Article Text
PDF
Abstract
Objective Study the MRI signal of fatty lesions (FL) by immunohistological analysis of vertebral body biopsies of patients with ankylosing spondylitis (AS) compared with degenerative disc disease (DDD).
Methods Biopsies obtained during planned surgery from vertebral edges where MRI signals of FL was detected were stained with H&E. Immunofluorescence (IF) staining was performed to quantify osteoblasts and osteoclasts. Bone marrow (BM) composition, grade of cellularity and quantification of cells were analysed on six randomly chosen high-power fields (HPF; 0.125 mm2) at 200-fold magnification per patient by two experienced researchers in a blinded manner.
Results Biopsies of 21 patients with AS and 18 with DDD were analysed. Adipocytes were found in the BM of 19 patients with AS (90.5%) versus 5 with DDD (27.8%) (p<0.001), while inflammatory infiltrates were found in in the BM of 8 patients with AS (38.1%) versus 14 with DDD (77.8%) (p=0.035) and fibrosis in 6 patients with AS (28.6%) versus 4 with DDD (22.2%) (p=n .s.). The most frequently detected cells were adipocytes in AS (43.3%) versus DDD (16.1%, p=0.002) and inflammatory mononuclear cells in DDD (55%) versus AS (11.0%, p=0.001). Using IF staining, there was more osteoblastic than osteoclastic activity (6.9 vs 0.17 cells/HPF) in FL as compared with inflammatory BM (1.3 vs 7.4 cells/HPF), respectively.
Conclusion MRI FL correspond to presence of adipocytes, resulting to change of cellular homeostasis towards diminution of osteoclasts in the BM of patients with AS. The cross-talk between the different cell types and osteitis, fat and new bone formation needs further study.
- axial spondyloarthritis
- magnetic resonance imaging
- fatty lesions
- biopsy
Statistics from Altmetric.com
Key messages
What is already known about this subject?
Fatty lesions (FL), similar to bone marrow (BM) oedema, are characteristic MRI findings of patients with ankylosing spondylitis (AS) and degenerative disc disease.
FL are associated with syndesmophyte formation in AS.
What does this study add?
The presence of FL on MRI corresponds to fat deposition in the BM of patients with advanced AS.
AS seems to lead to a change in the BM microenvironment with local disruption of haematopoiesis and replacement by fat.
How might this impact on clinical practice or future developments?
These data add to our understanding of structural progression in patients with axial spondyloarthritis, indicating a possible ‘window of opportunity’ for earlier stages of the disease with respect to development of new bone formation.
Introduction
Ankylosing spondylitis (AS), the prototype of axial spondyloarthritis, is characterised by inflammation and new bone formation predominantly in the axial skeleton. The MRI signal corresponding to ‘inflammation’ in patients with AS is what has been internationally defined and named as bone marrow (BM) oedema or osteitis. In contrast, new bone formation in form of syndesmophytes and ankylosis are better detected by conventional radiography. Using MRI, an intermediate step between osteitis and bone formation in the form of tissue metaplasia to development of fatty lesions (FL) has been detected.
Immunohistological features of inflammation in form of osteitis have been studied in biopsies obtained from both sacroiliac joints1–3 and the spine,4 and infiltrates of different cell types and inflammatory cytokines and other mediators have been described.5 Furthermore, based on the association with human leucocyte antigen B27 and endoplasmic reticulum aminopeptidase 1, a role for the immune system in the pathogenesis of AS seems likely.6
As recently shown,7 8 new bone formation in AS is linked to both osteitis and also to the so-called ‘fatty lesions’—both of which are well detected by MRI.7 The exact sequence of pathological events leading to new bone formation is not really known because sequential MRIs have not been performed in observational studies in short sequence, but it seems clear that the combination of osteitis with a fat signal have the strongest predictive capacity for syndesmophyte formation later on.7 8 The most likely scenario for the sequence of events is that chronic spinal and entheseal inflammation leads to an alteration of cells of the longitudinal spinal ligaments.7 However, MRI derived fat signals in the endplates of vertebral bodies are also observed in acute and chronic stages of degenerative disc disease (DDD), as first described by Modic.9
While the anatomic background of osteitis has been nicely studied in rheumatoid arthritis,10 the anatomic nature of the fat signal detected by MRI has not been well studied to date. Current hypotheses on the underlying cellular processes of this signal are not based on histological analyses.
