Serum levels of matrix metalloproteinase 3 and macrophage colony-stimulating factor 1 correlate with disease activity in ankylosing spondylitis




To assess the usefulness of measuring serum matrix metalloproteinase 3 (MMP-3) and macrophage colony-stimulating factor (M-CSF) in patients with ankylosing spondylitis (AS).


Serum levels of MMP-3 and M-CSF were measured in AS patients who did and did not receive infliximab treatment. These were compared with those of 28 healthy subjects.


In the group of AS patients not treated with biologics, both M-CSF and MMP-3 correlated with the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) values, but not with each other. Logistic regression analysis showed that MMP-3 values were high in those with severely active disease. Infusions of infliximab in AS patients led to a significant decrease in the values of the BASDAI as well as the serum MMP-3, but no change in the serum M-CSF values.


MMP-3 and M-CSF are potentially useful markers of AS disease activity.


Despite decades of intense research, the molecular pathways that mediate the spondylarthropathies are still completely unknown. Although HLA–B27 is a contributing factor, how it initiates arthritis remains controversial. No universally accepted hypothesis is available to allow investigators to pursue an infallible hypothesis-driven approach. What is clear is that there are multiple genetic factors, and there must be many more critical factors other than HLA–B27 (1). Hence, identifying these factors would be critical to this field of research. Since the identities of some of these factors are probably beyond our imagination, one reasonable approach is to use microarrays to randomly screen for genes differentially expressed in spondylarthropathy (SpA) tissues. In this article, we use a 1,176 gene microarray to study SpA synovial tissues.

There are two challenges in using microarrays for this purpose. The first is to be able to follow stringent guidelines for reproducibility so that the results are not simply a massive amount of uninterpretable information. In this article, we have set up strict criteria for selection of candidate transcripts, taking into account variability among duplicate assays, power analysis to predict sample sizes, comparison of variance for appropriate use of parametric or nonparametric analysis, and statistical correction when comparing multiple genes. By employing these strategies, we identified 3 candidates in our microarray results: macrophage colony-stimulating factor (M-CSF), matrix metalloproteinase 3 (MMP-3), and interleukin 7 (IL-7).

The second challenge in microarray is that tissue samples consist of heterogeneous cell types, so that the above-identified genes might appear to be highly expressed only because certain cell types are preferentially enriched. To ensure that the candidates uncovered by microarray are participants in the disease processes, we arbitrarily postulate that any observed microarray transcripts will be acceptable as candidates only if the serum concentrations of the corresponding proteins correlate with the disease activity index. In this way, we bypass even the step of verification of expression of candidate transcripts by reverse transcriptase–polymerase chain reaction (PCR). This is because, regardless of whether they are false positives of the microarrays, such candidates are still promising targets for future research. Equally important, their serum concentrations might also help us to assess SpA disease activity.

We collected 2 sets of serum samples for this purpose. The first set was derived from 41 ankylosing spondylitis (AS) patients in Beijing. The characteristics of these patients have been carefully recorded, including scores for Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) (2) and Bath Ankylosing Spondylitis Functional Index (BASFI) (3). Using these samples, we tested if serum levels of MMP-3 and M-CSF would correlate with the BASDAI values. This was indeed the case. The second set of serum samples consists of 13 of 21 previously reported AS patients in Edmonton, Canada (4). These are paired samples from patients before and 14 weeks after initiation of infliximab infusions. Infliximab is an anti-tumor necrosis factor (anti-TNF) antibody that, in the majority of AS patients, provides suppression of signs and symptoms (5, 6). We postulated that if our observed genes were relevant to the disease, the serum concentrations of the encoded proteins would be suppressed by infliximab infusions. This is indeed the case with MMP-3.


Clinical materials.

