RESEARCH ARTICLE

CD40-mediated HIF-1α expression underlying microangiopathyin diabetic nerve pathologyHung-Wei Kan1, Jung-Hsien Hsieh1,2, Hsiung-Fei Chien1,2,*, Yea-Huey Lin3, Ti-Yen Yeh1, Chi-Chao Chao3,‡

and Sung-Tsang Hsieh1,3,4,‡

ABSTRACTTo understand the pathology and molecular signatures ofmicroangiopathy in diabetic neuropathy, we systemically andquantitatively examined the morphometry of microvascular andnerve pathologies of sural nerves. In the endoneurium of diabeticnerves, prominent microangiopathy was observed, as evidenced byreduced capillary luminal area, increased capillary basementmembrane thickness and increased proportion of fibrin(+) bloodvessels. Furthermore, capillary basement membrane thickness andthe proportion of fibrin(+) blood vessels were correlated with smallmyelinated fiber density in diabetic nerves. In diabetic nerves, therewas also significant macrophage and T cell infiltration, and cluster ofdifferentiation 40 (CD40) expression was increased. The molecularalterations observed were upregulation of hypoxia-inducible factor-1α(HIF-1α), mitogen-activated protein kinase-activated protein kinase 2(MK2; MAPKAPK2) and phosphatase and tensin homolog (PTEN). Inaddition, HIF-1α was correlated with small myelinated fiber densityand capillary luminal area, while both MK2 and PTENwere correlatedwith capillary basement membrane thickness. The molecularcascades were further demonstrated and replicated in a cell modelof microangiopathy on human umbilical vein endothelial cells(HUVECs) exposed to high-glucose medium by silencing of CD40,PTEN and HIF-1α in HUVECs using shRNA. These data clarified thehierarchy of the molecular cascades, i.e. upregulation of CD40leading to HIF-1α expression in endothelium and nerve fibers. Inconclusion, this study revealed the association of microangiopathy,thrombosis and inflammatory infiltrates with nerve degeneration indiabetic nerves, demonstrating that CD40 is a key molecule for theupregulation of HIF-1α and PTEN underlying the severity ofmicroangiopathy.

KEY WORDS: Diabetic neuropathy, Microangiopathy, Ischemia,Thrombosis, Inflammation

INTRODUCTIONPeripheral nerve degeneration in diabetic neuropathy, a majorcomplication of diabetes with a frequency of 25-60%, has beenattributed to multifactorial mechanisms, such as metabolicdysregulation of sorbitol, oxidative stress and neurotrophicdeficiency, according to studies on experimental diabetes inrodents and human nerve pathology (Zochodne, 1999; Brownlee,2001; Kennedy and Zochodne, 2005; Edwards et al., 2008; Bakovicet al., 2018). Although microangiopathy is listed as one of the majormicrovascular complications of diabetes, including reducedendoneurial capillary density and endothelial dysfunction, thecascades leading to this cellular pathology and its molecularconsequences in diabetic nerves have not yet been systemicallyinvestigated.

Chronic inflammation is frequently coupled with microvasculardisease in diabetes (Crook, 2004; Said, 2006). The recruitment ofinflammatory cells probably occurs through hyperglycemia andglycation (Brownlee, 2001), which are also risk factors forendothelial dysfunction and thrombus formation (Grant, 2007).Inflammatory mediators can augment the expression of tissue factorgenes in endothelial cells, leading to thrombosis, which, in turn, hasbeen postulated as a cause of leukocyte adhesion and inflammatoryresponse amplification (Libby and Simon, 2001). Among theinterweaved mechanisms between inflammation and thrombosis,the cluster of differentiation 40 (CD40) signaling pathway plays animportant role (Mach et al., 1997). The CD40 ligand binds to itsreceptor CD40 and induces tissue factor expression in macrophagesand endothelial cells. However, the relationship between the CD40-CD40 ligand system and microangiopathy in diabetic nerves hasremained unexplored.

Tissue ischemia, one of the major consequences ofmicroangiopathy (Agnic et al., 2015), activates a series ofsignaling pathways, such as hypoxia-inducible factor-1α (HIF-1α)(Semenza, 2000). Among these cascades, both the mitogen-activated protein kinase (MAPK) family and phosphoinositol 3-kinase (PI3K; PIK3CA)/Akt pathways constitute importantresponses to hypoxia and regulate HIF-1α protein levels andtransactivation (Emerling et al., 2005; Maynard and Ohh, 2007;Khandrika et al., 2009). The MAPK family mediates the protectionof nerve fiber function and structure (Price et al., 2004; Jolivaltet al., 2011). Phosphatase and tensin homolog (PTEN), anantagonist of PI3K, is elevated during experimental diabetes andcontributes to impaired regeneration of diabetic axons (Park et al.,2010; Singh et al., 2014) as well as in diabetic kidneys (Delic Jukicet al., 2018). However, the roles of the MAPK family and PI3K/Aktsignaling molecules in microangiopathy have not been fullyexamined in human diabetic tissues. These observations raised thepossibility of exploring the underlying molecular mechanismsleading to microangiopathy and its relationship with nervepathology, either beneficial or detrimental to nerve integrity.Received 16 January 2018; Accepted 12 March 2018

1Department of Anatomy and Cell Biology, National Taiwan University, Taipei10051, Taiwan. 2Department of Surgery, National Taiwan University Hospital, Taipei10002, Taiwan. 3Department of Neurology, National Taiwan University Hospital,Taipei 10002, Taiwan. 4Graduate Institute of Brain and Mind Sciences, College ofMedicine, National Taiwan University, Taipei 10051, Taiwan.*Present address: Department of Surgery, Taipei Medical University Hospital andCollege of Medicine, Taipei Medical University, Taipei 11042, Taiwan.

‡Authors for correspondence (b1401019@ms17.hinet.net; shsieh@ntu.edu.tw)

H.-W.K., 0000-0002-5361-4511; C.-C.C., 0000-0001-6499-5789; S.-T.H., 0000-0002-3188-8482

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

To address the above issues, this study aimed to investigate(1) microangiopathy and its correlation with nerve pathology, and(2) molecule signatures related to microangiopathy.

RESULTSEndoneurial capillary morphometry, thrombosis andinflammatory cell infiltrationTo explore the pattern of microangiopathy, we analyzed the vascularmorphometry of diabetic nerves. The pattern of microangiopathy indiabetic nerves was distinct from the vascular appearance of controlnerves in which the capillary lumen was evident, with anappropriate thickness relative to the vascular basement membrane(Fig. 1A,B). In diabetic nerves, the capillary basement membrane

was thicker than that of control nerves by 52.9% (4.28±0.73 versus2.80±0.43 μm, P<0.0001; Fig. 1C), and the capillary luminal areawas reduced by 73.1% (4.09±1.61 versus 15.2±5.12%, P<0.0001;Fig. 1D). However, endoneurial capillary densities were similar inboth control and diabetic groups (72.4±28.0 versus 58.2±10.6capillaries/mm2, P=0.398).

To identify the compositions of the capillary basementmembrane, we performed multiple labeling of CD31 (PECM1),smooth muscle α-actin (SMA; ACTA2) and collagen IV forendothelial cells, pericytes and basement membrane, respectively.The increased capillary basement membrane thickness in diabeticnerves mainly resulted from the thickening of the basementmembrane components (Fig. 1E,F). The presence of cleavedcaspase 3(+) endothelial cells provides further signs ofmicrovascular insults in diabetic nerves (Fig. 1G,H).