In the present study, we examined the areas with FL detected by MRI in the edges of vertebral bodies of patients with AS or DDD by histology. We aimed to compare the cellular composition and osteoblastic or osteoclastic activity that underlies this fatty signal in both diseases.
Methods
Patients with AS or DDD undergoing planned kyphosis correction surgery including anterior spinal osteotomy (in AS) or surgery to correct spinal stenosis (in DDD) were prospectively included into this biopsy study after giving informed consent for the use of the biopsy material. All patients underwent conventional radiographs and MRI examinations (T1 and T2 sequences, in some cases also short tau inversion recovery sequences) as part of preparation for surgery, of the area of the surgical intervention. Inclusion criterion for obtaining the biopsy was the occurrence of FL on MRI in the area of the planned surgery. Surgery was performed in a single centre and by the same experienced orthopaedic surgeon team (HB, AS, RB). Biopsies were obtained and documented under direct visual control via an anterior approach to the ankylosed spine from the area close to the vertebral edge, in which FL had been seen by MRI (figure 1A for AS and figure 1B for DDD). Biopsies were decalcified, embedded in paraffin, cut and stained by H&E by a separate investigator, blinded to patient‘s diagnosis. Four different compositions of the BM were differentiated: fat, inflammation, fibrosis and normal haematopoiesis (figure 2A–D). The marrow composition was analysed, and the cellularity was quantified as proportion of the surface area covered by the respective cell types. Six randomly chosen high-power fields (HPF; 0.125 mm2) at 200-fold magnification per patient were evaluated by two experienced researchers in a blinded manner. Finally, the histological results were compared with the MRI examinations of the area of the biopsy by a separate investigator. Furthermore, immunofluorescence (IF) staining was performed for quantification of osteoblastic and osteoclastic activity. Epitopes were retrieved from the formalin-fixed, paraffin-embedded axial spine sections of patients with AS using a heat-induced method. Briefly, sections were alternately bathed in boiling sodium citrate buffer (10 mM Sodium citrate, pH 6.0) or Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9.0). The slides were blocked with 5% bovine serum albumin and 2% horse serum in phosphate-buffered saline and stained with antibodies against Cav 1.2 (1:100 dilution; #NBP1-22439AF647, Novus Biologicals, USA) TRas osteoblast marker and tartrate-resistant acid phosphatase (1:100 dilution; ab191406, Abcam, UK). Alexa Fluor 555 or Alexa Fluor 647-conjugated antibodies (1:200 dilution; Thermo Fisher scientific, USA) were used as secondary antibodies targeted against the primary antibodies. Nuclei were counterstained using 4′,6-diamidino-2-phenylindole (Santa Cruz Biotechnology, USA). The staining was analysed using a Nikon Eclipse 80i microscope (Nikon, the Netherlands). Total bone as well as osteoblast and osteoclast surfaces were quantified in the fluorescence images using the ImageJ software V.1.52i (National Institutes of Health, Bethesda, Maryland, USA). Then, the ratio of osteoblast or osteoclast surface to the bone surface was calculated. For the IF stainings in the FL or inflammatory BM, three randomly chosen HPF (0.125 mm2) at 200-fold magnification per patient were evaluated by two experienced researchers in a blinded manner. The results are given as the mean number of cells per patient with the indicated SEM.
Example of sagittal and axial MRIs of a patient with AS (A) and a patient with DDD (B) who participated in the study. The arrowheads point the area of the vertebral body from where the respective biopsies were obtained, hyperintense signal indicating fatty lesions and ankylosis in AS or osteophyte in DDD. AS, ankylosing spondylitis; DDD, degenerative disc disease.