For microarrays, synovial biopsy samples were obtained from 11 SpA patients during needle arthroscopy, as reported previously (7). Demographics of these patients are summarized in Table 1. Peripheral blood mononuclear cells (PBMCs) of 10 healthy subjects (age 31–59 years, 3 female, 7 male) provided microarray controls. Six of these healthy subjects were white and 4 were Asian. For an enzyme-linked immunosorbent assay (ELISA), serum samples were obtained from AS patients: 41 from Beijing and 13 pair samples from Edmonton. For the Edmonton patients, serum samples were collected before and 14 weeks after the start of infliximab infusions. Infliximab was administered at 3 mg/kg at weeks 0, 2, and 6 (4). The Beijing patients had not been treated with any anti-TNF agents or thalidomide or methotrexate. Twenty-eight healthy subjects from Beijing (age 15–49 years, 14 female, 14 male) provided control serum samples. Synovial fluid samples were obtained by arthrocentesis from 15 SpA patients with synovial effusions. Eight of these patients were diagnosed as having AS, 6 as having undifferentiated spondylarthropathy (USpA), and 1 as having reactive arthritis. These procedures and their consent forms were approved by the respective institutes. Diagnosis of AS and SpA followed standard criteria (8).

Table 1. Characteristics of patients providing synovial biopsy samples*
Designation no.SexAge, yearsDiagnosisHLA–B27
  • *

    uSpA = undifferentiated spondylarthropathy; Pos = positive; AS = ankylosing spondylitis; PsA = psoriatic arthritis; Neg = negative; SpA = spondylarthropathy; IBD = inflammatory bowel disease.

5F18SpA with IBDPos


The methods for microarray assay followed those provided by BD Biosciences Clontech (Palo Alto, CA) and have been reported previously (8). Briefly, RNA was extracted from tissue and cell samples by solution D (Strategene, LaJolla, CA) and purified by repeated phenol:chloroform reextractions. RNA (150 ng) was then amplified by SMART-PCR (BD Biosciences Clontech), the cycle numbers being monitored to avoid amplification beyond the exponential phase. Amplified complementary DNA (cDNA) was labeled with 33P-dCTP using rediprime II random prime labeling system (Amersham, Piscataway, NJ) and hybridized to a cDNA-based nylon membrane (Atlas Human 1.2 array; BD Biosciences Clontech). After exposure to a phosphor screen, signals were scanned and recorded with a STORM scanner (Molecular Dynamics, Sunnyvale, CA). Signal intensities were computed by the AtlasImage 1.5 software (BD Biosciences Clontech). For each gene, local background intensity value was subtracted from the gross intensity value. When results of one membrane were compared with those of another, the intensities of the genes in both membranes were normalized to the mean intensities of glyceraldehyde-3-phosphate dehydrogenase. Calculation of threshold and statistical evaluation will be described in the Results section.


ELISA kits were purchased from commercial sources: Human M-CSF Immunoassay (R&D Systems, Minneapolis, MN), Human Matrix Metalloproteinase-3 Biotrak ELISA system (Amersham), Human IL-7 ELISA kit (Cell Sciences, Norwood, MA), and Human Tissue Inhibitor of Metalloproteinase-2 BioTrak ELISA system (Amersham). All assays were carried out in duplicate and followed the manufacturers' recommendations. The MMP-3 ELISA kit measures total MMP-3 including pro-MMP-3, active MMP-3, and MMP-3/tissue inhibitor of metalloproteinase (TIMP) complexes. The M-CSF kit utilizes recombinant 158 amino-terminal amino acid residues of the extracellular domain of native M-CSF.


Statistical evaluations were carried out with Microsoft (Redmond, WA) Excel and the XLStat software (available at URL:


Results of microarrays.

In the first preliminary experiment, the variability of the microarray technique was assessed by analyzing duplicate experiments of 2 different samples: 1 was a PBMC sample, the other a lipopolysaccharide-stimulated PBMC (LPS-PBMC) sample. The 2 were included because they provided different numbers of genes in which the intensity values exceeded the background values (149 in PBMC versus 204 in LPS-PBMC). Within each replicate, the extent of variability in intensity signal for each of the 1,176 genes was calculated. Because the absolute intensity signal for each gene also depends on the background intensity, the variability values were expressed as a ratio of the corresponding background intensities. In the duplicate PBMC samples, the number of genes with variability exceeding 1×, 2×, 3×, 4×, and 5× background values were 48, 12, 2, 1, and 0, respectively. In the LPS-activated samples, they were 6, 2, 0, 0, and 0, respectively (Figure 1). This suggested that there would be only a very small margin of error if we take into consideration only those genes in the SpA samples that exceed the control samples by more than 3 times the background values.

