Statistical methods
Statistical analysis was by Graphpad Software v8 (p ≤ 0.05 was considered significant). Normality was assessed with the Kolmogorov-Smirnov test and unless stated otherwise, data were found to be normally distributed. Comparisons were made between wound edge, proximal and noninvolved using one-way ANOVA with post hoc testing performed using

| CD31 expression is maintained in the vasculature up to the edge of the wound
Based on its importance in maintaining endothelial homeostasis and vascular health, 23 we first assessed the immunoreactivity for CD31 in noninvolved, proximal and wound edge tissue ( Figure 1). Strong immunoreactivity was evident in non-involved and tissue proximal to the wound in capillaries, venules and arterioles ( Figure 1A B) with a similar pattern seen in tissue proximal to the wound. In wound edge tissue, CD31 localisation to endothelial cells was evident, although vessels shape was more irregular ( Figure 1B). Cluster of CD31 + blood vessels were observed up to the end of the epithelial front, with some blood vessels evident in the wound bed.
No significant differences were seen in CD31 + blood vessel area between non-involved, proximal and wound edge ( Figure 1C), with average vessel size ( Figure 1D) and density of vessels ( Figure 1E) also similar.
T A B L E 3 List of antibodies used in the study. 3.3 | CD146 + and endoglin + blood vessels are increased at the edge of human chronic skin wounds To further assess the angiogenic phenotype of the vasculature surrounding chronic wounds, we labelled tissue with antibodies to CD146, which is known to be involved in angiogenesis and is expressed by endothelial cells, smooth muscle and pericytes 24 ( Figure 2). In non-involved tissue, immunoreactivity was observed in all blood vessels, localising mainly to the endothelial layer ( Figure 2A).
In wound edge tissue, vessel number was increased, with CD146 + present in all vessel irrespective of their size ( Figure 2B). In many CD146 + vessels, no lumen could be distinguished, although they were still maintained in larger vessels. Quantification of CD146 + vessel area showed a significant increase from non-involved to proximal tissue F I G U R E 1 CD31 expression in the endothelium of the dermal microcirculation is similar in non-involved and wound edge tissues. In (A) noninvolved uninjured skin and (B) skin proximal to the wound, CD31 localises to the endothelial layer in vessels irrespective of their size, but it particularly evident in capillaries. In (C) wound edge dermis, although present, CD31 expression was associated with vessels with poor organisation and fewer capillaries were evident. Analysis of (D) percentage area of vessels, (E) average vessel size, and (F) microvasculature density CD31 + associated vessels showed no significant differences between any tissue region investigated ( p > 0.05, One-way ANOVA, with Tukey's multiple comparisons test). Images in (A-C) are from AP5.
F I G U R E 2 CD146 expression in the dermal microcirculation increases significantly in proximal and wound edge tissue compared to noninvolved. In non-involved dermis (A), high immunoreactivity was evident in the endothelium and mural cells surrounding capillaries. In (B) proximal and (C) wound edge tissue, the density of vessels was qualitatively higher, although lumens were hard to discern in many of the vessels.
(D) Percentage area of vessels was increased in both proximal and wound edge versus non-involved, (E) average vessel size was similar in all tissue areas, and (F) microvasculature density was significantly increased in wound edge versus both proximal and non-involved. In all cases, p < 0.05, One-way ANOVA, with Tukey's multiple comparisons test. Images in (A-C) are from AP14.
F I G U R E 3 Endoglin microvascular density increases in wound edge tissue compared to non-involved dermis. Although expressed in noninvolved tissue (A) and tissue proximal to the wound (B), the overall density of endoglin + vessels remained low compared to wound edge tissue (C). In the lower reticular dermis in non-involved dermis, the immunoreactivity was also lower than in the papillary region. In wound edge tissue, the density of endoglin + vessels increased, but the vessel structure is unorganised and in many vessels lumens are not evident. Quantification of (D) percentage area of endoglin + vessels showed significantly more vessels in both proximal and wound edge tissue, with average vessel size only increased in wound edge tissue (E). Microvasculature density of endoglin + associated vessels (F) was significantly higher in both proximal and wound edge versus non-involved dermis. In all cases, p < 0.05, One-way ANOVA, with Tukey's multiple comparisons test. Images in (A-C) are from AP12.

