Serum Neutrophil Gelatinase B-associated Lipocalin and Matrix Metalloproteinase-9 Complex as a Surrogate Marker for Mucosal Healing in Patients with Crohn's Disease.

BACKGROUND AND AIMS
Although costly and uncomfortable for the patient, the current standard to assess mucosal healing in Crohn's disease [CD] patients is endoscopy. The aim of this study was to evaluate NGAL-MMP-9 as surrogate marker for mucosal healing in CD patients.


METHODS
Serum NGAL-MMP-9 levels were determined with sandwich enzyme-linked immunosorbent assay before and up to 5 years after first infliximab infusion in 108 active CD patients [median age at first infliximab 36 years, 57% female] and 43 healthy controls [HC, median age 27 years, 60% female]. Serum samples were matched to the time of endoscopy and complete endoscopic healing was defined as absence of ulcerations. Histological healing was defined as absence of epithelial damage [D'Haens score].


RESULTS
At baseline, median [interquartile range] NGAL-MMP-9 levels were significantly higher in active CD patients vs HC (77.6 [36.9-141.0] vs 25.5 [17.8-42.8] ng/ml; p < 0.001). After treatment, NGAL-MMP-9 levels significantly decreased in completely healed CD patients [n = 38] (84.5 [36.7-138.4] to 23.4 [7.4-42.5] ng/ml; p < 0.001) and--to a lesser extent--in non-healed CD patients [n = 36] (100.9 [43.4-152.6] to 43.8 [27.0-96.8] ng/ml; p = 0.001). Receiver operating characteristic analysis defined a NGAL-MMP-9 cut-off level of 45 ng/ml corresponding to complete endoscopic healing (area under the curve [AUC] = 0.79, 82% sensitivity, 65% specificity) and histological healing [AUC = 0.72, 79% sensitivity, 53% specificity]. At baseline, C-reactive protein [CRP] was not elevated in 33% of active CD patients, whereas 53% of these patients did have elevated NGAL-MMP-9 levels.


CONCLUSIONS
In the search for surrogate markers to assess mucosal healing in inflammatory bowel disease, NGAL-MMP-9 supplements and outperforms CRP in both ulcerative colitis and CD patients.


Introduction
Crohn's disease [CD] is a chronic, relapsing disease of the gastrointestinal tract with increasing prevalence and incidence in both industrialised and developing countries. 1 In CD, a defective acute immune response with impaired neutrophil accumulation and interleukin [IL]-8 production is observed. 2,3 This may lead to delayed or incomplete removal of bacteria that breach the mucosal barrier. These bacteria will subsequently be taken up by macrophages 4 that elicit a granulomatous reaction 5 and produce a secondary chronic inflammation. 3 However, neutrophil migration to the active site of inflammation is stimulated by bacterial components, eg through the pro-inflammatory NOD2/CARD15 pathway.
The primary function of neutrophils is to kill bacteria, with the use of potent digestive enzymes present in and secreted from the granules. Neutrophil gelatinase B-associated lipocalin [NGAL, lipocalin-2] 6,7 is found in secondary neutrophil granules. NGAL is expressed in response to Toll-like receptor activation during infections 8,9 and can inhibit bacterial growth by sequestering iron-laden siderophores. 10 Moreover, NGAL has been correlated with parameters of active disease in IBD patients. [11][12][13][14][15][16] Tertiary neutrophil granules contain matrix metalloproteinase-9 [MMP-9, gelatinase B], a member of the MMP family. 17 MMPs are zinc-dependent endopeptidases involved in many developmental processes, including angiogenesis, wound healing, and extracellular matrix [ECM] degradation. Dysregulated MMP-9 levels have been previously described in inflammatory bowel disease [IBD]. 11,18,19,20,21,22,23,24,25,26 Recently, the EMBARK study showed that a combination of fecal calprotectin, serum MMP-9, and serum IL-22 had a strong association with imaging/endoscopy-defined inflammation. 27 Moreover, neutralising antibodies with tissue inhibitor of MMPs [TIMP]-like mechanisms against MMP-2 and MMP-9 were shown to attenuate the development of colitis in IBD mouse models. 28 In addition to their separate circulating forms, MMP-9 and NGAL occur in covalent complexes, mainly in neutrophil degranulates. 17 However, the functional role of this NGAL-MMP-9 complex is still debated. One hypothesis is that by formation of this complex, NGAL protects MMP-9 from autodegradation. 29 In a few studies, NGAL or MMP-9 levels were investigated separately in blood or biopsies after anti-inflammatory treatment in patients with CD. 30,31 Based on our original view that by measuring the covalent complex of NGAL with MMP-9 we would evaluate two markers in one assay and eventually combine the information content of two assays, we recently described that serum NGAL-MMP-9 complex levels decrease after infliximab treatment in ulcerative colitis [UC] patients achieving mucosal healing, and that levels correlate well with the Mayo endoscopic subscore. 32 With the present study, we aimed to investigate whether NGAL-MMP-9 complex levels are elevated in CD patients in comparison with healthy controls and whether levels decrease after anti-inflammatory treatment in patients with mucosal healing. Moreover, since C-reactive protein [CRP] is not elevated in 20-30% of CD patients, 33 we aimed to compare the diagnostic accuracy of NGAL-MMP-9 with CRP.

