Objectives The pathophysiology of dermatomyositis (DM) remains unclear, combining immunopathological mechanisms with ischaemic changes regarded as a consequence of membranolytic attack complex (MAC)-induced capillary destruction. The study is a reappraisal of the microvascular involvement in light of the microvascular organisation in normal human muscle.
Methods Muscle microvasculature organisation was analysed using 3D reconstructions of serial sections immunostained for CD31, and histoenzymatic detection of endogenous alkaline phosphatase activity of microvessels. An unbiased point pattern analysis-based method was used to evaluate focal capillary loss. Double immunostainings identified cell types showing MAC deposits.
Results The normal arterial tree includes perimysial arcade arteries, transverse arteries penetrating perpendicularly into the endomysium and terminal arterioles feeding a microvascular unit (MVU) of six to eight capillaries contacting an average of five myofibres. Amyopathic DM cases (n=3) and non-necrotic fascicles of early DM cases (n=27), showed patchy capillary loss in the form of 6-by-6 capillary drop-out, corresponding to depletion of one or multiple MVUs. MAC deposits were also clustered (5–8 immunostained structures, including endothelial cells, but also pericytes, mesenchymal cells and myosatellite cells).
Conclusions Capillary loss may not be the primary cause of muscle ischaemia in DM. The primary event rather stands upstream, probably at the level of perimysial arcade arteries around which inflammatory infiltrates predominate and which lumen may show narrowing in chronic DM. Ischaemia-reperfusion injury, which is favoured by autoimmune backgrounds in experimental models and which activates the complement cascade in capillaries, could represent an hitherto unsuspected (and potentially preventable) mechanism of muscle damage in DM.
Statistics from Altmetric.com
Myopathology of dermatomyositis (DM) typically combines inflammatory and microvascular alterations1–3 Inflammatory infiltrates are perivascular; they predominate in the perimysium and the perifascicular endomysium, and consist of a mixture of CD3+T-cells, with CD4+>CD8+cells, CD20+B-cells, CD68+macrophages4 and CD4+BDCA2+ plasmacytoid dendritic cells.5
Myofibre alterations include: (i) perifascicular atrophy; (ii) upregulation of major histocompatibility complex class I antigens with perifascicular enhancement; (iii) punch-out vacuoles; and (iv) microinfarcts.
Microvascular changes mainly include: (i) early capillary deposition of the complement C5b9 membranolytic attack complex (MAC); (ii) endothelial cell hyperplasia with typical tubuloreticular inclusions at electron microscopy; (iii) destruction of endothelial cells assessed focal capillary loss predominating in perifascicular areas.6
Despite increasing attention paid to microcirculatory changes,6–8 the pathophysiological mechanism of capillary depletion remains uncertain in DM.8 DM is usually regarded as a complement-mediated microangiopathy. It is believed to begin when putative antibodies directed against endothelial cells activate the complement cascade, leading to MAC formation. Consistently, MAC, C3b and C4b are detected early in serum and are deposited at both the adluminal and abluminal aspect of microvascular cells. Capillary necrosis/loss ensues, causing endofascicular hypoperfusion, microinfarcts and, presumably, perifascicular atrophy.9 In fact, neither putative antigens nor factors activating complement, such as immune complexes, have been reliably identified in DM patients. However, ischaemia remains unquestionable in DM immunopathogenesis.10
Routine observation that capillary loss is heterogeneous in DM, led us to define the micro-anatomic organisation of the human deltoid muscle vessels, in order to finely characterise muscle microcirculatory changes in DM. In contrast to the common belief, capillary loss in DM likely results from upstream pathogenic events.
Material and methods
Patients and material
Four normal deltoid muscle samples were used for the microanatomical study. They were collected from adult patients (38–48 years: two women and two men) without muscle weakness in whom suspected sarcoidosis or mitochondriopathy remained unproven. Forty deltoid muscle biopsies of adult patients with DM diagnosed from 2001 to 2007 according to the conventional clinicopathological criteria1 were reviewed. In accordance with Henri Mondor hospital research ethics committee, all patients gave written individual informed consent to participate in the study.