| Distinct profile of blood vessel phenotype exists surrounding the wound bed
In non-involved tissue, the average size of CD31 + blood vessels was larger than those associated with endoglin or CD146 ( Figure 5A) (p < 0.05, one-way ANOVA, with Tukey's multiple comparisons test).
However, the number of CD146 + vessels was higher than those F I G U R E 4 Endoglin + and CD146 + microcirculation is increased surrounding chronic wound edge dermis. Using dual immunofluorescence with antibodies specific for endoglin and CD146, we observed in non-involved tissue co-localisation for endoglin and CD146 as well as vessels that only labelled for one of the markers (white arrows) or none of the markers (white arrowheads). In wound edge dermis, endoglin consistently labelled the endothelium of vessels, with CD146 labelling both the endothelium and mural cells surrounding the vessels in both the reticular and papillary dermis. Images in from AP15.
positive for endoglin or CD31 ( Figure 5A) (p < 0.05, one-way ANOVA, with Tukey's multiple comparisons test). In proximal tissue, the average size of CD31 + vessels was only larger than endoglin + vessels

| Co-localisation of endoglin with ALK-1 and ALK-5
Endoglin is expressed by activated endothelial cells acting as a coreceptor for the TGF-β superfamily and is essential for angiogenesis. 36 ALK-1, which is an endothelial specific TGF-β type I receptor, interacts with endoglin to promote endothelial proliferation, but in contrast when associated with ALK-5, endoglin suppresses proliferation. 28,37 Tissue F I G U R E 5 Distinct patterns of endoglin, CD31 and CD146 exist in the microcirculation in wound edge dermis. The localisation of each marker was compared in each tissue region (A) non-involved, (B) proximal and (C) wound edge with respect to average size of blood vessels and the total number of blood vessels. In non-involved tissue, CD31 + were largest, but at the wound edge this difference was attenuated. In total number of vessels per mm 2 , CD31 + vessels were significantly reduced in number in both proximal and wound edge tissue compared to endoglin and CD146 positive vessels. In all cases, p < 0.05, One-way ANOVA, with Tukey's multiple comparisons test. 3.6 | Endoglin-ALK1 signalling axis and endothelial proliferation is maintained at the edge of human chronic skin wounds As association of endoglin with ALK-1 is upstream of endothelial cell proliferation, we next investigated in whether endoglin-ALK-1 + vasculature was associated with proliferation. We first confirmed colocalisation of endoglin and ALK-1 in the vasculature of wound edge tissue ( Figure 8A,B) and assessed proliferation using antibodies to PCNA in serial sections ( Figure 8C,D). PCNA immunoreactivity was evident throughout the wound edge tissue, particularly in the nuclei of epithelial cells up to the end of the epithelial tongue ( Figure 8C). PCNA + nuclei were evident in the endothelial layers of the majority of blood vessels, independent of their size ( Figure 8D). Co-labelling of endoglin/ALK-5 and endoglin/ALK-1 and its relation to PCNA labelling in wound edge tissue from additional individuals is shown in Figure S3.

| Chronic skin wound edge micro-vasculature is not associated with p16 INK4A labelling of the endothelium
As the endoglin-ALK1 signalling axis, endothelial proliferation and increased microvascular density was evident at the wound edge, we next assessed whether cellular senescence was evident in any of the wound edge tissue using antibodies specific to p16 INK4A (Figure 9). In all patients, immunoreactivity for p16 INK4A was localised predominantly to keratinocytes around the wound edge. Basal keratinocytes and cells in the stratum spinosum showed sporadic p16 INK4A labelling at the wound edge, although no obvious pattern was discernable in any tissue from any individual (Figure 9).
F I G U R E 7 ALK-1 and ALK-5 show distinct co-localisation patterns in endoglin + microvasculature. To quantify the overlap of endoglin+ microvessels with ALK-1 or ALK-5, sections were dual labelled with antibodies and co-localisation performed using ImageJ. Quantification of endoglin positive vessels demonstrated a similar pattern as shown in Figure 3A,B. Co-localisation of endoglin with ALK-1 showed an almost 100% overlap in signals (C), but ALK-5 overlap was more variable (D). In (A, B), p < 0.05, One-way ANOVA, with Tukey's multiple comparisons test. In (C, D), p < 0.05 Spearman's rank correlation coefficient test.