Characteristic images of the different compositions of the bone marrow that were differentiated from the obtained biopsies, in H&E staining and 100-fold magnification. (A) Fatty bone marrow with adipocytes as seen mainly in patients with AS, not displaying signs of damage related to their original cellular appearance. (B) Inflammatory bone marrow with lymphocytic cells as seen mainly in patients with DDD. (C) Fibrotic bone marrow, as seen less frequently in both patients with AS and DDD. (D) Bone marrow with normal haematopoiesis for comparison to the pathological findings as described in 2A–C. AS, ankylosing spondylitis; DDD, degenerative disc disease.
Descriptive data are presented as mean values when referring to quantitative variables and as percentages when referring semiquantitative measures. The Mann-Whitney U test was used to compare data between AS and DDD.
Results
Biopsies mostly obtained from the lower thoracic spine and the lumbar spine of 21 patients with AS (mean age 51.7±11.9 years, mean disease duration 24.6±14.0 years, 18 (86%) male) and from the lumbar spine of 18 patients with DDD (mean age 60.1±12.0 years, 5 (28%) male) were available. In the AS group, 12/21 (57.1%) patients were receiving non-steroidal anti-inflammatory drugs (NSAIDs) and 9/21 (42.9%) were receiving biologics which were all withdrawn two half-lives prior to surgery. In comparison, in the DDD group, 12/18 (66.7%) patients were receiving NSAIDs and six patients were not receiving any medication preoperatively. None of the patients in either indication was receiving any antiosteoporotic drug.
Differentiation of BM cellular composition on the patient level
The histological appearance of the sites with FL on MRI was different between the groups, with fat marrow being present in biopsies of 19 patients with AS (90.5%) but in only 5 with DDD (27.8%) (p<0.001). In contrast, inflammatory marrow changes resembling mononuclear infiltrates were seen in 8 patients with AS (38.1%) and 14 with DDD (77.8%) (p=0.035), all of which had appearance of osteitis concomitant to FL on MRI. Finally, histological findings of marrow fibrosis were found in 6 patients with AS (28.6%) and 4 with DDD (22.2%) (p=n.s.), also all of which corresponded to findings with sclerosis concomitant to FL on MRI (table 1).
Prevalence of each subtype of lesions (fat alone, fat with inflammatory findings, fat with sclerosis) as seen on MRI (A) and in the biopsies (B), separated by diagnosis subgroup (AS and DDD)
Differentiation of BM cellular composition based on the affected BM area
In the semiquantitative analysis, the mean distribution of the various BM tissue types in the biopsies differed between AS vs DDD, with 43.3% vs 16.1% (p=0.002) of the biopsy surface being covered by complete and well defined adipocytes, while 11.0% vs 55.0% (p=0.001) of the biopsy surface was covered by inflammatory marrow and 9.0% vs 12.8% (p=0.890) by fibrotic marrow, respectively.
Quantification of osteoblastic and osteoclastic activity
Using immunofluorescence staining, a clear distinction was found between areas with that showed FL versus areas that showed inflammatory BM on T1-weighted MRI sequences. While in the areas with FL a high number of osteoblasts was counted (6.917±1.066 cells/HPF), osteoclastic activity was almost not present (0.1667±0.09623 cells/HPF), p<0.001 (figure 3A). In contrast, the opposite result was found in the areas with inflammatory BM, where significantly more osteoclasts were counted (7.375±1.599 cells/HPF), as compared with a minimal count of osteoblastic cells (1.292±0.3359 cells/HPF), p<0.001 (figure 3B). Similar results were seen in the calculation of surface covered by osteoblasts (figure 4A) or osteoclasts (figure 4B) in relation to the entire surface of the analysed bone biopsy. The relation of osteoclasts vs osteoblasts to the examined bone surface was 0.480%±0.289% vs 9.570%±1.3925% in the areas of FL, while the reverse result was seen in the areas of inflammatory lesions, with 11.900%±0.297% vs 1.980%±0.440%, respectively (all comparisons p<0.001).