Patient sampling
Consecutive serum sampling and endoscopy were performed in 108 CD patients before and after first treatment with infliximab [Remicade; Centocor]. Baseline characteristics of the CD patients are shown in Table 1. The baseline samples were obtained from active CD patients within 1 month before first infusion of infliximab, and follow-up samples were obtained up to 5 years after start of treatment. The median (interquartile range [IQR] time to follow-up endoscopy was 13  weeks. The median [IQR] interval between follow-up serum sampling and last infusion with infliximab was 49  days. Serum samples were matched to a current endoscopy with a maximum interval of 30 days between serum sampling and time of endoscopy. CRP levels and neutrophil counts were measured in a centralised laboratory facility before and after infliximab, at time points corresponding to endoscopy. Furthermore, we collected serum samples from 43

Sandwich ELISA
The commercial anti-human NGAL-MMP-9 complex ELISA [enzymelinked immunosorbent assay] kit [R&D Systems, Abingdon, UK] was used to determine NGAL-MMP-9 complex levels in the serum of CD patients and HC according to the manufacturer's guidelines. Briefly, two antibodies with different antigen specificities were used: an antibody directed towards MMP-9 was pre-coated on the plate and another against NGAL was used as the detection antibody. Hence, only NGAL-MMP-9 complexes were measured. The absorbance was measured at 450 nm with a spectrophotometer [Omega, Nazareth, Belgium]. NGAL-MMP-9 complex levels were quantified with the use of a calibration curve using purified human NGAL-MMP-9 as a standard [Mars software, BMG Labtech, Ortenberg, Germany].  Figure 1A and Table 2]. After treatment, NGAL-MMP-9 levels significantly decreased in completely healed CD patients [p < 0.001], although four patients [10%] showed increased NGAL-MMP-9 levels after therapy [ Figure 1A and Table 2]. Patients with partial healing had mild decrease of NGAL-MMP-9 levels after treatment [p = 0.048] [ Figure 1A and Table 2], and the decrease of NGAL-MMP-9 levels [difference between before and after treatment] was significantly lower than in patients with complete healing [p = 0.001]. Moreover, 10 patients with partial healing [34%] had increased levels after treatment. In non-healed CD patients, NGAL-MMP-9 serum levels also decreased after treatment [p = 0.001] [ Figure 1A and Table 2]. However, the decrease was significantly less profound than in completely healed CD patients [p = 0.020]. No significant difference in extent of decrease of NGAL-MMP-9 levels was observed in completely healed patients compared with partially healed patients [p = 0.294]. Moreover, NGAL-MMP-9 levels in completely healed CD patients decreased after treatment to levels equivalent to HC levels (median [ Figure 1A and Table 2]. No significant differences were found between NGAL-MMP-9 levels in completely, partially, or non-healed CD patients at start of treatment [ Figure 1A and Table 2].

