Granulomatous vasculitis and persistent measles virus infection in Crohn ' s disease

AJ WAKEFIELD. Granulomatous vasculitis and persistent measles virus infec, tion in Crohn's disease. Can J Gastroenterol 1994;8(2):70,74. Based upon the recent observation of vasculitis in Crohn's disease, a process that is more widespread than was recognized previously, the author investigated the possibil, ity that this mechanism may provide an explanation for some of the clinical and histological idiosyncracies of this condition. In addition, it was suggested that the mesenteric microvascular endothelium may be a source of the persistent antigen that is responsible for ongoing cellular immunity in Crohn's disease. This review discusses some of the studies designed to test these hypotheses, and discusses recent evidence for the presence of a measles,like virus in the endothelium in inflammatory foci, which may be relevant in the etiology of Crohn's disease.

T HE PREVAILJNG PERCEPTION OF Crohn's disease (CD) as a primary mucosa!pathology, either as an epithelial abnormality or an aberrant response of the mucosa!immune system, fails to explain many of the idiosyncracies of this condition.ln the absence of a plausible pathogenic basis for skip lesions, transmural inflammatio n, panenteric distribution of disease and anastomotic recurrence, and the deleterious effects of smoking, many have avoided these issues in favour of more elaborate hypotheses.
It is reasonable to assume that CD is characteri zed by a persistent cell-mediated immune response to an, as yet, unidentified antigen or antigens (1 ).T he capacity of cells both within the lamina propria and stimulated epithelial cells to present class 11 antigens (2) has, in the mi nds of many, reasserted the central role of the mucosa in the development of CD.The abundance of luminal antigen seems to suggest that we need go no further in identify ing the source of the stimulus for persisting cellular immunity; all that remains is to isolate and characterize the antigens responsible.
Before taking this final step into the darkness of the intestinal lumen, we should reflect upon CD's phe norypic idiosyncracics when conside ring the underlying pathogenic mechanismwithout a mechanism and, by extension, an appropriate source of persistent antigen, we are not in a position to seek the nature and origin of that antigen.
C lues as to the tissue localizati on of this putative antigen are present in CD in the form of the granuloma.This reaction localizes co persistent antigen, and in those diseases in wh ic h an e tiological agent has been ide ntified (including tuberculosis, pa ratubcrculosis and schistosomiasis), the granu loma localizes to the causati ve antigen.Therefore the granuloma provides a logical starting point when considering both the mechanism and the cause of CD.

TISSUE RELATIONSHIPS OF THE GRANULOMA IN CD
The gra nu loma is an early histo logical fea ture in the evolutio n of intestinal inflammation in CD; these hallma rk lesions occur principally in a submucosal location and can be present prior co mucosa[ inflammation or ulceration (3-5), an obse rva tion that argues strongly for the priority of this lesion in CD evolution.Although described previously, granulomatous vasculitis in CD has been reported as an occasional fea ture of no pathological significance (6,7).
For a variety of reasons, the relationship of the vasculature to foci of inflammation, such as granulomas, may be obscured in immersion-fi xed, he matoxy lin a nd cosin stained t issue sections.First, vessels coll apse in th e absence of an intraluminal hydrostatic pressure.
Furthermore, advanced granu lomatous vasculitis could be expected to destroy the vessel and to occlude the lumen of the affected vessels, obscuring the apparent vascular o rigin of the lesion.This is suggested by the observation chat extensive vascula r disruption in 'primary' granulomatous vascu litides can make it difficult to identify che vesse l whe rein the granuloma origi nated (8).
T he aut hor has recently examined 485 granulomas in L5 specimens of resected CD a nd shown that at least 85% of granulomas are directly involved in vascular injury (5).These stud ies we re facilitated greatly by the combination of perfusion-fixation and immunosta ining fo r vascular structures.G ranu lomas were seen to a rise from within the walls of blood vessels and, in many cases, from within the vessel lumen, in intimate contact with the vascular endothe lium (Figure 1 ), suggesting the presence of a previously unrecognized macrophage-endothe lial cell interaction in CD.In addition , compute rized three-dimensional reconstruction of serial sections of CD h as confirmed that granulomas follow the course of the affected blood vessels, rathe r than coinc identally being adj acent to these vessels in a single plane of section (5 ).Granulomatous vasculitis -and venu litis in particular -is an integral pa rt of the spectrum of vascular inflammation that rhe autho r beli eves is central to the pathogenesis of CD. S imilar observations in sa rco idosis suggest that this process may be more common than was t hought previously (9).The consequence of this vascular lesion in CD is likely to be isch emia of the dependent tissues.In te rms of a mechanism, the refore, can t his process explain some of the idiosyncracies of CD?