Quantification of osteoblastic and osteoclastic activity using immunofluorescence staining. (A) Results from areas with fatty bone marrow and significantly higher content of osteoblasts as compared with almost no osteoclasts. (B) Results from areas with inflammatory bone marrow with significantly higher content of osteoclasts as compared with a minimal count of osteoblasts. ***P value <0.001 for comparison of the cell count.
Quantification of osteoclast (A) and osteoblast (B) surface in relation to bone surface in areas of fatty and inflammatory lesions (detected on MRI). BS, bone surface; OB, osteoblasts; OC, osteoclasts. ***P<0.001.
Discussion
This is, to our knowledge, the first study to analyse the cellular composition of FL as detected by MRI in the vertebral edges of patients with AS, as compared with similar MRI lesions from patients suffering from a common diagnosis causing back pain in the general population, such as DDD. For our analyses, we used biopsy material obtained by spinal surgery from the anterior part of the vertebral bodies, which is the region of the spine that is known to be most commonly affected by osteophyte formation (syndesmophytes in AS or spondylophytes in DDD).
Our results suggest that the histological composition of the hyperintense signal seen on T1-weighted MRI sequences of the spine in patients with AS and which may be considered as a step prior to formation of syndesmophytes7 consists of adipocytes which do not display signs of damage related to their original cellular appearance. In contrast, in patients with DDD, we found a higher frequency of BM with signs of inflammation and significantly less adipocytes despite an appearance of the MRI signal (hyperintensity on T1-weighted MRI) that is largely similar to what was seen in AS. A possible explanation for these observations may be that in the biopsies obtained from patients with AS, the bone quality was more intact due to the local ankylosis and stability of the examined areas, which may have preserved the local cellular composition (local tissue metaplasia to an FL, following an earlier BM oedema). In patients with DDD, the instability of the respective segments caused by the disc degeneration may have led to local stress and microfractures of the bone, consequently leading to damage of the local cellular composition including adipocytes and extrusion of fatty acids. The latter may still be depictable as fat signal by T1-weighted MRI sequences, but the origin of this signal is not from adipocytes in the area of the biopsy. Similarly, the ‘inflammatory’ lymphocytic BM observed in patients with DDD does most probably also originate from microfractures occurring in patients with DDD but not in those with AS because of the established ankylosis. Furthermore, cell proliferation is also representing a well-described phenomenon in different stages of the osteoarthritic process.11 Hence, the presence of inflammatory cells in the BM of patients with DDD may be explained by (1) mechanical stress (these patients had a significant burden of pain, which was also a reason to undergo spinal surgery) and (2) local osteoporotic processes may be expected in addition to degenerative changes in these relatively old patients with DDD.
Finally, it is of interest that we did not detect an increased signal of fibrotic BM neither in the number of patients nor in the mean proportion of the area of the biopsies in the patients examined, irrespective of the diagnosis. These findings are in contrast to the sequence proposed about a decade ago,12 suggesting that a dense infiltration of mononuclear cells accompanied by high microvessel density and osteoclasts are destroying the bone structure in a first step, being followed by formation of a fibrous repair tissue with fibroblasts with remaining high microvessel density, a few mononuclear cells and a decreased density of osteoclasts and later substitution of the fibrous tissue by new bone formation with high osteoblast activity. Other observations from biopsies from facet joints of patients with AS have revealed the occurrence of parallel findings of granulation tissue and fat tissue in patients with AS, showing that in this region of the spinal skeleton, replacement of the subchondral BM by fat tissue is not related to progressive joint remodeling,13 something that might be due to the different mechanical load in the area of the facet joints or in areas of pseudarthrosis due to trauma,14 as compared with the anterior part of the spine which was examined in our study. However, it needs to be taken into account that these pathological processes may differ according to the disease stage.