Given the effects of hyperglycemia and glycation on fibrinstructures (Grant, 2007), we next asked whether thrombus formationwas enhanced in diabetic nerves and examined its consequences withfibrin immunohistochemistry. All blood vessels and capillaries incontrol nerves were negative for fibrin (Fig. 1I). By contrast, 16 of 28(57.1%) sural nerves in the diabetic group had fibrin deposition inboth endoneurial capillaries and epineurial vessels (Fig. 1J; Fig. S1),indicating enhanced thrombus formation in diabetic nerves.

Because chronic inflammation plays a crucial role in initiatingendothelial cell dysfunction and thrombus formation (Libby andSimon, 2001), we first examined the patterns of inflammatory cellinfiltration in diabetic nerves, including macrophages (CD68), Tcells (CD3 proteins) and B cells (CD20; MS4A1) (Fig. S2). Thedensity of CD68(+) cells was higher in diabetic nerves than incontrol nerves (244.6±117.0 versus 26.6±19.6 cells/mm2,P=0.026). T cell infiltration was absent in control nerves, butthere was significant infiltration of CD3(+) cells around theepineurial blood vessels of diabetic nerves (4.34±3.04 versus 0±0cells/vessel, P=0.019). CD20 immunoreactivity was absent in thesural nerve of both control and diabetic nerves. Notably, bothmacrophage and T cell infiltration indices were significantlycorrelated with the proportion of fibrin(+) blood vessels indiabetic nerves (r=0.45, P=0.042 for macrophage infiltrationindex; r=0.68, P=0.015 for T cell infiltration index; Fig. 2J,K).

Fig. 1. Microangiopathy in diabetic nerves. (A,B) Semithin sections revealedcapillaries that were obviously identified by endothelial cells (arrows), lumen (L)and capillary basement membrane (BM). The endoneurial capillary of controlsubjects appeared normal, with appropriate capillary basement membranethickness and adequate capillary luminal area (A). The capillary basementmembrane thickness was increased in diabetes patients, with hypertrophicendothelial cells and a decreased capillary luminal area (B). (C,D)Quantificationof microangiopathy by semithin sections showed significantly increasedcapillary basement membrane thickness and significantly reduced capillaryluminal area in diabetic nerves (n=28) compared with control nerves (n=6).****P<0.0001 by two-tailed Student’s t-test. Data are mean±s.d. (E,F) Multiplelabeling of CD31 (green), smooth muscle actin (SMA, red) and collagen IV(COLIV, cyan) was performed on endoneurial capillaries. DAPI (blue) was usedfor nuclear staining. Comparable to the semithin sections, the capillarybasement membrane consisted of pericyte nuclei (arrows), SMA and COLIV;and the capillary basement membrane thickness was increased in diabeticcapillaries (F) compared with control capillaries (E). Arrowheads indicate nucleiof endothelial cells. (G,H) Epineurial blood vessels were examined with cleavedcaspase 3 (Cas3, green) and SMA (red) immunochemistry. Compared withcontrol blood vessels (G), diabetic blood vessels exhibited increased Cas3 inendothelial cells (H), which indicated microvascular insults. (I,J) Thrombosescontaining fibrin clots were stained with FITC-conjugated anti-fibrin antibody(green) and endoneurial capillaries were revealed by SMA (red)immunohistochemistry. Fibrin(+) clots were absent (I), whereas there weremarkedly increased fibrin(+) clots (J) in diabetic capillaries. Scale bars: 10 μm.

2

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

Given the enhanced coagulation and inflammation in diabeticnerves, and CD40 has been implicated in the interweavedmechanisms of these complications (Libby and Simon, 2001;Mach et al., 1997), CD40 immunostaining was performed. CD40immunoreactivity was only minimally detectable in the endothelialcells of control nerves but became upregulated in diabetic nerves(Fig. 2A-I). In diabetic nerves, CD40 expression was enhanced inthe endothelial cells and became marked in inflammatory cells.Quantitatively, CD40(+) cells were significantly increased in theendoneurium of diabetic nerves (63.6±29.4 versus 8.9±8.2 cells/mm2, P=0.002) and were correlated with the proportion of fibrin(+)vasculatures of diabetic nerves (r=0.73, P=0.0028; Fig. 2L). Theseresults suggested an association of microvascular insults withchronic inflammation, and CD40 might serve as a mediator ofenhanced thrombus formation between inflammatory cells andendothelial cells in the diabetic nerves.

Sural nerve pathology and correlation with microangiopathyTo investigate diabetic effects on nerve morphometry, we analyzednerve pathology and its relationship with microangiopathy. In

control subjects, myelinated fibers were abundant with two distinctlarge and small subpopulations (Fig. 3A). In diabetic nerves, bothlarge and small myelinated fibers were significantly reduced(Fig. 3B). This result was confirmed by the reduced myelinatedfiber density (2540±1934 versus 7458±1519 fibers/mm2,P=0.0001), with a similar reduction in both large (1234±1055versus 3973±1029 fibers/mm2, P=0.0002) and small myelinatedtypes (1306±1149 versus 3485±703 fibers/mm2, P=0.0004). Incontrol nerves, the fiber diameter histogram showed a bimodaldistribution, and the pattern became unimodal in diabetic nerves(P=0.028 by Chi-squared goodness-of-fit test; Fig. 3C,D). Giventhis change in the fiber diameter distribution representing patterns offiber type pathology, we focused on small myelinated fibers in thefollowing analyses.

We further analyzed the relationship between nerve pathologyand the parameters of microangiopathy. Capillary basementmembrane thickness and the proportion of fibrin(+) blood vesselswere determinants of small myelinated fiber density (SMFD)[r=−0.53, P=0.0035 for capillary basement membrane thickness;r=−0.59, P=0.017 for the proportion of fibrin(+) blood vessels;

Fig. 2. CD40 expression in endothelial cells andinflammatory cells. (A-I) Double staining of CD68(green, A,D), CD3 (green, G), and CD40 (red, B,E,H) was performed in sural nerves. CD40immunoreactivity in the endothelial cells of controlnerves was minimal but was increased in diabeticnerves. In addition, there was strong colocalizationof CD40 with CD68 (F) and CD3 (I) in diabeticnerves. Scale bar: 100 μm. (J-L) In diabetic nerveswith fibrin(+) immunostaining (n=16), inflammatorycell infiltration and CD40 expression weresignificantly related to fibrin(+) vessels. Theproportions of fibrin(+) vessels are plotted with theinfiltration indices of macrophages (J), T cells (K)and CD40(+) cells (L). Solid lines, regression lines;dashed lines, 95% confidence intervals (CIs) byPearson’s correlation analysis.

3

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

Fig. 3E,F]. The above observations indicated that the quantityof small myelinated nerve fibers was correlated with the severityof microangiopathy.