Figure 1.

Extent of variability in microarray. Number of genes with variability exceeding the background values shown on the x-axis. The broken line represents results from the peripheral blood mononuclear cell (PBMC) in duplicate. The solid line represents results from the lipopolysaccharide-stimulated PBMC in duplicate.

In the second preliminary experiment, the microarray results of 5 different PBMCs from healthy adult subjects were compared with those of 6 different SpA tissue samples. In this and subsequent experiments, variance was compared by the F test before deciding whether the parametric or the nonparametric tests should be used. Using the t-test, the gene that showed the highest degree of statistical significance between the 2 groups was neurotrophin 4. However, the P value was only 0.005. Using the Bonferroni correction for multiple comparisons, a P value of 0.000042 was needed for actual statistical significance. A power analysis was then carried out using Statmate software (GraphPad Softwares, San Diego, CA). It indicated that the number of samples for each group should be increased to at least 8. The numbers of samples for PBMC and for SpA tissue were then increased to 10 and 11, respectively. Again, for each gene in the microarray, the mean value of the 10 PBMCs was compared with the mean value of the 11 SpA tissue samples. Taking the Bonferroni correction into consideration, only 1 gene showed results of statistical significance, with P = 0.000007. This was the M-CSF (mean ± SD of PBMC and SpA tissue samples: 3,234 ± 3,848 and 23,588 ± 8,832; values are intensity measurements of microarray results).

Many genes provided in the microarray membranes did not generate intensity signals exceeding the background. If the Bonferroni correction for multiple comparisons takes into account only those genes with positive intensity signals, 3 additional genes met the threshold. They were IL-7, MMP-3, and Bcl-xL/Bcl-2–associated death promoter (BAD) protein (P = 0.0003 for all 3 genes). Mean ± SD values of PBMC and SpA tissue samples were 1,325 ± 1,711 and 16,106 ± 9,356 for IL-7, 100 ± 260 and 23,155 ± 14,430 for MMP-3, and 1,866 ± 2,030 and 20,672 ± 11,961 for BAD protein). The neurotrophin gene was identified using the smaller number of samples to calculate power analysis and was ranked to the next highest statistical significance (P = 0.001).

Taking the intensity threshold determined in the first preliminary experiment into consideration (Figure 1), the 4 genes with the lowest P values listed above showed high likelihood of being true positives. Of the 4 genes identified, protein levels of the first 3 can be measured by ELISA.

Clinical parameters of a cohort of AS patients and their serum levels of M-CSF, MMP-3, and IL-7.

We recruited 41 AS patients attending a specialty clinic in Beijing. None of these patients were treated with biologics. Nine of the 41 patients were female. The mean ± SD age of the 41 patients was 25 ± 9.9 years and disease duration was 62.4 ± 70.4 months. The mean ± SD value for the BASDAI was 5.3 ± 0.24, with the median being 5.25 (Table 2). There is a significant correlation of the BASDAI with the erythrocyte sedimentation rate (ESR; r = 0.54, P = 0.0001; n = 41; Figure 2 lower panel) and the BASFI (r = 0.27, P = 0.05; n = 41; Figure 2, upper panel), but not with the C-reactive protein (CRP; r = 0.22, P = 0.1). However, in only 35 of the 41 patients was CRP measured. Because of this limitation, the potential clinical value of CRP was not analyzed in subsequent experiments. For BASFI, there is a significant correlation with BASDAI (r = 0.27, P = 0.05), finger to floor distance (r = 0.43, P = 0.003), and occiput to wall distance (r = 0.27, P = 0.043), but not with duration of disease, age, Schober measurement, ESR, or CRP.