F I G U R E 8 Co-localisation of endoglin and ALK-1 is associated with endothelial proliferation in the microvasculature. Co-localisation of endoglin and ALK-1 was confirmed in the vasculature (A, B), with PCNA labelling assessed in serial sections (C, D). Endothelial proliferation was evident in the endothelium of all microvasculature extending back from the wound edge tissue. Images are from AP4.

| DISCUSSION
Chronic skin wounds remain a major co-morbidity associated with vascular insufficiency and diabetes, with a significant socio-economic burden placed on the healthcare systems. 2 Understanding wound chronicity at the cell and molecular level is an important and necessary step towards identifying specific strategies to aid in wound closure as well as identifying markers associated with irreversible end stage disease. In this study, we specifically utilised a histological approach to assess the overall structure of the tissue and vasculature in the skin surrounding chronic wounds as well as within the wound bed itself. By understanding the structure and localisation of the microcirculation in the dermis, wound edge and wound bed, it provides important information that cannot be assessed through mRNA quantification, Western blotting or cell culture studies in which tissue structure is methodologically lost. By reconstructing large sections of the tissues, we were able to analyse how the viability and architecture of the tissue changes approaching the non-healing region of the skin, which has potential significance for guiding debridement that remains a largely wound and patient dependent methodology.
We show for the first time that a significant increase in the density of blood vessels in the microcirculation is evident in wound edge tissue. Increases in the endogenous angiogenic response have been previously described in muscle samples isolated from patients with critical limb ischemia, 38,39 with increased hypoxia-inducible factor-1 expression present, 38 although significant abnormalities in developing vessel structure were described. 39 A common abnormality associated with the pathological changes in the microcirculation of muscle is thickening of the basement membrane between the mural cells and the endothelium, which has been described generally in the F I G U R E 9 p16 INK4A is not associated with the microvasculature on the edge of chronic skin wounds but localises to keratinocytes in the overlying epidermis. Tissue was labelled with antibodies specific for p16 INK4A in tissue from multiple individuals. p16 INK4A was primarily associated with keratinocytes in the overlying epithelium and was largely absent in the microcirculation and mesenchymal populations in the dermis. Inset shows positive control for p16 INK4A , which in this instance are oropharyngeal cancer specimens.
microvasculature of patients with diabetes including in the dermis. 14 Moreover, pericyte degeneration and the presence of acellular capillaries emphasises the potential role of microvascular dysfunction upstream of tissue damage in diabetic patients. 40 Interestingly, we show here that irrespective of background co-morbidities such as diabetes, a similar microvascular angiogenic response was evident in all wound edge tissue isolated from all patients in the study, with consistent mural cell coverage of endothelium. However, a clear line of demarcation could be seen beyond which microvasculature sprouting into the wound bed was absent, suggesting factors that inhibit endothelial sprouting. Of significance, it is known in murine skin healing, that although vascular proliferation occurs, many of these vessels are never perfused. 7 Although CD31+ vessels increase in number post wounding, in particular the vessels on the wound margin and superficial granulation tissue are not perfused first. In normal healing, the assembly of the developing microvasculature follows a welldocumented series of events, with recruitment of pericytes and mural cell coverage a defining moment; once recruited, stabilisation of the endothelial plexus occurs which prevents further fine tuning and remodelling of the vessels in the microcirculation. 41 Regardless of aetiology, it has been well characterised that angiogenesis within the granulation tissue of chronic wounds is reduced, 21,22 although this can be modulated by certain treatments such as negative wound therapy. 42 We suggest that understanding further the factors that limit and colleagues demonstrated that the microvasculature density in the deeper skin layers did not change during treatment with negative pressure therapy. 44 In contrast, we found a similar pattern of CD31 localisation in non-involved tissue, but in proximal and wound edge tissue, we noted an increase in CD31 immunoreactivity in the deeper reticular tissue, particularly at the wound edge. However, as shown in mice, quantity of vessels does not equate to functional, perfused vasculature. 7,45 It is possible that negative pressure therapy specifically targets vessels in the superficial layers of the skin as opposed to the deeper reticular dermis, but the fact our tissue was harvested from end-stage disease could also account for differences.