The second important finding of our study are the immunohistological results suggesting that a major reason for the occurrence of new bone formation in AS may be a disturbance of the homeostasis between osteoblasts and osteoclasts at areas of FL (figure 5). This may be especially true for the formation of syndesmophytes rather than spondylophytes. In those areas where adipocytes and the respective FL detected on T1-weighted MRI sequences were found, we observed almost no osteoclasts (figure 5). Together with the findings described above, our data may suggest that the during progression from an inflammatory lesion to an FL, a degradation of local osteoclasts may be promoted leading to a relative overexpression of osteoblasts which finally leads to syndesmophyte formation and ankylosis. Of interest, this is in line with previously published data obtained using positron emission tomography, in which a high uptake of 18F-labelled fluoride, a tracer that is highly specific for osteoblastic activity, had been observed in areas where both BM oedema and FL were detected.15
Example of a microscopic image of a biopsy taken from a patient with ankylosing spondylitis. (A) Overview (40-fold magnification) of an area with both fatty (**) and inflammatory (*) bone marrow patterns. (B) 200-fold magnification of one of the areas with both inflammatory and fatty bone marrow. In the areas with inflammatory bone marrow, only osteoclasts are stained (arrowheads) and visualised in red in TRAP/DAPI immunofluorescence staining (C; 400-fold magnification). In contrast, in the areas with fatty bone marrow, no osteoclasts were seen, while osteoblasts were visualised using Cav/DAPI immunofluorescence staining (D; 400-fold magnification). DAPI, 4′,6-diamidino-2-phenylindole; TRAP, tartrate-resistant acid phosphatase.
The finding of a higher number of osteoclasts in comparison to osteoblasts in the biopsies containing inflammatory BM in patients with DDD can be explained by osteoporotic processes induced locally by the inflammation.
Our study does also have some limitations. Therefore, it needs to be stressed that we are only dealing with biopsies of patients with AS with longstanding disease—this may well be the reason for the low prevalence of inflammatory cells in the biopsies of these patients. Furthermore, the biopsies analysed were mainly obtained from areas identified as showing FL by MRI prior to surgery—thus, not allowing for analysis of tissue obtained from other areas. In line with that, many syndesmophytes were also found to develop in areas with no initial osteitis or FL on MRI.7 In our study, such analyses were not possible because the surgery performed was based on a clinical indication and not for the purposes of a study. Finally, in the absence of appropriate samples we could not compare our biopsy findings with serum levels of biomarkers related to new bone formation such as Dickkopf-116 16 sclerostin,17 leptin or high molecular weight adiponectin,18 as previously described.
In conclusion, our data from biopsies obtained by spinal surgery suggest that FL on T1-weighted MRI sequences of patients with AS are related to a high content of adipocytes. This does not seem to be the same in patients with DDD, where FL on T1-weighted MRI sequences had less adipocytes but more inflammatory BM. The ‘early stage’ of inflammatory activity in AS may well represent a ‘window of opportunity’ for treatment where osteoblasts and osteoclasts are still in a relative homeostasis. However, once FL with adipocytes develop in a ‘later stage’ of the disease, degradation of osteoclasts and a relative overexpression of osteoblasts may initiate new bone formation.
Acknowledgments
We would like to thank the patients for agreeing to participate in this study.
References
Footnotes
Handling editor Josef S Smolen
Contributors XB and JB: concept, coordination of the study, data analysis, data interpretation, writing of the manuscript. HB, RB and AS: collection of biopsies, data interpretation and editing of the manuscript. GS, ML and AR: immunohistological analyses, data interpretation and editing of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval The study was approved by the Ethical Committee of the Ruhr-University Bochum, Germany (Reg. Nr. 4360-12).
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement The exact steps of all analyses, including the tissue analyses, are included in the manuscript. For additional information, please contact the corresponding author.