Table 2. Clinical and enzyme-linked immunosorbent assay data of 41 ankylosing spondylitis patients*
Patient no.Age, yearsSexDisease duration, monthsHLA–B27Occiput-to-wall, cmChest expansion, cmFinger-to-floor, cmSchober's testWBC/mlPlatelet/ mlBASFIBASDAIESR, mm/1st hourCRPMMP-3, ng/mlM-CSF, pg/mlTIMP-2, ng/ml
  • *

    WBC = white blood cell count; BASFI - Bath Ankylosing Spondylitis Functional Index; BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; MMP-3 = matrix metalloproteinase 3; M-CSF = macrophage colony-stimulating factor; TIMP-2 = tissue inhibitor of metalloproteinase 2; Pos = positive; Neg = negative.

1720M60Pos023044.52454.54.5513 15.71,15733.3
1814M24Pos1031548.62205.44.658 01,60342.3
2012F14Neg104854.33304.95.745 24.91,59440.3
2153M72Neg1021048.61656.25.315 24.91,37439.1
2214M12Pos15450413.733676.144 16.32,06532.1
2313M60Pos15450410.74025.98.222 22.13,83336.5
Figure 2.

Correlation of Bath Ankylosing Disease Activity Index (BASDAI) withA, Bath Ankylosing Spondylitis Functional Index (BASFI) or B, with erythrocyte sedimentation rate (ESR) in 41 ankylosing spondylitis patients. Details of the 41 patients are shown in Table 2.

Serum levels of IL-7 were undetectable. For M-CSF, the serum levels in the 41 AS patients ranged from 398 to 3,833 pg/ml (Tables 2 and 3). There is no significant difference in the mean serum M-CSF levels between these 41 AS patients and 28 healthy subjects. We also collected synovial fluid samples from 8 patients with AS, 6 patients with USpA, and 1 patient with reactive arthritis. There is no statistically significant difference between the synovial fluid levels of M-CSF and the serum levels of the AS patients or the healthy subjects (Table 3).

Table 3. Serum and synovial fluid levels of M-CSF and MMP-3*
Samplenmean ± SD
  • *

    M-CSF = macrophage colony-stimulating factor; MMP-3 = matrix metalloproteinase 3; SpA = spondylarthropathy; AS = ankylosing spondylitis.

  • P = 0.0000000015 compared with serum levels.

 SpA synovial fluid15974 ± 879
 AS serum411,582 ± 952
 Healthy serum281,407 ± 892
 SpA synovial fluid152,177 ± 602
 AS serum4149.9 ± 111.4
 Healthy serum2835.1 ± 32.8

The MMP-3 serum levels of 41 AS patients ranged from 0 to 176 ng/ml (Table 2). The MMP-3 levels in the 15 SpA synovial fluids on the other hand are more than 4 times higher than the serum levels of the AS patients or the healthy subjects (P < 0.000000008; Table 3). There is no statistically significant difference in the serum MMP-3 levels between the 41 AS patients and the 28 healthy subjects.

Relationship between levels of M-CSF and MMP-3 with clinical parameters.

When compared with the various clinical parameters shown in Table 2, there is a significant correlation between the levels of M-CSF and BASDAI (r = 0.41, P = 0.004; Figure 3 upper panel). There is a weak correlation with the platelet count (r = 0.25, P = 0.05), but not the white blood cell count. There is no correlation between M-CSF and other clinical parameters listed in Table 2, including ESR (r = 0.2, P = 0.1).

Figure 3.

Correlation of serum macrophage colony-stimulating factor (M-CSF) levels with A, Bath Ankylosing Disease Activity Index (BASDAI) or B, with serum matrix metalloproteinase 3 (MMP-3) levels in the 41 ankylosing spondylitis patients shown in Table 2.