Expression of HIF-1α, mitogen-activated protein kinase-activated protein kinase 2 and PTEN in sural nervesAccording to the above observations of microangiopathy and itsconceivably consequent ischemia, we examined the expression ofHIF-1α in sural nerves. HIF-1α immunoreactivity was sparse incontrol nerves, whereas it was upregulated in the axons and bloodvessels of diabetic nerves (Fig. 4A,B). Based on the quantificationof western blots, HIF-1α expression was increased in the diabeticpatients compared with the controls (0.72±0.34 versus 0.27±0.19,P=0.020; Fig. 4C), and was correlated with the SMFD (r=0.43,P=0.039; Fig. 4D) and capillary luminal area in the diabetic group(r=0.52, P=0.012; Fig. 4E). Because the MAPK and PI3K(PIK3CA)/Akt signaling pathways regulate HIF-1α (Emerlinget al., 2005; Maynard and Ohh, 2007), we examined the

expression levels of phosphorylated extracellular signal-regulatedkinase 1/2 (pERK1/2), a downstream molecule of the MAPK/ERKpathway, mitogen-activated protein kinase-activated protein kinase2 (MK2; MAPKAPK2), a key enzyme marker of the p38 MAPKpathway, and PTEN, an inhibitory molecule of the PI3K/Aktpathway. pERK1/2 immunoreactivities were similar in both controland diabetic sural nerves (Fig. S3). MK2 immunoreactivity wasminimal in control nerves but became abundantly expressed inthe microvasculature of diabetic nerves (Fig. 5A; Fig. S4A). Forquantitative comparisons, MK2 expression was increased indiabetic nerves compared with control nerves (2.86±1.57 versus1.24±0.80, P=0.047; Fig. 5B), and was correlated with the capillarybasement membrane thickness in the diabetic group (r=0.51,P=0.016; Fig. 5C). In control nerves, there was limited PTENimmunoreactivity. By contrast, PTEN immunoreactivity wasincreased in the blood vessels of diabetic nerves (Fig. 6A;Fig. S4B). Quantitatively, PTEN expression was increased indiabetic nerves compared with control nerves (1.27±0.80 versus0.32±0.27, P=0.030; Fig. 6B), and was correlated with the capillarybasement membrane thickness in the diabetic group (r=0.52,P=0.011; Fig. 6C). Taken together, HIF-1α, MK2 and PTENexpression all correlated with the morphometric index of vascularintegrity, i.e. capillary luminal area or capillary basement membranethickness, indicating that the activation of these molecules wasassociated with diabetes-induced microvascular insults.

Molecular alterations in human umbilical vein endothelialcell cultures exposed to high-glucose mediumTo investigate whether the above molecular alterations were casuallyrelated to hyperglycemia, we examined the molecular expressionlevels in human umbilical vein endothelial cells (HUVECs) culturedin high-glucose medium. After exposure to added 3, 15 and 30 mMD-glucose for 7 days, there was significant dose-dependentupregulation of HIF-1α (P=0.038), MK2 (P=0.034), PTEN(P=0.033) and CD40 (P=0.031), particularly in the HUVEC culturesexposed to added 30 mM D-glucose (Fig. S5). These in vitro findingsrecapitulated those in the endothelial cells in diabetic nerves, indicatingthe detrimental effects of a high-glucose environment on endothelialcells, resulting in alterations of molecular signatures.

CD40 regulates HIF-1α and PTEN, but not MK2, in HUVECsGiven that CD40 induces intracellular signaling in endothelial cellsthrough the PI3K/Akt and p38 MAPK pathways (Deregibus et al.,2003; Chakrabarti et al., 2007), we next silenced CD40 in HUVECswith lentivirus-based short hairpin RNA (shRNA) and then culturedthe cells in a medium containing added 30 mMD-glucose for 7 daysto examine the effects on molecular cascades of microangiopathy.The knockdown effects among the four constructs were assessedwith western blotting. Two shRNAs with higher knockdownefficiency were used in the subsequent experiments. CD40 shRNAsignificantly silenced CD40 expression in HUVECs (P=0.0054 byone-way ANOVA; Fig. 7A). Accordingly, HIF-1α and PTEN weresignificantly downregulated in concert with CD40 by CD40 shRNA(P=0.0005 by one-way ANOVA for HIF-1α; P=0.0003 by one-wayANOVA for PTEN). However, there was no difference in MK2expression following CD40 silencing (Fig. S6), indicating thatHIF-1α and PTEN were downstream molecules of the CD40signaling pathway.

To clarify the relationship between CD40, HIF-1α and PTEN, weperformed knockdown experiments on HIF-1α (Fig. 7B) and PTEN(Fig. 7C). Western blotting demonstrated that the silencing ofHIF-1α or PTEN had no effect on CD40 expression. By contrast,

Fig. 3. Nerve degeneration in diabetes. (A,B) Sural nerveswere cut into 1-μmsections and then stained with Toluidine Blue. The sural nerve fascicle from acontrol subject revealed abundant myelinated fibers (A). By contrast, thediabetic sural nerve fascicle exhibited a marked reduction of both large- andsmall-diameter myelinated fibers (B). Scale bar: 100 μm. (C,D) The histogramsshowed the average frequencies of myelinated fibers according to myelinatedfiber diameters of the control subjects (C) and diabetes (D). In diabetic nerves(n=28), the histogram had a unimodal distribution which was distinct from abimodal distribution in control nerves (n=6) (χ2=25.7, P=0.028 by Chi-squaredtest). (E,F) SMFD was significantly correlated with capillary basementmembrane thickness (n=28) and the proportion of fibrin(+) vessels (n=16).Solid lines, regression lines; dashed lines, 95% CIs by Pearson’s correlationanalysis.

4

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

inhibition of PTEN resulted in HIF-1α downregulation (P<0.0001by one-way ANOVA), while PTEN expression was similarfollowing HIF-1α knockdown. Taken together, these aboveobservations proposed a hierarchy of molecular signalingpathways from CD40 to HIF-1α (Fig. 7D).

DISCUSSIONThis report documented CD40 as the major molecular signatureof microangiopathy in diabetic nerve pathology from thedimensions of inflammation-related thrombosis. We demonstrated(1) microangiopathy (basement membrane thickness and thrombosis)as a determinant of small myelinated fiber degeneration, (2) increasedinflammation with a pro-coagulation cascade, (3) the alteredexpression levels of HIF-1α, MK2 and PTEN and their correlationswith the pathologic severity of diabetic microvasculature (Fig. 7D),and (4) direct evidence of CD40 upregulation induced by high-glucose exposure resulting in the expression of HIF-1α and PTENin vitro.In diabetic patients, vascular integrity (the thickness and patency

of endoneurial capillaries) was a significant determinant of nervedegeneration, far more important than the quantity of endoneurialcapillaries, i.e. endoneurial capillary density. Despite the similar

quantity of endoneurial capillaries between diabetic nerves andcontrol nerves, the quality of capillaries in diabetic nerves was muchworse than that of control nerves, as assessed by capillary basementmembrane thickness, capillary luminal area and fibrin clots in bloodvessels, recapitulating the vasculopathy in ischemic limbs caused byperipheral vascular disease. However, the global and generalizedloss of nerve fibers in the current study was distinct to focal nervefiber injury involving certain fascicles or fibers in peripheralvascular disease (Nukada et al., 1996).