SKIP LESIONS AND MICROVASCULAR INJURY
Discontinuous inflammation and associated ulce ration are characterist ic features of CD; inflamed, ulcerated mucosa may occur immediately adjacent to mucosa of norma l histo logical appearances (10,11).]n an attempt to expla in this phenomenon by extrapolation of the vascula r hypothesis, the au thor's group selectively embolizcd the submucosal microvascular plexus of ferret intestine ( L2).Microsphe res (27 to 90 µm in diameter) were injected in to the arteria l arcade of a defined loop of intestine, causing focal inflammation a nd ulce ratio n 24 to 72 h later.This model did no t set out to reproduce the ch ronic damage of CD but it did reprod uce the juxtaposition of inflamed and ulce rated mucosa a longside histologically normal mucosa .In CD, repeated episodes of focal submucosal vascular occlusion could cause the ch aracterisric patchy mucosa[ damage.The grn,mloma adheres to the endothelium and iltere is no generalized inflammaiion associated with this lesion.l-lema1oxyli11 and eosin X400

ANASTOMOTIC RECURRENCE OF CD
The a uthor's group believes that occ ul t residua l granulomarous vascu li cis is the mechanism for anastomotic recurrence.The capacity of the bowel to h eal after a focal, intramura l vascu lar injury depends largely upon the integrity of the collateral plexus fo rmed by the submucosal microvasculature (12).Thus, for any point in the bowel wall, submucosal collateral vessels from e ither side may preserve the c irculation to chat point, but division of the bowel before anastomosis will halve this collateral suppl y.Tissue pe1fusion at an anastomosis will be compromised further by the local pressure of a suture line.T h is devascularization is of little consequence when normal intestine is a nastomosed.If, however, the re is a vascu litis in the macroscopically norma l bowel used to make the a nastomosis, the reduced capacity of the a lready compromised col late ra l circulation may be sufficient to induce focal tissue ischemia and anastomotic recurrence.S imi larl y, subacute obstruction of the bowel, by increasing intraliminal pressure, will cause a profound decrease in perfusion of the mucosa[ vascu lature ( 13).
Osborne et a l ( L 4) recently adapted the model described by Hudson (L2) to study anastomotic heali ng in the presence of a compromised submucosal collateral ci rculation.Following emboli za- Crohn's disease.Focus of granulomatous mucosa!inflammation.Nuclear immunosraining with measles virus polyclonal antibody (x/ 000) Figure 3) Crolm's disease.Submucosal lymJ1hoid aggregate.In situ hybridization for measles virus genomic RNA.Characteristic nucleolar pattern of hybridi zation (39) is seen (X400) tion of the intestine, which produced a self-limiting mucosa!injury, an anastomosis was created through the e mbo lized loop of intestine 72 h fo llowing the initial procedure.At two weeks there was macroscopic and microscopic inflammation in the embolized limb of the anastomosis with a combination of transmural inflammation, mucosa!ulceration and occasional granuloma formation; these fea tures were not seen in controls who received e ithe r anastomosis or microsphe res alone, and the resul ts suggest that the superimposition of two self-limiting ischemic insults can produce c hron ic inflammation at an anastomosis.