As for serum levels of MMP-3, when compared with the clinical parameters shown in Table 2, there is high correlation with ESR (r = 0.54, P = 0.0001) and with BASDAI (r = 0.48, P = 0.0007; Figure 4). There is no significant correlation with CRP or other clinical parameters listed in Table 2. Although serum levels of MMP-3 and M-CSF correlate with BASDAI, they do not correlate with one another (r = 0.2, P = 0.1; Figure 3 lower panel). Twenty-eight of the 41 AS patients showed peripheral arthritis in having both swelling and tenderness in the joints peripheral to the spine, shoulders, and hips. However, when analyzed by logistic regression, there was no correlation between the levels of MMP-3 or M-CSF with the presence of peripheral arthritis.

Figure 4.

Correlation of serum matrix metalloproteinase 3 (MMP-3) levels with A, Bath Ankylosing Disease Activity Index (BASDAI) or with B, erythrocyte sedimentation rate (ESR) in the 41 ankylosing spondylitis patients shown in Table 2.

As a control, we also measured the serum level of TIMP-2. In our microarray study, this metalloproteinase inhibitor was not found to be increased in the SpA tissue samples. The serum levels of TIMP-2 in the 41 AS patients ranged from 32.1 to 113.4 ng/ml (mean ± SD 43 ± 15.4). There is no statistically significant correlation between serum TIMP-2 levels and BASDAI (r = 0.1, P = 0.29), BASFI (r = 0.07, P = 0.3), or ESR (r = 0.1, P = 0.3).

Next, we tested the value of combining several parameters to assess disease activity. Using multiple regression analysis, we tested if there was a linear relationship model in which values of ESR, BASFI, MMP-3, and M-CSF all together would account for the variability of the BASDAI values. With the combined values of ESR and BASFI, the r2 value was 0.46. When values of MMP-3 and M-CSF were added, the r2 value was improved to 0.57. Hence, 57% of the variability of the BASDAI could be accounted for by the combination of ESR, BASFI, MMP-3, and M-CSF.

In the above analysis, disease activity was considered as a continuous variable. Using a predesigned BASDAI value as cutoff, one can also divide patients qualitatively into those with severe or mild disease. In the 41 patients in Table 2, the median BASDAI value was 5.25. We then arbitrarily grouped those with BASDAI values exceeding this median value as having severely active disease. Using univariate logistic regression analysis, we tested how well the laboratory measurements ESR, CRP, MMP-3, and M-CSF would correctly classify the patients into the categories of severe and mild disease. As shown in Table 4, ESR and MMP-3 were almost the same in chi-square values. Their P values were 0.04 and 0.025, respectively.

Table 4. Logistic regression analysis of disease severity in comparison with other parameters*
  • *

    Median of Bath Ankylosing Spondylitis Disease Activity Index was used as the cutoff to determine whether disease was severely active. MMP-3 = matrix metalloproteinase 3; M-CSF = macrophage colony-stimulating factor; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein.


Effect of infliximab infusions on serum MMP-3, M-CSF, and IL-7.

Serum from 13 Canadian AS patients were examined for the effect of infliximab infusions on serum M-CSF, MMP-3, and IL-7. The serum MMP-3 levels of 10 of these patients have been reported previously (4). The current data are completely new measurements of these patients. Patient characteristics are shown in Table 5. This is a subset of 21 patients whose response to infliximab has been reported previously. Only 13 patients were used in the present study because they were the only ones in which serum samples were still available. These 13 patients received infusions of 3 mg/kg of infliximab at weeks 0, 2, and 6. Measurements were carried out before infusion and also at week 14. In these 13 patients, the BASDAI scores decreased significantly from 6.4 ± 1.5 at week 0 to 3 ± 2.1 at week 14 (P = 0.00007 by paired t-test; Figure 5, upper panel). In only 2 subjects was there no clinical response, in that the decrease in BASDAI scores were either <2.0 or <20% of values before infusion. When we measured the serum MMP-3 levels of these 13 patients, there was a statistically significant decrease in MMP-3 at week 14 (P = 0.013 by Wilcoxon matched-pairs signed-ranks test; Figure 5 middle panel). For samples taken before infliximab infusion, there was a high degree of correlation between MMP-3 and ESR (r = 0.74, P = 0.002) and CRP (r = 0.65, P = 0.008). The correlation became insignificant for samples taken after the infliximab infusions. There was no statistical relationship between changes in BASDAI values and changes in MMP-3. ESR, CRP, and the serum levels of MMP-3 taken before the infusions were not predictive of an absence of clinical response to the anti-TNF agent. There was no statistically significant difference of M-CSF serum levels before and after infliximab (Figure 5 lower panel).