This study provided the morphometry evidence of microvascularpathology, i.e. basement membrane thickness and fibrin(+) clotsin capillaries as the major contributors of SMFD, which was incontrast to most studies considering myelinated fibers as a whole.Despite the fact that both large and small myelinated fiber densitieswere diminished in diabetic patients, small nerve fibers are morevulnerable to diabetic insults, and nerve regeneration is frequentlyobserved in diabetic patients (Løseth et al., 2008), indicating that therole of small myelinated fibers is much more exclusive than that oflarge myelinated fibers. In addition, given that small myelinatedfibers contribute to skin nerve terminals (Herrmann et al., 1999) andsmall fiber neuropathy constitutes an integral part of diabeticneuropathy (Shun et al., 2004), it is important to determine how

Fig. 4. Upregulation of HIF-1α in diabetic suralnerves. (A,B) Sural nerve biopsy sections wereimmunostained. Nerve fibers (A) and endoneurialcapillaries (B) were revealed by β-tubulin III and SMA(red), respectively. HIF-1α (green) expression wasmainly upregulated in axons, endothelial cells,perivascular infiltrating cells and vascular smoothmuscle cells of diabetic patients comparedwith thoseof control subjects. Scale bars: (A) 25 μm; (B) 10 μm.(C) HIF-1α expression was validated and quantifiedwith western blotting. The expression of HIF-1α wassignificantly increased in the sural nerves of diabeticpatients (n=23) compared with those of controlsubjects (n=3). *P<0.05 by two-tailed Student’st-test. (D,E) The HIF-1α expression level of diabeticnerves was plotted against SMFD and capillaryluminal area. These parameters were significantlycorrelated with the HIF-1α expression level amongthe diabetic group. Solid lines, regression lines;dashed lines, 95% CIs by Pearson’s correlationanalysis.

5

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

small myelinated fibers were affected in diabetes-induced nervedegeneration.This study provided direct evidence for the simultaneous

existence of two pathological processes, inflammation andthrombosis, in microangiopathy of diabetic neuropathy. Althoughinflammation has been proposed as an important pathophysiologyassociated with diabetes, direct evidence of inflammation as animportant pathology in diabetic neuropathy is sparse. The currentreport documented that diabetic microvasculature was associatedwith the inflammation of macrophage and T cell infiltration. Mostmacrophages expressing CD68, a total infiltrated macrophagemarker (Wentworth et al., 2010), were also immunoreactive forCD40 as one of the M1 polarization markers (Vogel et al., 2014),indicating that the inflammatory response was prompted in diabeticsural nerves. Furthermore, the presence of fibrin(+) clots in diabeticmicroangiopathy indicated that the inflammatory response resultingfrom long-term diabetes could serve as an important mechanism forthrombus formation. A critical question was whether there was amolecular candidate contributing to both pathological processesof inflammation and thrombosis. The upregulation of CD40 inendothelial cells, macrophages and T cells in diabetes suggested thatenhanced CD40 expression could play a key role in mediatingleukocyte adhesion through the CD40-CD40L signaling pathway,resulting in the induction of pro-inflammatory and pro-thromboticprocesses (Libby and Simon, 2001; Lutgens et al., 2007). Takentogether, the association of CD40 with the proportion ofthrombosed capillaries in diabetic neuropathy underpinned theinteractions between these two networks of inflammation andthrombosis in the pathogenesis of microangiopathy leading to nervedegeneration.

This study demonstrated the three key molecular signatures ofmicroangiopathy-mediated diabetic neuropathy downstream ofCD40: HIF-1α, MK2 and PTEN. According to previous studies,the functions of HIF-1α are controversial. For example, HIF-1αinduces erythropoietin to promote neuron survival (Ostrowski et al.,2011), but another study indicated that HIF-1α promotes cell deathin the context of cerebral ischemia via p53 (Halterman et al., 1999).Although diabetes might have a complex repressive effect on thestabilization and transactivation of HIF-1α (Bento and Pereira,2011; Catrina et al., 2004), our results indicated that HIF-1αexpression was associated with better nerve profiles. In diabetes,this correlation presumably resulted from stimulating the recoveryof blood flow, similar to the induction of HIF-1α in the ischemiclimb (Sarkar et al., 2009). This assumption was comparable to ourfindings, i.e. HIF-1αwas also positively associated with the luminalarea of microvessels. Whether microangiopathy coexisted withsevere diabetic neuropathy or was the etiology of nerve degenerationhas been an issue of debate for decades. The current study suggestedthat tissue hypoxia might occur prior to nerve degeneration and thatthe capability for expressing HIF-1α was related to the protection ofnerve integrity. Further investigation on a cohort study of diabeticpatients could clarify this assumption.

In experimental diabetes, the p38 MAPK and PI3K/Akt signalingpathways have been thought to be responsible for reduced nerveconduction velocity and the impaired nerve regeneration of diabeticaxons, respectively (Price et al., 2004; Singh et al., 2014). This studyidentified additional mediators of MK2 and PTEN in disruptedmolecular cascades in epineurial blood vessels and endoneurialcapillaries, i.e. the upregulation of both MK2 and PTEN wasassociated with the degree of microangiopathy. Given that the p38

Fig. 5. Upregulation of MK2 in diabetic sural nerves.(A) Sural nerve biopsy sections were immunostained.Endoneurial capillaries were revealed by SMA (red). MK2(green) showed minimal expression in the smooth musclecells and endothelial cells of control subjects, but itsexpression was increased in the smooth muscle cells andendothelial cells of diabetic patients. Scale bar: 10 μm.(B) The expression of MK2 was quantified with westernblotting, and there was a significant increase in MK2expression in diabetic nerves (n=23) compared with controlnerves (n=3). *P<0.05 by two-tailed Student’s t-test.(C) The expression of MK2was significantly correlatedwithcapillary basement membrane thickness among thediabetic group. Solid lines, regression lines; dashed lines,95% CIs by Pearson’s correlation analysis.

6

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

MAPK and PI3K/Akt pathways are involved in the regulation of cellsurvival, this observationmight indicate that pericytes, smoothmusclecells and endothelial cells undergo apoptosis through the activation ofthe p38 MAPK pathway and inhibition of the PI3K/Akt pathway.These molecular alterations underlie the mechanisms of thickenedcapillary basement membrane in diabetic neuropathy, which includedthe cell debris of smooth muscle cells and pericytes (Giannini andDyck, 1994). In addition, MK2 is crucial for collagen IV degradationby regulating the expression of matrix metalloproteinase-2 (MMP-2)(Xu et al., 2006), possibly influencing the inflammatory infiltration ofmacrophages and T cells. A seminal study identified a set of genesignatures that classified the progression of diabetic neuropathies (Huret al., 2011). The current study thus provides another dimension toclassify diabetic neuropathies by molecular signatures.In addition to thickened basement membrane, diabetes contributes

to inflammation and coagulation in the microvasculature (Mach et al.,1997; Libby and Simon, 2001). CD40 plays an intermediary role inthese diabetic complications. By culturing HUVECs with high-glucose medium as a model mimicking microangiopathy inuncontrolled hyperglycemia of diabetes (Cagliero et al., 1991;Rubenstein et al., 2011; Patel et al., 2013), we demonstrated, for thefirst time, that CD40 was induced by high-glucose exposure. Wechose to investigate the relationship between HIF-1α, MK2 andPTEN, and CD40 as they were upregulated in the endothelial cells ofsural nerve microvasculatures and were all major determinants ofmicroangiopathy and nerve morphometry in diabetic neuropathy.Although the PI3K/Akt and p-38 MAPK signaling pathways areinvolved downstream of CD40 (Deregibus et al., 2003; Chakrabartiet al., 2007; Dormond et al., 2008), our data suggested that CD40