Statistical analysis
Finally, we investigated whether patients at time of follow-up endoscopy were maintained under the same type of concomitant treatment as at the time of first infliximab infusion. We found that 56%, 83%, 96%, and 73% of the patients who received 5-aminosalicylic acid [

Serum NGAL-MMP-9 complex levels correlate with neutrophil counts and complement CRP as an inflammatory marker
To assess its role as a serum marker of inflammation and mucosal healing, we correlated serum NGAL-MMP-9 complex levels with neutrophil counts and CRP levels. NGAL-MMP-9 serum levels correlated well with the amount of neutrophils in the blood [Spearman's rho [r] = 0.470, p < 0.001]. For most of the patients, neutrophil counts were found to be in the clinically determined interval of 2.5-7.8 10 9 /l. Nevertheless, 25% of the CD patients presented with neutrophilia [> 7.8 10 9 /l]. Neutrophil counts significantly decreased after treatment in completely healed CD patients [p < 0.001] [ Table 2 and Figure 1C], whereby 13% of the patients showed neutropenia with neutrophil counts lower than 2.5 10 9 /l. CD patients with partial healing after treatment also showed a decrease in neutrophil counts [p = 0.002] [ Table 2 and Figure 1C]. However, this decrease had a trend to be less profound than in patients with complete healing [p = 0.118]. In CD patients without healing, neutrophil counts also decreased after treatment [p = 0.005], but to a lesser extent than in patients with complete healing [p = 0.153] [ Table 2 and Figure 1C]. The decrease of neutrophil counts was not significantly different between partially healed or non-healed CD patients [p = 0.928]. Furthermore, no significant difference was observed between baseline neutrophil counts of CD patients with complete healing compared with CD patients without mucosal healing [p = 0.95] [ Table 2 and Figure 1C]. The amount of neutrophils correlated with CRP levels [r = 0.357, p < 0.001]. In patients with high CRP levels, the neutrophil count was also elevated [> 5.4 10 9 /l] in 56% of the patients whereas, in patients with low CRP levels, 70% of the patients also had low neutrophil levels. In cases of low or high NGAL-MMP-9 levels, neutrophil counts were high [> 5. 4 Figure 1B]. We observed no significant difference in the decrease of CRP levels between complete and partially healed patients [p = 0.420]. CRP levels also decreased in CD patients without mucosal healing after treatment [p = 0.037] [ Table 2 and Figure 1B]. However, the decrease of CRP levels was more profound in complete or partial CD healers after treatment than in CD nonhealers [p = 0.104 and p = 0.003, respectively]. Of importance, CRP was not elevated [< 5mg/l] in 33% of patients with active disease at start of treatment, whereas 53% of these patients did have elevated [> 45 ng/ml] NGAL-MMP-9 levels. Moreover, 47% of the patients   Figure 2A]. In contrast, CRP was markedly elevated in CD patients who did not heal, whereas CRP levels in patients with partial and complete healing were similar and therefore not discriminative [ Figure 2B]. Neutrophil count reflected the degree of endoscopic improvement, discriminating between no, partial, and complete mucosal healing [ Figure 2C] 3.4. Serum NGAL-MMP-9 complex can discriminate complete mucosal healing as defined by endoscopic evaluation ROC analysis was performed to evaluate the performance of NGAL-MMP-9, CRP, and neutrophil levels to discriminate mucosal healing. When including patients with partial mucosal healing in the non-healing group of patients, the AUC for NGAL-MMP-9 levels was 0.77, and levels lower than 45 ng/ml were determined to discriminate complete mucosal healing with 82% sensitivity, 60% specificity, 29% positive predictive value [PPV] and 95% negative predictive value [NPV]. The AUC for CRP levels was 0.74, and levels lower than 5 mg/l were able to discriminate complete mucosal healing with 79% sensitivity, 57% specificity, 28% PPV and 93% NPV. Neutrophil levels lower than 5.4 10 9 /l were able to discriminate complete mucosal healing with an AUC of 0.68, sensitivity of 79%, specificity of 48%, 25% PPV, and 91% NPV. ROC analysis, whereby patients with partial mucosal healing were excluded, indicated that NGAL-MMP-9 complex levels lower than 45 ng/ml could discriminate complete mucosal healing from no healing with a sensitivity of 82% and specificity of 64% [ Figure 3A and Table 3]. The AUC was 0.79, and a PPV of 44% and NPV of 91% were determined. The diagnostic accuracy of NGAL-MMP-9 was 68%. ROC analysis with CRP levels showed a comparable AUC of 0.75. CRP levels lower than 5 mg/l were able to discriminate complete mucosal healing with 79% sensitivity, 58% specificity, 39% PPV, and 89% NPV [ Figure 3A and Table 3]. To investigate the superiority of NGAL-MMP-9 over CRP, we analysed the performance of NGAL-MMP-9 to discriminate complete mucosal healing in patients with low CRP levels at baseline. Of the 36 patients with low CRP, 18 patients had complete healing and 11 patients had no healing after treatment. An AUC of 0.78 was determined, and a cut-off value of 38 ng/ ml NGAL-MMP-9 was able to discriminate complete mucosal healing with 83% sensitivity, 58% specificity, 47% PPV, and 88% NPV. Finally, a neutrophil count lower than 5.4 10 9 /l was able to discriminate complete mucosal healing with an AUC of 0.70, 79% sensitivity, 52% specificity, 36% PPV, and 88% NPV [ Figure 3A and Table 3].
In order to investigate whether the combination of NGAL-MMP-9 and CRP in clinical practice would improve the prediction of complete mucosal healing, we performed binary logistic regression analysis including both parameters. With the use of the predicted probabilities, ROC analysis was performed and indicated that the combination of the two markers was able to discriminate complete mucosal healing with an AUC of 0.81, 82% sensitivity, 73% specificity, 51% PPV, and 92% NPV [ Figure 3A and Table 3]. The combination of three markers [NGAL-MMP-9, CRP, and neutrophils] did not improve the discriminative power [AUC of 0.81] as compared with the combination of two markers [NGAL-MMP-9 and CRP].