Table 5. Characteristic of ankylosing spondylitis patients on infliximab*
Designation no.SexAge, yearsConcommitant conditionsPrevious DMARDMMP-3 before infusionsMMP-3 after infusions
  • *

    DMARD = disease-modifying antirheumatic drug; MMP-3 = matrix metalloproteinase; PA = peripheral arthritis; PSA = psoriatic arthritis; MTX = methotrexate; HCQ = hydroxy chloroguine; SSZ = sulfasalazine.

1M26PA, PsA 22051
2F39PA, PsA 5114
3F60PA, psoriasis 14214
4M55PA PsA 2522
5M35PA 188102
6M53PA 4257
7M51PA 2736
8F38  3518
9M40 Pamidronate2226
10M31 Pamidronate1,221394
11M49Crohn'sMTX, gold1913
12M48Crohn's 12542
13M31 MTX, HCQ, SSZ11343
Figure 5.

Effect of infliximab infusions on A, Bath Ankylosing Disease Activity Index (BASDAI), B, serum levels of matrix metalloproteinase 3 (MMP-3), and C, macrophage colony-stimulating factor (M-CSF). Total of 13 ankylosing spondylitis patients were studied.


Other than genome scanning and microarray screening, most research effort on the spondylarthropathies have been narrowly focused because they are hypothesis driven, and so limited by the availability of candidate causative factors (1). Our random screening has generated 2 research candidates that are relatively new to the field of SpA: MMP-3 and M-CSF.

MMP-3 has been an subject of intense study in inflammatory arthritis conditions, especially rheumatoid arthritis (RA) (9). One example is a study that prospectively collected monthly serum MMP-3 measurements for 3 years in 33 patients with early RA. Ribbens et al discovered that the serum MMP-3 levels correlate with swollen joint counts, RA disease activity scores, ESR, CRP, and outcome of radiologic joint space narrowing (9). In contrast to RA, very few SpA studies address the significance of serum MMP-3. This is probably because the mean level of serum MMP-3 in SpA is within the range of normal subjects, as observed in this study (10, 11).

There are at least 3 observations in our results that are remarkable. The first observation is that MMP-3 levels are much higher in synovial fluids compared with serum samples. This would indicate that there is high expression of the MMP-3 protein in the joints. Interestingly, MMP-3 has also been reported to be highly expressed in discs of the spine (12), a target of SpA disease process. Being generated in the articular areas, measurements of serum MMP-3 would in theory reflect more closely the events at the joints compared with those markers generated elsewhere, such as ESR or CRP. Our second new observation is that there is a certain degree of correlation between serum MMP-3 levels and disease activity. Measuring disease activity has always been a challenge in SpA. The standard measurement, the BASDAI, consists of a number of question-based scale evaluations (2). It is completely subjective. In the case of those 41 Beijing patients we studied, it probably has a certain degree of reliability because there is a correlation of the BASDAI values with the values of the functional impairment index. ESR and CRP are objective measurements, and have been used as adjuncts in evaluating disease activity of several inflammatory rheumatic diseases. In our 41 patients, the BASDAI values do correlate with ESR. However, at least 2 studies have found ESR and CRP to be questionable measurements of SpA disease activity (13, 14). Our discovery that there is a certain degree of correlation between serum MMP-3 levels and BASDAI would indicate that it could be studied as another objective parameter of measurement of disease activity. Accurate assessment of AS disease activity has become very important in clinical practice. Many patients with active AS will respond favorably to treatment with etanercept or infliximab, as we reported here. Because of the high cost of the drugs, it is important to generate an accurate method to measure disease activity to ensure that a particular patient does have a favorable response to the drugs. Our third observation here addresses this problem. Similar to what one of us have reported with a smaller subset of this same group of patients (4), and also in RA (15), we confirmed that infliximab infusions did induce a significant decrease in MMP-3 levels. Our 3 findings with MMP-3 would strongly indicate that it should be studied as a measurement of disease activity, especially during treatment with infliximab.