As considerable evidence particularly from tumour formation shows that CD31 is not always the most reliable marker of angiogenesis, we next investigated the localisation of the angiogenesis associated CD146. 24 CD146 is expressed by both endothelial cells and mural cells including pericytes and we show here that microvasculature at the edge of chronic wounds label extensively with this marker.
Immunoreactivity was strongly associated with mural cells surrounding the endothelium of vessels, showing significant pericyte coverage.
While endothelial localisation of CD146 is associated with neovascularisation, 46 as discussed previously, coverage of endothelial plexus by mural cells and pericytes can actually prevent further modifications of the microvasculature. 41 This could suggest that the pattern of vessel formation evident on the edge of non-healing skin wounds is dysfunctional and that these vessels are not capable of sprouting into the wound bed to restore oxygen and nutrient supply.
Endoglin (CD105), is also considered a more reliable marker of developing microvasculature than CD31, [47][48][49] and we show here that it is highly expressed in the endothelium of the microvasculature adjacent to the chronic wound bed. The expression of endoglin is particularly significant as it is required for efficient TGF-β signalling in angiogenesis, and it demonstrates that even in end stage tissue, endothelial cells are still attempting to initiate processes associated with angiogenesis. Interestingly, mice that are heterozygous for endoglin (eng +/À ) exhibit large numbers of irregular, dilated, and thinner-walled vessels compared with wild-type counterparts. 50 In endothelial cells, endoglin acts as an ancillary receptor for the TGF-β receptors ALK-1 and ALK-5. 28 ALK5 mediates TGF-β signalling in quiescent endothelial cells, inhibiting endothelial proliferation, 51 but association of endoglin with ALK-1 has the opposite effect, triggering endothelial cell division. 52 We show here that on the edge of non-healing skin wounds, this signalling axis appears to be functional, with ALK-1/endoglin colocalisation prominent and PCNA labelling of endothelial cells evident in serially cut sections of the same wound edge regions. In general, we show that ALK-5 is present at a much lower level, although is not completely absent. However, it is known that to get optimal ALK-1 kinase activity, ALK-5 signalling is also required at a basal level. 51 With the caveat that our analysis is based on histological assessment, it does appear that the endoglin/ALK-1/ALK-5-endothelial proliferation signalling axis is still functional in end stage chronic wound tissue, suggesting that the failure of capillaries to sprout into the wound bed is likely due to a toxic cellular microenvironment as opposed to inappropriate angiogenic signalling.
One of the fundamental cellular abnormalities considered to play a major role in impaired skin healing is cellular senescence. 53 63 Our data in this article supports this with most p16 INK4A labelling present in the keratinocytes. The exact role of cellular senescence in wound repair could be considered controversial, with research in murine skin healing models demonstrating that senescence in fibroblasts and endothelial cells is required for proper healing. 64 It does open the possibility that cells on the edge of chronic wounds could be quiescent in contrast to senescent, with quiescence associated with arrest in G 0 , but the cells can re-enter the cell cycle upon stimulation by an appropriate signal. 65 In this study, we harvested chronic skin wound tissue at elective amputation of the affected limb. Therefore, the tissue source we used in the study must be considered end-stage of the disease process for the individual concerned. Indeed, nearly all of the individuals recruited into the study had extensive PAD and were diagnosed with CLI level 6. The major advantage of our approach is that we can isolate large pieces of tissue from the wound and surrounding tissue for histological reconstruction, as well as remove skin sections from other parts of the limb to serve as non-involved control tissue. By quantifying how the microcirculation changes in end-stage disease, it is theoretically possible to use this a baseline with which to assess these same markers in tissue removed through debridement in patients in which amputation is not considered necessary due to their disease state. In essence, at which stage of wound management would these markers be detectable in debrided tissue and how would it correlate with clinical outcome for that wound.
In conclusion, we identified that even in end-stage chronic wound tissue, normal angiogenic processes appear to be active in the dermis.
The results of this study have implications for our understanding of chronic skin wound pathology and treatments, with our data suggesting that the wound edge microenvironment may be a major hurdle to be overcome in resolution of the wounds to re-establish the microcirculation into the wound bed.