Discussion
Identification of non-invasive biomarkers has become an important research topic, since frequent endoscopic examinations are costly and uncomfortable for the patient. Several serological and fecal markers have been investigated for their use in diagnosis and assessment of disease activity, but none have been shown to be superior to the commonly used CRP. 35 In this study, we investigated the diagnostic accuracy of NGAL-MMP-9, CRP and neutrophil levels in a considerable CD cohort and found that NGAL-MMP-9 performed better than CRP and neutrophil count in discriminating complete endoscopic and histological healing. With ROC analysis, we showed that NGAL-MMP-9 was the best discriminator of complete endoscopic healing and that the combination with CRP discretely enhanced the discriminative power. Importantly, we found that CRP was not elevated in one-third of CD patients with active disease at start of treatment, whereas more than half of these patients did have elevated NGAL-MMP-9 levels. This was further investigated by ROC analysis, identifying that NGAL-MMP-9 was a good surrogate marker to discriminate complete mucosal healing in patients without elevated CRP levels at baseline. These data suggest that NGAL-MMP-9, which is a combination of two marker molecules in one assay, can be considered as a new surrogate marker for IBD patients to assess mucosal healing and can complement or even replace CRP measurements. In a previous study, we already showed the value of the NGAL-MMP-9 complex as surrogate marker for mucosal healing in UC patients. 32 Besides CRP, other markers currently used include fecal calprotectin. Fecal calprotectin is a sensitive marker of intestinal inflammation and correlates well with the degree of endoscopic activity. 36 Patients with ileal CD are reported to have significantly lower fecal calprotectin levels than those with [ileo-]colonic disease, even in the presence of large and/or very large ulcers. 37 Moreover, the time of sampling during the day may affect calprotectin levels. 38 In our cohort, we found that NGAL-MMP-9 levels were lower in patients with active ileal disease compared with patients with [ileo-]colonic disease before start of treatment [ Supplementary Figure 2A, B]. Since neutrophils are the source of both calprotectin and NGAL-MMP-9, we further looked whether the amount of neutrophils was lower in patients with ileal disease compared with [ileo-]colonic disease; however, we could not document a significant difference.
A limitation of our study is the lack of fecal calprotectin measurements for a direct comparison with NGAL-MMP-9. Interestingly, the EMBARK study showed that a combination of fecal calprotectin, serum MMP-9, and serum IL-22 had a strong association with imaging/endoscopy-defined inflammation. 27 Moreover, recent reports indicate the emergence of fecal MMP-9 levels in the IBD biomarker field. Annahazi et al. described that fecal MMP-9 is a good marker for the non-invasive evaluation of disease activity and mucosal healing in UC. 39 Moreover, Kolho et al. showed that fecal MMP-9 performed equally well as fecal calprotectin in UC, suggesting its use as a surrogate marker of inflammation. 19 Recently, fecal MMP-9 was also tested in a cohort of CD patients, but was not correlated with any of the activity indices of CD. 40 Therefore, it might be interesting in the future to evaluate NGAL-MMP-9 complex levels in fecal samples.
As was shown by previous studies, endoscopic mucosal healing does not necessarily reflect quiescent histological disease. 41,42 Most of the studies investigating histological healing were performed in UC patients, and there is scarce information in CD. 43 In the present study, we used the D'Haens score 34 which is also known in literature as the Colonic or Ileal Global Histologic Disease Activity Score [GHAS]. Despite the evidential importance of microscopic activity, histological remission has yet to be recommended as a therapeutic endpoint for clinical trials or practice in IBD. 44 In our cohort of CD patients, we found a good correlation between histological and endoscopic mucosal healing. A high concordance was found, especially in patients with active disease. Importantly, 72% of the patients with complete endoscopic healing also had histological healing with an absence of epithelial damage. These data are in line with current literature, since it was reported that persistent histological inflammation occurs in 25-37% of patients with clinical and endoscopic quiescent CD. 45,46 Since CD is characterised by transmural disease, endoscopic and histological mucosal evaluations are not able to actually determine 'deep' remission in CD patients. Recent studies indicate that magnetic resonance enterography [MRE] can evaluate ulcer healing with a high level of accuracy compared with ileocolonoscopy. 47 This method is, however, not routinely performed in all patients, is costly, and requires technical and analytical skills. Moreover, it has been recommended that the combination of mucosal and histological healing should be achieved as a minimum therapeutic target in IBD patients. 48 In order to investigate the diagnostic value of NGAL-MMP-9, we compared the levels between active UC and CD patients; however, we could not document a significant difference [Supplementary Figure 3]. Moreover, with ROC analysis we defined a higher cut-off value [97.7 ng/ml] in our UC cohort than in the present CD cohort [45 ng/ml]. This may be in part due to the fact that, in general, higher NGAL-MMP-9 levels were found in UC patients compared with CD patients. In addition, the cut-off values were chosen in order to identify as accurately as possible patients with mucosal healing. In the UC cohort, a high specificity of the test was chosen since levels higher than 97.7 were positively associated with no healing. In the present CD cohort, a high sensitivity was chosen since levels lower than 45 ng/ml were positively associated with complete mucosal healing.
Although NGAL and MMP-9 have been discussed separately as good markers of disease, 13,14,15,16,27 our previous study in UC patients was the first to investigate the NGAL-MMP-9 complex as a surrogate serum marker of mucosal healing. 32 Urinary NGAL-MMP-9 has been reported to predict paediatric IBD, 11 and in cancer research NGAL-MMP-9 complex is a well-known biomarker. 49,50,51 In conclusion, we propose that serum NGAL-MMP-9 complex is useful and recommended as a surrogate marker of endoscopic and histological mucosal healing after treatment with infliximab in both UC and CD patients. NGAL-MMP-9 complex hereby outperforms CRP and can be used as a single marker in patients without elevated CRP levels or in combination with CRP to discriminate mucosal healing. Prospective studies to evaluate this new marker are needed, and the efficacy of the marker to discriminate mucosal healing under other emerging gut-selective biological treatments [eg vedolizumab] should be investigated.