Our second microarray candidate, M-CSF, is a completely unexpected one. M-CSF has never been reported in SpA patients. A very small number of reports suggest that its level is increased in the serum and synovial fluids of RA patients (16). We discovered that in the 41 AS patients from Beijing, serum levels of M-CSF correlate both with BASDAI and ESR. However, in the Canadian patients, they are not significantly suppressed by infliximab treatment, which would suggest that the M-CSF and MMP-3 constitute independent parameters.

Our discoveries regarding MMP-3 and M-CSF are also helpful for research on the pathogenesis of SpA. The molecular pathway mediating the spondylarthropathies is unknown. Even the particular cell type that is responsible is controversial. The once-favorite hypothesis, that the spondylarthropathies are mediated by a CD8+ T cell reactivity toward self peptides, has been challenged by multiple findings both in human subjects and in animal models (1). One of the alternate hypotheses, for example, is that the spondylarthropathies are induced by an endoplasmic reticulum (ER) overloading response secondary to the slow rate of folding of the HLA–B27 heavy chains (17). What is clear is that to induce this ER response, there must be factors in addition to the expression of HLA–B27, and that most likely these factors target only certain cell types, and are localized only at the disease areas. Our preliminary experiments have suggested that synovial monocytes/macrophages are targets of this overloading response (18). However, no additional factors contributing to this overloading response have yet been proposed. M-CSF is unique among potentially arthritis-causing factors in that it affects only cells of the monocyte lineage. It can induce replication of such cells and activate the monocytes to contribute to inflammatory processes. Experimental animals that are deficient in M-CSF are resistant to development of collagen-induced arthritis (19). With our observation of its possible high expression in the synovial tissues, and the positive correlation of serum levels with disease activity, it will be useful to examine its role in the SpA processes. The observation that it is not affected by infliximab infusions would place the expression of M-CSF in events upstream of those targeted by anti-TNF agents. The decrease in MMP-3 by infliximab would place its expression downstream of those events. It is possible, for example, that the suppression of MMP-3 is secondary to the suppression of IL-1 expression by infliximab.

The reason these 2 factors have not been subjects of SpA research in the past is probably that their serum levels in SpA patients are within the range of healthy subjects. There are several possibilities why the serum levels of MMP-3 and M-CSF in AS are within the range of healthy subjects, and yet still correlate with disease activity. MMP-3 can exist in at least 3 forms: pro-MMP-3, active MMP3, and MMP-3 complexed with TIMP (10). The ELISA method used here measures the total of all 3 forms. It is possible that differences between AS patients and healthy subjects can be detected if we measure these 3 forms separately. M-CSF can also exist in several forms (20). Measurements of each of these might differentiate between healthy subjects and AS patients. Lastly, even if the MMP-3 and M-CSF levels in AS are in the same range of healthy subjects, it is not a deterrent for using them in clinical practice. Serum CRP levels in apparently healthy individuals, for example, are frequently measured as a predictor of heart disease (21). In the case of MMP-3, it is not a deterrent to using it to monitor response to infliximab, provided that serum samples are obtained prior to therapy. To assess whether measurement of MMP-3 and M-CSF are useful in AS patients unrelated to anti-TNF agents, we have used the multiple regression and the logistic regression analyses to test whether they account for the variability of the BASDAI values. The results do show a certain degree of relationship. Although BASDAI is the current standard for evaluation of disease activity, additional assistance by objective parameters, such as those described here, might lead to a higher degree of accuracy. This will of course require validation.

In summary, the results of our experiments have highlighted the potential roles of MMP-3 in particular in measuring disease activity, and also both MMP-3 and M-CSF as foci of research on disease mechanisms.