initiated the high-glucose-induced upregulation of PTEN, indicatingsuppressed PI3K/Akt signaling. Therefore, an increase in MK2 underthe high-glucose condition possibly resulted from hypoxia and theinflammatory response (Limbourg et al., 2015). Although it was notclear which process was the most major one in HIF-1α accumulationinduced by high-glucose exposure (Yan et al., 2012), we provideddirect evidence that the regulation of HIF-1α was, in part, associatedwith PTEN as the intermediate molecule in the CD40 signalingpathway. Collectively, these observations support the notion thatmetabolic derangements resulting from hyperglycemia induced bydiabetes disrupt themolecular profiles in endothelial cells. Furthermore,the current study provides a putative signaling pathway that could bea therapeutic target for diabetic complications; that is, becauseHIF-1α was beneficial for the integrity of the sural nerve, acombination of HIF-1α overexpression and CD40 inhibition couldbe a therapeutic strategy for improving microvascular and nerveintegrity in diabetic neuropathy.

ConclusionIn summary, this study demonstrated that microangiopathy,including morphological and molecular alterations, is a majorpathology substrate of diabetic nerve degeneration, and providesevidence that the CD40 signaling pathway mediates the interactionsbetween inflammation and thrombosis in diabetes. Furthermore,knockdown of CD40 attenuates high-glucose-induced upregulationof HIF-1α and PTEN in vitro. However, this cross-sectional studymight be limited by the fact that the sural nerves came from anadvanced stage of neuropathy with vasculopathy as the primaryetiology. Ideally, diabetic microangiopathy without neuropathy

Fig. 6. Upregulation of PTEN in diabetic sural nerves.(A) Sural nerve biopsy sections were immunostained.Endoneurial capillaries were revealed by SMA (red).PTEN showed limited expression in the endothelial cells ofcontrol nerves, but its expression was increased in thesmooth muscle cells and endothelial cells of diabeticnerves. Scale bar: 10 μm. (B) Measurement of PTENexpression by western blotting demonstrated that PTENwas significantly increased in diabetic nerves (n=23)compared with control nerves (n=3). *P<0.05 by two-tailedStudent’s t-test. (C) Capillary basement membranethickness was significantly correlated with PTENexpression level. Solid lines, regression lines; dashedlines, 95% CIs by Pearson’s correlation analysis.

7

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

would be the best control. Nevertheless, this report provides newinsights into microangiopathy and its contribution to nervepathology. Certainly, further studies are mandatory to include

diabetic microangiopathy in the absence of neuropathy for aprospective design.

MATERIALS AND METHODSStudy subjects, demographic data and nerve morphometryThis study examined the sural nerves from type 2 diabetic patients fromNational Taiwan University Hospital (Taipei, Taiwan). Patients usuallyvisited the hospital once every 1-3 months for adjustments of diabetic control.All glycemic data were collected for analysis before the nerve biopsy. Controlsural nerves came from a portion of the sural nerve during grafting for subjectswith nerve trauma. Sural nerve biopsies were obtained from a standard siteposterior to the lateral malleolus under local anesthesia or from the same siteof amputated limbs. The study was approved by the Institutional ReviewBoard at National Taiwan University Hospital, and signed written consentforms were obtained from all participants in order to use the biopsies andrecordings during their treatments. All clinical investigation was conductedaccording to the principles expressed in the Declaration of Helsinki.

There were 28 type 2 diabetic patients (15 males and 13 females). Amongthese patients, five had sural nerves taken for regular diagnostic biopsy, and 23had sural nerves taken at the time of amputation (Table 1). All amputatedpatients could ambulate independently before amputation. There were sixhealthy control subjects (two males and four females, aged 44.6±25.5 years)who were determined to be neuropathy-free by neurological examination andsymptom score. There was no difference in age between the biopsy subgroupof diabetic patients and control subjects (adjusted P=0.943 by one-wayANOVAwith Tukey’s post hoc test). Control subjects and the biopsy subgroupof diabetic patients were younger than the amputated subgroup of diabeticpatients (adjusted P=0.008 and P=0.021, respectively, by one-way ANOVAwith Tukey’s post hoc test). Because age was a potential confounder ofmicroangiopathy, we addressed this issue with multiple linear regressionmodels (Table S1), with each morphometric parameter as a dependentvariable, and age and disease status (control versus diabetes) as independentvariables. Disease status, i.e. diabetes, was associated with eachmorphometricparameter, indicating that the effects of diabetes on microangiopathy wereindependent of age. There was no difference in plasma glucose, nerve fiberdensities, and microvascular morphometry between the two subgroups ofdiabetic patients (Table 1). Thus, the data of these two subgroups were pooledtogether in the following analyses and presentations.

AntibodiesAnti-CD31 (1:500, M0834), anti-CD68 (1:50, M0876), anti-CD3 (1:50,M0742), anti-CD20 (1:50, CD20cy), and fluorescein isothiocyanate(FITC)-conjugated anti-fibrin (1:100, F0111) primary antibodies werepurchased fromDako (Glostrup, Denmark); anti-SMA (1:1000, A5228) andβ-tubulin III (1:1000, T5076) were obtained from Sigma-Aldrich (St. Louis,MO, USA); anti-collagen IV (1:50, 2150-0140) and anti-CD40 (1:100,MCA 1590) were purchased from AbD Serotec (Oxford, UK); anti-pERK1/2 (1:100, #9101) and anti-PTEN (1:500, #9559) were purchased from CellSignaling Technology (Danvers, MA, USA); anti-HIF-1α (1:100, sc-10790) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA,USA); anti-MK2 (1:200, ADI-KAP-MA015) was obtained from Stressgen(Victoria, Canada). Alexa Fluor 488-conjugated goat anti-rabbit IgG, AlexaFluor 488-conjugated goat anti-mouse IgG1a, Cy3-conjugated goat anti-mouse IgG2a and Alexa Fluor 647-conjugated donkey anti-rabbit IgGsecondary antibodies were purchased from Jackson ImmunoResearchLaboratories (West Grove, PA, USA).

Morphometry of nerve and microvascular pathologiesGlutaraldehyde-fixed nerves were post-fixed in 2% osmium tetroxide for2 h, dehydrated through a graded ethanol series, and embedded in Epon 812resin (Polyscience, Philadelphia, PA, USA). Semithin sections were cut on aReichert Ultracut E (Leica, Wetzler, Germany) and stained with ToluidineBlue.