COLONIC CD: DIVERSION OF THE FECAL STREAM AND RESTING THE GUT
Patien ts with severe colonic CD often enjoy a remission fo llowing a diverting ileostomy, but relapse of the d isease usually follows restoration of the fecal strea m (1 5-17).This sequence has prompted the hypothesis that some constituent of the feca l stream is responsible for either initiating or provoking CD ( 17 ,l 8 ).
T he bloodflow of t he intestinal mucosa increases greatly in response to the metabolic demand placed on the mucosa by digestion and absorption ( 19 ).In CD, stenotic vascular lesions in the bowel wall would place a rate-limiting influence upon the hyperemic response to mucosa!activity.T his situation is analogous to intermitte nt claudication or exercise-induced a ngina -ischemia or infarction will occur when t he demand fo r blood exceeds supply.
Fecal diversion rests the excluded bowel, helping reduce mucosa!bloodflow and oxygen demand.Provided with a respite, the intestinal ischcmic damage eases and healing dominates.T his phenomenon may also help to explain, at least in part, the a meliorating effect of a proximal di verting ileostomy upon a nastomotic recurrence (20).However, reinstitution of the fecal scream is li kely to induce relapse due to further ischemic damage.This may also explain the transient clinical benefit of an elemental or polymeric diet (21).

MULTIFOCAL GASTROINTESTINAL INFARCTION AND APHTHOID ULCERATION
It h as been argued that aphthoid ulceration in CD is n ot a feature of mucosa!ischemia (22) .However, in cases of Wegener's granulomatosis or Beh -~et's d isease, both of which are idiopat hic necrotizing vascu li tides involv ing the intestine, aphthoid ulceration of the mucosa (including the oral mucosa) is a prominent featu re (23,24 ).It is likely that these diseases, which produce occlusive vascula r inflammation in affected segments of intestine, produce focal ulce ration by causing mucosa!ischemia; it is inte resting that anastomotic recurrence is also a feature of Behc;:et's disease (24 ).The authors group has proposed that ulceration in CD represents focal infa rction of the mucosa that may progress, depend ing on the location a nd severity of the underlying vascular d isease, to penetrating ulceration and fistulation (25).To :c1ccept that this da mage is due to 'multifocal gastrointestinal infarction' needs a reappraisal of the concepts of infarction (22).'Classical' intestinal infa rc tion involves fibrosis, mucosa!atroph y and hemosiderin depositionall features of large mesente ric vessel occlusion.But this definition precludes a spectrum of tissue injury t hat may result fro m c hronic damage at the level of the microvasculature of the bowel wall.Infarction of the mucosa will not be immediately apparent, not only because of the bowel's abi lity co shed dead tissue into the lumen, but also because of its remarkable capacity for repair and regeneration without sequelae.Mucosa!ulcera tion may be seen in a variety of other conditions in which ischemi a is implicated -for example, pseudomembranous colitis, pigbel and necrot mng enterocolitis (26)(27)(28).These appearances are not those of classical infarction but, if infarction is defined as 'tissue necrosis due to in terruption of its blood suppl y', t hese cond itions qualify and so, the author suggests, does CD.
By examining macroscopicall y normal, non infla med areas of bowel in patients with CD, Sankey and colleagues (29) ic.lent1fied subtle mucosa!changes that woulc.lnormally be obscurec.lin areas of acttve inflammation.Using a combinauon of perfusion-fixation at mean arterial pressure anc.l immuno-h1stochem1cal techniques, she was able to examine small mucosa( capillaries in c.letail.Focal, early mucosa!changes appear to be associated with damage to small capillaries.This was manifest as <lisruption of the capillary basement membranes or the vascular endothelium with hemorrhage and trails of fibrinogcn in the surrounding lamina propria.These were very much more common in patients with CD (50%) than with ulcerative colitis ( 15%) or in controls ( 15%).This vascu lar c.lamage prccedec.lboth the focal accumulation of inflammatory cells and necrosis of the overlying epithelium anc.l cannot, therefore, be a consequence of inflammation.
The mucosa( changes described by Sankey ct al occurrcc.lat all levels of the mucosa, hoth in areas close to, and distant from, Peyer's patches.The relation between these mucosa!changes and the classical 'aphthoid' ulcers occurring over Peyer's patches is unclear.l11c centres of lymphoid follicles are supplied by end arterioles.le has been suggested that as a consequence, they are panicularly prone to ischemia (30); thus, Peycr's patches may provic.lea favoured site for this kind of ulceration.Altemattvely, many 'aphchoid' ulcers overlie lymphoid aggregates rather than true follicles; it is possible that progression and extension of the early features that were described would be compatible with the ultimate development of an 'aphchoid' ulcer.