DM biopsies were classified into three groups: amyopathic DM (n=3, typical cutaneous rash with neither muscle weakness nor serum creatine kinase increase at presentation), early stage of mild DM (n=27, recent onset <2 months of both skin and muscle involvement and ischaemic changes restricted to one fascicle, usually consisting of mild perifascicular myofibre alterations) and chronic full-blown DM (n=10, chronic course of 3–12 months, and extensive ischaemic changes defined as abundant punched-out vacuoles, or microinfarcts, or conspicuous perifascicular atrophy affecting the majority of fascicles) (table 1).
Histological and immunohistochemical stainings
Unfixed 7 µm cryosections from samples stored at −80°C were routinely processed and subjected to automated immunohistochemistry (Ventana NEXES, Faulquemont, France) for endothelial cells (CD31/PECAM-1, 1:20, DAKO, Trappes, France; NoM0823), pericytes (NG2, 1:200, Chemicon Millipore, Molsheim, France; NoAB5320), myosatellite cells and regenerating myofibres (CD56/NCAM:1:100, Novocastra, Antony, France; NoNCL-CD56–1B6 or CD56/NKH-1:1:50, Coulter Clone, Villepinte, France; No6602705), macrophages (CD68/KP1/macrosialin, 1:300, DAKO, NoM0814), C5b-9/MAC (1:50, DAKO, NoM0777 or AbCam, Cambridge, UK; Noab55811), basement membranes (laminin-1:1:500, AbCam, Noab14055) and either fluorescent or peroxidase-coupled secondary antibodies with diaminobenzidine as chromogen (Ventana). Local production of reactive oxygen species was detected using antibody against 4-hydroxy-2-nonenal (HNE)-modified proteins (1:200, Calbiochem, Fontenay-sous-Bois, France; No393207).11 HNE is the most reactive lipid peroxidation product and detection of the protein adducts it forms in tissues attests local lipid peroxidation.12 Dystrophic muscles were used as positive controls.13
3D Reconstructions from serial sections to visualise the largest vessels
Vessels with relatively small lumen and thick wall and transversally oriented smooth muscle cells were considered as arteries; vessels with larger lumens and thinner walls and longitudinally oriented smooth muscle cells as veins. 3D reconstruction of normal muscle was done using 50–70 serial transverse 7 µm sections immunostained for CD31. Images (×10) acquired from each section of the series were processed in the align module of Axiovision4.2 (Zeiss, Pecq, France) to compensate for misalignment of vessels due to deformation during sectioning and acquisition. This could be reliably done for larger vessels, but not for capillaries. Aligned vessels were subjected to colour segmentation in the RGB colour space based on the dark-brown colouration of CD31+ structures.
3D Reconstructions of muscle treated to visualise small endomysial vessels
Endogenous alkaline phosphatase (ALP) activity, associated with pericytes and smooth muscle cells in capillaries and arterioles was detected as previously described.14 ,15 Muscle sections of 20 µm were permeabilised in 0.1% Triton X 100 in PBS (10 min), incubated (3 h) in 50 ml TRIS 0.1 M pH8.2 buffer containing 10 mg Naphtol AS-TR phosphate dissolved in 1ml of dimethyl formamide and incubated with 50 mg Fast RedTR salt (10 min) (Sigma-Aldrich, Lyon, France). Using a fluorescent Zeiss Axioplan2 microscope motorised in Z we specifically examined terminal arteries and capillaries. Optical stacks of 30–200 images (0.2–0.8 µm thick) were acquired by a set of rhodamine filters as Fast-Red product fluorescence is roughly similar to this fluorochrome. Stacks acquired in each channel were subjected to iterative deconvolution; vessel ‘growing volume’ segmentation, Fast-Red minus 4’,6-diamidino-2-phenylindole channel subtraction and 3D reconstruction, were done in Axiovision4.2.
Detection and estimate of focal capillary loss
Capillary loss was assessed by point-pattern analysis (PPA), a general method developed in astronomy, material science and ecology, and previously applied to muscle capillaries16 to analyse the distribution (and density) of objects represented as points with X,Y coordinates.17 ,18 Points, in our case, were CD31+ capillaries and the 2D space whole muscle biopsy reconstructions from 12–20 adjacent images (×20), using a Coolscope microscopy station (Nikon, Champigny-sur-Marne, France), and an analysis software (Olympus, Rungis, France). The X,Y coordinates of all capillaries (Axiovision4.2) were introduced in the PPA module of the ADE-4 software (http://pbil.univ-lyon1.fr/ade4/). A grid composed of 19–30 equidistant sampling points was superposed to the image, each grid point becoming the centre of virtual incremental concentric rings (15–20) in which the number of capillaries was counted by the software. The local density function calculated by the software, allowed determination of capillary-depleted area contours.