Myelinated fibers were photographed at an original magnification of ×20,and endoneurial capillaries in each fascicle were photographed at anoriginal magnification of ×100 under a Leica DM2500 microscope.Morphometric parameters were measured using Image-Pro PLUS

Fig. 7. CD40 knockdown and putative signaling pathway in HUVECs afterexposure to high-glucosemedium. (A-C) HUVECs were infected with CD40shRNA, HIF-1α shRNA and PTEN shRNA. Following lentiviral infection for24 h, HUVECs were exposed to high-glucose (30 mM D-glucose) medium for7 days. Western blotting using anti-CD40, anti-HIF-1α and anti-PTENantibodies confirmed the knockdown efficiency of the molecules. CD40knockdown was associated with downregulation of HIF-1α and PTEN (A),whereas the expression of CD40 remained unchanged after knockdown ofHIF-1α or PTEN (B,C). Knockdown of HIF-1α did not affect the expression ofPTEN (B). By contrast, knockdown of PTEN significantly reduced theexpression of HIF-1α (C). The mean results of three independent experiments(±s.d.) are shown, and data were normalized using the amount of proteinexpression of a scrambled construct (SC). n.s., not significant. *P<0.05,**P<0.005, ***P<0.0005, ****P<0.0001 by one-way ANOVA with Dunnett’spost hoc test. (D) The knockdown experiments demonstrated the hierarchy ofthe high-glucose-induced signaling pathway. The illustration summarizes theassociations between the molecular signatures and morphometry in thecapillaries of diabetic sural nerves.

8

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

software (Media Cybernetics, Silver Spring, MD, USA). All myelinatedfibers in the entire fascicle were counted and were expressed as fibers/mm2

and further classified as large and small myelinated fibers, with 5 μm as thecutoff (Thomas et al., 2001). The capillary pathology criteria were based onprevious studies (Dyck et al., 1985; Estrella et al., 2008). Briefly, thecapillary basement membrane thickness was assessed by the mean widthfrom the base of endothelial cells to the outline of the basement membraneof pericytes; and the luminal area was depicted by the outline of endothelialcells facing the lumen using Image-Pro PLUS software. Endoneurialcapillaries in each endoneurium of all nerve fascicles were counted anddivided into each comparative area of endoneurium. The endoneurialcapillary density was derived from the average of each nerve fascicle andwas expressed as capillaries/mm2.

Immunofluorescence and quantification of microvascularinsultsSural nerve specimens were fixed overnight in 2% paraformaldehyde-lysine-periodate and then changed to phosphate buffer for storage. Frozensections (10 μm) were immunostained following established protocols(Hsieh et al., 2008). Briefly, nonspecific binding was blocked by 0.1%Triton X-100 and 0.5% nonfat milk in Tris buffer, and nerve sections wereincubated at 4°C overnight with primary antibodies suspended in 0.1%Triton X-100 and 0.5% nonfat milk in Tris buffer. After the sections wererinsed with Tris buffer, they were incubated with secondary antibodies atroom temperature for 1 h. After a rinse in Tris buffer, 4′,6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich) was used for nuclear staining, ifnecessary. Fluorescent samples were viewed and scanned under a Leica SP5confocal imaging system.

For fibrin quantification, the proportion of fibrin(+) blood vessels wasderived according to the number of fibrin(+) blood vessels divided by thetotal number of blood vessels in that nerve tissue. For CD40 quantification,CD40(+) cells in the endoneurium of all nerve fascicles were counted anddivided by the area of endoneurium and were expressed as cells/mm2.

Immunohistochemistry and quantification of inflammatory cellinfiltrationAdditional sural nerve frozen sections were quenched by 1%H2O2 and wereimmunohistochemically stained at 4°C overnight for CD68, CD3 and CD20with mouse spleen as positive controls. Sections were incubated withbiotinylated goat anti-rabbit IgG (Vector Laboratories, Burlingame, CA,USA) at room temperature for 1 h, followed by incubation with the avidin-biotin complex (Vector Laboratories) for 1 h. Diaminobenzidine (DAB,Sigma-Aldrich) was used as the chromogen.

For macrophage quantification, CD68(+) cells in each endoneurium of allnerve fascicles were counted and divided by each comparative area ofendoneurium. For T cell quantification, CD3(+) cells around epineurialblood vessels (within 100 μm) were counted and divided by the number of

epineurial blood vessels. The macrophage infiltration and T cell infiltrationindices were expressed as cells/mm2 and cells/vessel, respectively.

HUVEC culture and high-glucose exposureHUVECs were purchased from Bioresource Collection and Research Center(Hsinchu, Taiwan). Cultures were maintained with M199 (Sigma-Aldrich)containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA), 30 μg/ml endothelial cell growth supplement (Merck Millipore, Darmstadt,Germany), 25 U/ml heparin (Sigma-Aldrich), 100 units/ml penicillin(Gibco), and 100 μg/ml streptomycin (Gibco) at 37°C with 5% CO2/95%air in a humidified environment; and the entire medium was replenishedevery 2-3 days. HUVECs at passage 4-8 were used in the experiments. Forhigh-glucose exposure, D-glucose stock solution was added to the growthmedium to achieve the final added concentrations of 3, 15 and 30 mM. Thecell medium was changed, and high glucose was freshly added every otherday for 7 days.

For protein knockdown, lentiviruses expressing shRNA targeting CD40,HIF-1α and PTEN were produced and designed by RNAi Core, AcademiaSinica, Taiwan (Table S2). The PLKO.1 expression vector was used forshRNA expression. HUVECs (seeded at 2800 cells/mm2 cell density) wereinfected with the indicated lentivirus (multiplicity of infection=3)containing 8 μg/ml polybrene for 24 h and with fresh medium containing1 μg/ml puromycin and incubated for 48 h to select infected cells. HUVECswere further exposed to added 30 mM D-glucose for 7 days and were thenlysed in radioimmunoprecipitation assay (RIPA) buffer with a cocktail ofprotease inhibitors (Sigma-Aldrich) for analysis of proteins using westernblotting.

Western blot analysisFor each sural nerve, a 0.5-cm segment was desheathed by removing most ofthe connective tissue in the epineurium and homogenized with ice-coldlysis buffer containing proteinase inhibitor cocktail (Sigma-Aldrich).Homogenates were centrifuged at 12,000 g at 4°C for 15 min, and thesupernatants were collected. Proteins (10 μg) were separated by 10% (w/v)SDS-PAGE and were transferred to immobilon polyvinylidene difluoridemembranes (Millipore, Billerica, MA, USA). Nonspecific binding sites wereblocked with 5% (w/v) nonfat milk for 1 h, and blots were incubated withanti-HIF-1α, anti-MK2 or anti-PTEN antibody suspended in 5% nonfat milkat 4°C overnight. After washing in buffer (0.05 M Trizma base, 0.5 M NaCland 0.5% Tween 20), the blots were incubated with horseradish peroxidase-linked secondary antibodies (Promega, Madison, WI, USA) at roomtemperature for 1 h. The blots were washed again, and the bands weredetected using enhanced chemiluminescence (Thermo Fisher Scientific,Rockford, IL, USA) and were exposed to medical X-ray film (Fuji, Tokyo,Japan). The relative density of protein bands was analyzed by ImageJ v.1.47 h(National Institutes of Health, MD, USA), and the relative density of eachprotein band was normalized to the corresponding amount of β-actin.