SYSTEMIC THROMBOGENESIS, SMOKING AND CD
Independent risk factors for systemic thrombogcnesis and an increased incidence of thrombotic disease include smoking, elevated plasma levels ofl ipoprotein(a) (31) anc.l factor V ile (32).The author's group recently showed that only six of 49 patients with CD die.I nm have one or more of these additional risk factors for thrombosis (unpublishcc.ldata).It is tempting to postulate chat a primary mesenteric microvascular injury anc.l a systemic thrombotic tendency combine to precipitate clinically acnve en.

MESENTERIC MICRO VASCULAR ENDOTHELIUM AND MEASLES VIRUS
The observation of granulomatous vasculttis in CD, and the possibility chat many of the idiosyncrac1es of this condition might be explainec.lby a compromised intramural circulation, focused attentton upon the mesenceric microvasculature.Vascular endothelium plays a central role in both immunity and hemoscasis (33 ); furthermore, it can harbour both productive and persistent viral infections (34,35).There was sufficient epidemiological evidence of an i nfcccious etiology for CD to progress this hypothesis in search of a source for persistent antigen.Since CD is neither vertically nor horizontally transmissible, it is likely chat the putative infection is nonproductive and therefore cell-associated.If so, is it possible that the microvascular endothelium of the gur is the source of chis infection and, thus, the persistent antigenic stimulus chat drives cellular immunity in CD?
It was hypothesized that measles virus could be the etiological agent in CD.Measles virus-infected lymphocytes 'home' to the lymphoid tissues of the gut during the primary vircmia; subsequently, the virus infects the microvascular endothelium of these tissues and induces vasculiris-secondary transmural lymphoid follicles develop in the gut wall and there is necrosis of the overlying epithe lium (36), all of which are characteristic features of CD (3,4, l I).Thus, measles can produce Koplik's spots throughout the gut that are similar to the aphthoid ulcers characteristic of CD (36)(37)(38).Giant cell syncycia at the centre of secondary lymphoid follicles in the intestine are a prominent featu re of measles virus infection (36).Hadfield (3) reported similar changes in his classical hisco-log1cal description of CD.Acute endothelial cell mfection by measles virus occurs principally in the submucosal microvasculature throughout the entire CAN j GA~"Tl\OENTEROL VOi. 8 No 2 MARCIi/APRiL 1994 Granulomatous vasculitis and measles virus le ngth of the gascroincescinal trace (36), the lesional topography of CD (3,5).
Recent derailed analysis of 24 cases of intestine affected by CD using a combination ultrastructure, immunoh1stochemistry (Figure 2) anc.l in situ hybridization (Figure 3) suggests that measles virus may play an etiological role in this condition (38,39), and that intestinal microvascular endothelial cells may be the principal reservoir of infection (infected cells were often observed specifically in foci of granulomacous vascular injury).A profound cellular immune response to apparently low levels of endothelial cell-associated virus, with rhe consequent vascular compromise chat is seen, might account for the extensive tissue injury in CD.Identical intranuclear paramyxovirus inclusions in CD have been observed recently by others ( 40).There is an urgent need co confirm these observations at a molecular level and provic.leamplificants for viral sequencing studies.
A possible role for measles virus in CD has received recent support from epidemiological studies undertaken in Sweden: Ekbom et al ( 41) observed chat perinatal measles virus infection was a strong risk factor for the development of CD lacer in life, a phenomenon chat exhibits similarities co subacute sclerosing panencephalicis.

CONCLUSIONS
In early nonulceraced CD there is evidence of granulomarous and lymphocyric vasculitis: the combmation of chis lesion with either a systemic prothrombotic environment or mechanical devasculanzation of the intestine, before anastomosis, may induce mucosa( ischemia anc.l ulceration.We hypothesised chat the mesenteric microvascular endothelium may harbour the antigenic stimulus for vasculitis in CD, and early morphological and epic.lcmiologicaldata suggest char this stimulus may be persistent measles virus 111- fection.A great deal more work needs to be undertaken ro confirm or refute these early ettological observations before any firm conclusions can be reached.

CAN J GASTROENTEROL VOL 8
Figure l)Crolm'sdisease: incraluminalgranuloma in a perfusion-fixed submucosal vessel.The grn,mloma adheres to the endothelium and iltere is no generalized inflammaiion associated with this lesion.l-lema1oxyli11 and eosin X400