The theoretical number of capillaries that should have been found therein was determined by superposing the contours of depleted areas over areas of healthy muscle in which capillaries were counted (3–7 superpositions were done depending on specimen size). Capillary depletion zones belonging to adjacent fascicles were considered separately. The number of myofibres was counted in each depleted zone.
Counting of C5b9+ microvascular deposits
C5b9+ deposits forming clusters and entirely located within the same fascicle were counted. Large zones showing extensive immunolabelling involving the entire microvasculature were deliberately avoided.
Quantitative data were expressed as histograms. Cubic splines were used to highlight histogram peaks constructed in StatView V.5.0 (SAS Institute Inc).
Anatomy of human muscle microcirculation
Muscle microcirculation displayed stereotyped organisation. Serial histological sections of three normal human deltoid muscles allowed visualisation of feeder arteries in the epimysium wrapping the muscle (not shown). They continued into a network of arcade arterioles (diameter: 40–90 µm) in the perimysium (figure 1A). At regular intervals (200–300 µm) perimysial arcades gave off transverse arterioles bifurcating at nearly right angle to penetrate into the endomysium (figure 1B). Veins had a very similar organisation. Three dimensional reconstructions of transverse arterioles showed a trunk and several branches roughly perpendicular to the orientation of myofibres, with a mean of four bifurcations detected from the penetration of transverse arteries in the endomysium to the terminal arterioles (figure 1C). One additional fresh muscle sample was used for visualisation of terminal arterioles and capillaries by an histoenzymatic reaction detecting endogenous microvascular ALP activity as a red fluorescent signal. figure 1D reconstructed from 50×1 µm stacks is representative of the forked ending of one terminal vessel, presumably an arteriole, which schematic representation is shown figure 1E. The terminal arteriole ending feeding one microvascular unit (MVU), was typically surrounded by five myofibre sections (figure 1D). It was possible to identify, image after image, the beginning of each capillary forming this MVU (figure 1F shows images # 24 and 30 of the stack): the terminal arteriole ending had a feather duster-like appearance resolving into 6–8 capillaries (up to 10), the orientation of capillaries changing after one micrometer to adopt a longitudinal orientation (figure 1E).
The pattern of capillary loss and in DM
Muscle vessels can proliferate or regress in reaction to various stimuli,19–22 blurring elementary changes. Thus muscle morphometry was done in selected patients (detailed table 1). Capillary loss was deliberately assessed in patients with either ‘amyopathic’ DM, a paradigm of pure capillary loss,16 or early mild DM in which the single fascicle affected by myonecrosis was excluded from the analysis (figure 2).
Using a grid of equidistant points (figure 2A) and computer based analysis, patchy zones of capillary depletion were delineated (figure 2B,C). A total number of 219 capillary-depleted zones detected in 14 patients were studied. They corresponded to a theoretical number of 4–19 lost capillaries. Distribution histogram was multimodal (figure 2E). Cubic splines smoothing clearly showed at least two peaks at six capillaries (5–8 in the crude histogram) and 12 capillaries (10–12 in the crude histogram) corresponding to a 6-by-6 capillary drop-out (figure 2F). Consistently, distribution histogram of myofibres enclosed in capillary-depleted zones was also multimodal, each peak roughly corresponding to a multiple of the mean of five myofibres perfused by a terminal arteriole (figure 2E). Thus, whole MVU drop-out appeared as a prominent pattern of capillary depletion in early DM. Capillary loss often occurred in the absence of perifascicular atrophy, but, in full-blown DM, perifascicular atrophy was constantly associated with conspicuous capillary depletion.23
The C5b9 deposit pattern
C5b9 deposits often appeared as clusters (figure 2D). Although isolated C5b9+capillaries were commonly detected, distribution histogram of immunolabelled structures mainly showed one large peak at 5–8, and, possibly, much smaller ones at 10–12 and 18 (figure 2H). Thus, microvascular MAC deposits also roughly obeyed a MVU-like distribution. The main peak of C5b9 deposition was broader than that of capillary loss. Complement deposition was not restricted to endothelial cells as assessed by double immunostainings (figure 3A–F). C5b9+ microvascular cells were CD31+ endothelial cells (figure 3A) or CD31− periendothelial cells (figure 3B). Some of these periendothelial cells expressed the NG2 marker of pericytes (figure 3C). In addition, myosatellite cells, identified by the expression of CD56 or their location beneath the myofibre basement membranes were also occasionally decorated (figure 3D). Juxtavascular C5b9+ cells devoid of basement membrane, thus distinct from steady state pericytes and myosatellite cells, were also observed (figure 3E). These C5b9+ cells, which did not express the CD68 macrophage marker (figure 3F), were presumably activated fibroblasts or mesenchymal stem cells.