Table 1. Clinical profiles, laboratory tests and nerve pathology

All Biopsy Amputation P-value

Demographic dataAge (years) 64.3±14.2 47.6±10.8 67.9±12.2 0.003Sex (male: female)* 15: 13 3: 2 12: 11 1.000Diabetic profilesDuration of diabetes (years) 17.0±11.0 11.7±9.7 17.8±11.1 0.378Fasting glucose (mg/dl) 144.4±44.1 123.3±33.6 148.9±45.6 0.379Postprandial glucose (mg/dl) 181.9±54.2 200.0±32.5 179.2±57.3 0.631HbA1c (%) 7.11±1.67 6.53±0.56 7.24±1.83 0.455Nerve morphometryMyelinated fiber density (fibers/mm2) 2540±1934 2357±2140 2580±1936 0.821Small myelinated fiber density (fibers/mm2) 1306±1149 1262±1184 1316±1168 0.926Microvasculature morphometryCapillary basement membrane thickness (μm) 4.28±0.73 4.12±0.75 4.31±0.73 0.601Capillary luminal area (%) 4.09±1.60 3.43±1.77 4.23±1.57 0.321

Data are expressed as the mean±s.d. P-values are derived from Student’s t-tests comparing regular biopsy and amputation results.*Analyzed by Chi-squared test.

9

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

Statistical analysisData were analyzed by statistical software Prism 6 (GraphPad Software, SanDiego, CA, USA). Numeric variables are expressed as the mean±s.d. andwere compared using two-tailed Student’s t-tests if the data followed aGaussian distribution. The differences in age between control subjects, thebiopsy subgroup of diabetic patients and the amputated subgroup of diabeticpatients were analyzed by one-way ANOVAwith Tukey’s post hoc test. Thecomparisons of sex and the distribution of nerve fiber diameter wereanalyzed by Chi-squared goodness-of-fit test. Pearson’s correlation betweenvariables was analyzed with the slope of the regression line, including the95% confidence interval (CI). One-way ANOVA with Dunnett’s post hoctest was used to evaluate the effects of different shRNA constructs onprotein expression. The results were considered significant at P<0.05.

AcknowledgementsWe thank the staff of the Imaging Core at First Core Labs, National Taiwan UniversityCollege of Medicine; Bioresource Collection and Research Center, Taiwan; andRNAi Core, Academia Sinica, Taiwan, for technical assistance.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsMethodology: H.-W.K., T.-Y.Y.; Formal analysis: H.-W.K.; Investigation: H.-W.K.;Resources: J.-H.H., H.-F.C., Y.-H.L.; Data curation: J.-H.H., H.-F.C., Y.-H.L.; Writing -original draft: H.-W.K., S.-T.H.; Writing - review & editing: H.-W.K., C.-C.C., S.-T.H.;Supervision: C.-C.C., S.-T.H.; Project administration: C.-C.C., S.-T.H.; Fundingacquisition: S.-T.H.

FundingThis work was supported by the Ministry of Science and Technology, Taiwan (104-2320-B-002-019-MY3 and 104-2321-B-002-070) and National Taiwan University(106R881001 to S.-T.H.).

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.033647.supplemental

ReferencesAgnic, I., Filipovic, N., Vukojevic, K., Saraga-Babic, M., Vrdoljak, M. andGrkovic, I. (2015). Effects of isoflurane postconditioning on chronic phase ofischemia-reperfusion heart injury in rats. Cardiovasc. Pathol. 24, 94-101.

Bakovic, M., Filipovic, N., Ferhatovic Hamzic, L., Kunac, N., Zdrilic, E., VitlovUljevic, M., Kostic, S., Puljak, L. and Vukojevic, K. (2018). Changes inneurofilament 200 and tyrosine hydroxylase expression in the cardiac innervationof diabetic rats during aging. Cardiovasc. Pathol. 32, 38-43.

Bento, C. F. and Pereira, P. (2011). Regulation of hypoxia-inducible factor 1 and theloss of the cellular response to hypoxia in diabetes. Diabetologia 54, 1946-1956.

Brownlee, M. (2001). Biochemistry and molecular cell biology of diabeticcomplications. Nature 414, 813-820.

Cagliero, E., Roth, T., Roy, S. and Lorenzi, M. (1991). Characteristics andmechanisms of high-glucose-induced overexpression of basement membranecomponents in cultured human endothelial cells. Diabetes 40, 102-110.

Catrina, S.-B., Okamoto, K., Pereira, T., Brismar, K. and Poellinger, L. (2004).Hyperglycemia regulates hypoxia-inducible factor-1alpha protein stability andfunction. Diabetes 53, 3226-3232.

Chakrabarti, S., Blair, P. and Freedman, J. E. (2007). CD40-40L signaling invascular inflammation. J. Biol. Chem. 282, 18307-18317.

Crook, M. (2004). Type 2 diabetesmellitus: a disease of the innate immune system?An update. Diabet. Med. 21, 203-207.

Delic Jukic, I. K., Kostic, S., Filipovic, N., Gudelj Ensor, L., Ivandic, M., Dukic,J. J., Vitlov Uljevic, M., Ferhatovic Hamzic, L., Puljak, L. and Vukojevic, K.(2018). Changes in SATB1 and PTEN expression in the kidneys of diabetic ratsduring ageing. Nephrol. Dial. Transplant gfy003.

Deregibus, M. C., Buttiglieri, S., Russo, S., Bussolati, B. and Camussi, G.(2003). CD40-dependent activation of phosphatidylinositol 3-kinase/Akt pathwaymediates endothelial cell survival and in vitro angiogenesis. J. Biol. Chem. 278,18008-18014.

Dormond, O., Contreras, A. G., Meijer, E., Datta, D., Flynn, E., Pal, S. andBriscoe, D. M. (2008). CD40-induced signaling in human endothelial cells resultsin mTORC2- and Akt-dependent expression of vascular endothelial growth factorin vitro and in vivo. J. Immunol. 181, 8088-8095.

Dyck, P. J., Hansen, S., Karnes, J., O’Brien, P., Yasuda, H., Windebank, A. andZimmerman, B. (1985). Capillary number and percentage closed in humandiabetic sural nerve. Proc. Natl. Acad. Sci. USA 82, 2513-2517.

Edwards, J. L., Vincent, A. M., Cheng, H. T. and Feldman, E. L. (2008). Diabeticneuropathy: mechanisms to management. Pharmacol. Ther. 120, 1-34.

Emerling, B. M., Platanias, L. C., Black, E., Nebreda, A. R., Davis, R. J. andChandel, N. S. (2005). Mitochondrial reactive oxygen species activation of p38mitogen-activated protein kinase is required for hypoxia signaling. Mol. Cell. Biol.25, 4853-4862.

Estrella, J. S., Nelson, R. N., Sturges, B. K., Vernau, K. M., Williams, D. C.,LeCouteur, R. A., Shelton, G. D. and Mizisin, A. P. (2008). Endoneurialmicrovascular pathology in feline diabetic neuropathy. Microvasc. Res. 75,403-410.

Giannini, C. and Dyck, P. J. (1994). Ultrastructural morphometric abnormalities ofsural nerve endoneurial microvessels in diabetes-mellitus. Ann. Neurol. 36,408-415.

Grant, P. J. (2007). Diabetes mellitus as a prothrombotic condition. J. Intern. Med.262, 157-172.

Halterman, M. W., Miller, C. C. and Federoff, H. J. (1999). Hypoxia-induciblefactor-1alpha mediates hypoxia-induced delayed neuronal death that involvesp53. J. Neurosci. 19, 6818-6824.

Herrmann, D. N., Griffin, J. W., Hauer, P., Cornblath, D. R. and McArthur, J. C.(1999). Epidermal nerve fiber density and sural nerve morphometry in peripheralneuropathies. Neurology 53, 1634-1640.