Other microvascular changes
Besides capillary loss and perivascular horseshoe-like C5b9 labelling (figure 4A), other microvascular changes were detected. In one full-blown DM case, conspicuous obliteration of an arcade artery associated with juxtavascular inflammation was detected (figure 4B). The lumen appeared severely narrowed by a non-inflammatory endoarterial fibrous thickening without media destruction (figure 4C). In full-blown DM, capillary depletion remained patchy, but the pattern was blurred by constant (10/10) intrication with neoangiogenesis (figure 4C) assessed by abundant transversally oriented capillaries.24 Neaoangiogenesis was associated (10/10) with the appearance of large microvessels with widely opened lumens and thick walls (figure 4D), with several layers of smooth muscle cells suggesting arteriolar-like transformation described in chronically ischaemic tissues.25 ,26 Oxidative stress was spotted by immunohistochemical detection of HNE-modified proteins.13 HNE immunopositivities were detected in 9/10 studied patients, including a clear microvascular pattern in four early DM, where HNE was expressed by transversally oriented endomysial vessels, presumably transverse arterioles (figure 4E), and by longitudinally oriented endomysial capillaries with a MVU-like pattern (figure 4F). In chronic DM, myofibres expressed HNE, mainly along sarcolemma, with enhancement in zones of perifascicular atrophy (figure 4G). Microinfarcted zone also expressed HNE (figure 4H). In severe DM cases, all cell types, including myofibres, vessels, fibroblasts and myosatellite cells, conspicuously expressed HNE (not shown), a finding consistent with extensive ischaemic myofibre damage assessed by NCAM expression (figure 4I).
We showed that the normal human muscle MVU is formed of 6–8 capillaries, and that early capillary loss in DM obeys a 6-by-6 rule, consistent with whole MVU drop out rather than random capillary damage.
Historic studies on muscle microcirculation were conducted in small animals (rats, hamsters or cats) using in vivo observations by transillumination, selective injection of opaque media, whole body perfusion and vascular casting.27–29 China ink injection studies in the rat have shown feeder arterioles in the epimysium continuing into a network of arcade arterioles in the perimysium.30 ,31 Interconnections allow for compensation if blood flow is compromised by occlusion of one arcade arteriole.32 Arcades feed transverse arterioles which penetrate into the endomysium, and divide asymmetrically to yield terminal arterioles which, in turn, resolve into capillaries. Transverse arterioles do not intercommunicate, and their occlusion could not be compensated for. Once formed, capillaries run in parallel to the muscle fibres to collect in venules situated up or down the length of the fibre.33 One MVU irrigates a cylinder of muscle tissue of 750–1000 µm in length in all animals studied.28 ,32 The MVU is the smallest functional unit for control of blood flow. It contains 7,6±2 capillaries in rats. In other small mammals, they comprise 6–8 capillaries located in-between three or four adjacent myofibres.34
The different techniques used to study muscle microvasculature in small animals cannot be applied to human muscle biopsy, thus explaining the lack of data on the human MVU in the literature. Herein, we used serial histological sections to visualise the three orders of larger vessels, and 3D images from serial optic stacks of thick sections of one fresh muscle treated to reveal ALP to visualise the terminal arteriole and its capillaries. Serial sectioning, confirmed the overall vascular organisation previously reported in rats and cats on the basis of intravascular ink injection.27 ,31 ,32 ,35 In the same way, optic stacks showed that human MVU organisation and size are strikingly similar to those of small animals,28 ,29 ,36 as represented in figure 1E. A single human post-mortem study using corrosion cast method in the temporal muscle does not provide data on the MVU size but shows an arteriole-to-venule distance of about 900 µm.37