Hsieh, Y.-L., Chiang, H., Tseng, T.-J. and Hsieh, S.-T. (2008). Enhancement ofcutaneous nerve regeneration by 4-methylcatechol in resiniferatoxin-inducedneuropathy. J. Neuropathol. Exp. Neurol. 67, 93-104.

Hur, J., Sullivan, K. A., Pande, M., Hong, Y., Sima, A. A. F., Jagadish, H. V.,Kretzler, M. and Feldman, E. L. (2011). The identification of gene expressionprofiles associated with progression of human diabetic neuropathy. Brain 134,3222-3235.

Jolivalt, C. G., Fineman, M., Deacon, C. F., Carr, R. D. and Calcutt, N. A. (2011).GLP-1 signals via ERK in peripheral nerve and prevents nerve dysfunction indiabetic mice. Diabetes Obes. Metab. 13, 990-1000.

Kennedy, J. M. and Zochodne, D. W. (2005). Impaired peripheral nerveregeneration in diabetes mellitus. J. Peripher. Nerv. Syst. 10, 144-157.

Khandrika, L., Lieberman, R., Koul, S., Kumar, B., Maroni, P., Chandhoke, R.,Meacham, R. B. and Koul, H. K. (2009). Hypoxia-associated p38 mitogen-activated protein kinase-mediated androgen receptor activation and increasedHIF-1alpha levels contribute to emergence of an aggressive phenotype inprostate cancer. Oncogene 28, 1248-1260.

Libby, P. and Simon, D. I. (2001). Inflammation and thrombosis: the clot thickens.Circulation 103, 1718-1720.

Limbourg, A., von Felden, J., Jagavelu, K., Krishnasamy, K., Napp, L. C.,Kapopara, P. R., Gaestel, M., Schieffer, B., Bauersachs, J., Limbourg, F. P.et al. (2015). MAP-kinase activated protein kinase 2 links endothelial activationand monocyte/macrophage recruitment in arteriogenesis. PLoS ONE 10,e0138542.

Løseth, S., Stålberg, E., Jorde, R. and Mellgren, S. I. (2008). Early diabeticneuropathy: thermal thresholds and intraepidermal nerve fibre density in patientswith normal nerve conduction studies. J. Neurol. 255, 1197-1202.

Lutgens, E., Lievens, D., Beckers, L., Donners, M. and Daemen, M. (2007).CD40 and its ligand in atherosclerosis. Trends. Cardiovasc. Med. 17, 118-123.

Mach, F., Schonbeck, U., Sukhova, G. K., Bourcier, T., Bonnefoy, J.-Y., Pober,J. S. and Libby, P. (1997). Functional CD40 ligand is expressed on humanvascular endothelial cells, smooth muscle cells, and macrophages: implicationsfor CD40-CD40 ligand signaling in atherosclerosis. Proc. Natl. Acad. Sci. USA 94,1931-1936.

Maynard, M. A. andOhh, M. (2007). The role of hypoxia-inducible factors in cancer.Cell. Mol. Life Sci. 64, 2170-2180.

Nukada, H., van Rij, A. M., Packer, S. G. K. and McMorran, P. D. (1996).Pathology of acute and chronic ischaemic neuropathy in atherosclerotic peripheralvascular disease. Brain 119, 1449-1460.

Ostrowski, D., Ehrenreich, H. and Heinrich, R. (2011). Erythropoietin promotessurvival and regeneration of insect neurons in vivo and in vitro.Neuroscience 188,95-108.

Park, K. K., Liu, K., Hu, Y., Kanter, J. L. and He, Z. (2010). PTEN/mTOR and axonregeneration. Exp. Neurol. 223, 45-50.

Patel, H., Chen, J., Das, K. C. and Kavdia, M. (2013). Hyperglycemia inducesdifferential change in oxidative stress at gene expression and functional levels inHUVEC and HMVEC. Cardiovasc. Diabetol. 12, 142.

Price, S. A., Agthong, S., Middlemas, A. B. and Tomlinson, D. R. (2004).Mitogen-activated protein kinase p38mediates reduced nerve conduction velocityin experimental diabetic neuropathy: interactions with aldose reductase. Diabetes53, 1851-1856.

Rubenstein, D. A., Maria, Z. and Yin, W. (2011). Glycated albumin modulatesendothelial cell thrombogenic and inflammatory responses. J. Diabetes Sci.Technol. 5, 703-713.

Said, G. (2006). Inflammation in diabetic neuropathies. Eur. Neurol. Rev. 6, 73-76.Sarkar, K., Fox-Talbot, K., Steenbergen, C., Bosch-Marce, M. and Semenza,

G. L. (2009). Adenoviral transfer of HIF-1alpha enhances vascular responses tocritical limb ischemia in diabetic mice. Proc. Natl. Acad. Sci. USA 106,18769-18774.

10

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms

Semenza, G. L. (2000). Surviving ischemia: adaptive responses mediated byhypoxia-inducible factor 1. J. Clin. Invest. 106, 809-812.

Shun, C. T., Chang, Y. C., Wu, H. P., Hsieh, S. C., Lin, W. M., Lin, Y. H., Tai, T. Y.and Hsieh, S. T. (2004). Skin denervation in type 2 diabetes: correlations withdiabetic duration and functional impairments. Brain 127, 1593-1605.

Singh, B., Singh, V., Krishnan, A., Koshy, K., Martinez, J. A., Cheng, C.,Almquist, C. and Zochodne, D. W. (2014). Regeneration of diabetic axons isenhanced by selective knockdown of the PTEN gene. Brain 137, 1051-1067.

Thomas, P. K., Kalaydjieva, L., Youl, B., Rogers, T., Angelicheva, D., King,R. H. M., Guergueltcheva, V., Colomer, J., Lupu, C., Corches, A. et al. (2001).Hereditary motor and sensory neuropathy-russe: new autosomal recessiveneuropathy in Balkan Gypsies. Ann. Neurol. 50, 452-457.

Vogel, D. Y. S., Glim, J. E., Stavenuiter, A.W. D., Breur, M., Heijnen, P., Amor, S.,Dijkstra, C. D. and Beelen, R. H. J. (2014). Human macrophage polarization in

vitro: maturation and activation methods compared. Immunobiology 219,695-703.

Wentworth, J. M., Naselli, G., Brown,W. A., Doyle, L., Phipson, B., Smyth, G. K.,Wabitsch, M., O’Brien, P. E. and Harrison, L. C. (2010). Pro-inflammatoryCD11c+CD206+ adipose tissue macrophages are associated with insulinresistance in human obesity. Diabetes 59, 1648-1656.

Xu, L., Chen, S. and Bergan, R. C. (2006). MAPKAPK2 and HSP27 aredownstream effectors of p38 MAP kinase-mediated matrix metalloproteinasetype 2 activation and cell invasion in human prostate cancer. Oncogene 25,2987-2998.

Yan, J., Zhang, Z. and Shi, H. (2012). HIF-1 is involved in high glucose-inducedparacellular permeability of brain endothelial cells.Cell. Mol. Life Sci. 69, 115-128.

Zochodne, D. W. (1999). Diabetic neuropathies: features and mechanisms. BrainPathol. 9, 369-391.

11

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033647. doi:10.1242/dmm.033647

Disea

seModels&Mechan

isms