Capillary loss in DM was present in both amyopathic and intact fascicles of early DM, pointing out capillary loss as an early hallmark of the disease.22 ,38 By reference to normal areas, point pattern analysis supported that capillary depletion in DM occurs in clusters and identified a distinct 6-by-6 loss pattern indicating that whole MVU drop-out is a prominent mean of capillary depletion. The multimodal curve of capillary-depleted myofibres supported this view. Microvascular C5-b9 deposits39 also tended to occur in roughly MVU-sized clusters, which was not suggestive of random complement activation.10 Moreover, in addition to endothelial cells, a variety of cell types, such as pericytes, neighbouring myosatellite cells16 and mesenchymal cells, were often decorated by C5b9 deposits. This makes antibody deposition directed against a cell-specific antigen or circulating immune complex deposition very unlikely mechanisms of complement activation in DM.10 The underlying mechanism of MVU depletion in DM remains unknown. Microthrombi distributed in a MVU-like fashion are exceptionally observed. Although thrombotic events are sufficient to activate the complement cascade40 and may be transient, the finding seems too rarely detected to represent by itself a firm explanation. It seem possible that changes in muscle perfusion originating in perimysial arcades around which inflammatory changes predominate in DM could contribute to whole MVU damage. In one chronic full blown DM case, obliterating arteriopathy was observed, suggesting some link with altered muscle perfusion, including hindered intercommunication of the arterial tree, and downstream damage of endomysial transverse and terminal arterioles. Although, arterial obliteration has not been noted as important in the classic literature on DM, lumenless perimysial vessels and statistical association of avascular perimysium with perifascicular atrophy were recently reported in DM.6 Arcade arteriole changes in DM clearly deserve further investigation. Chronic DM, showed typical combination of neoangiogenesis and capillary ‘arteriolisation’, that is, the replacement of normal muscle capillaries by microvessels with widely opened lumens and thickened walls and abundant smooth muscle cell in place of pericytes, as observed after arterial occlusion.25 ,26
It seems possible that altered perfusion regimen include ischaemic and reperfusion episodes in early DM. Ischaemia followed by reperfusion of the ischaemic region is known to lead to supplementary damage called ischaemia/reperfusion (I/R) injury which results from oxidative stress and causes microvascular alterations at the origin of the ‘no reflow’ phenomenon.41 ,42 Local production of reactive oxygen species inferred from the HNE labelling43 was detected in DM. It was observed in the microcirculation, notably at the level of transverse vessels and capillaries, and, to various extents in myofibers and other cell types in areas of ischaemic damage. Although not disease-specific, such an expression pattern seems consistent with I/R damage. Of note: (i) microvascular lesions are similar in DM and experimentally induced I/R, including endothelial abnormalities, thrombi, vessel hyperpermeability, and oedema42 (ii) endothelial tubulo-reticular inclusions typical of DM, may reflect endoplasmic reticulum stress which is inducible by both hypoxia or viral infections44 and may be assessed in DM by increased microvascular staining with the NADH-TR (nicotinamide adenine dinucleotide tetrazolium reductase) reaction routinely used in myopathology8 (iii) autoimmunity/inflammatory prone backgrounds enhance severity of I/R lesions, as demonstrated in the lupus mouse model.45 In addition, there is ample evidence that I/R injury also activates complement. It could do so either by the classical way, through fixation of natural low avidity IgMs on newly formed antigens such as non-muscular myosin in a variety of cells43 ,46 or by the alternative pathway47 or by the lectin pathway.48 Future studies aiming at substantiating experimentally the role of I/R in the pathophysiology of DM are mandatory, since a variety of anti-oxidant treatments are currently developed to limit I/R stress in organ transplant receivers,49 ,50 and might offer novel therapeutic approaches in DM.
Correction notice This article has been corrected since it was published Online First. Figures 1 and 2 have been updated.
Contributors CG contributed to the morphometric analyses, handled data and wrote the manuscript. EK interpreted double-immunostainings and pictures. PL provided antibodies and contributed to immunostainings. FJA collected clinical data and muscle biopsies. CC designed morphometric approaches, handled and interpreted data, performed histoenzymatic reaction and tissue imaging. RG conceptualised the study design, interpreted data and histological pictures and wrote the paper.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.