Circulating cathelicidin levels correlate with mucosal disease activity in ulcerative colitis, risk of intestinal stricture in Crohn’s disease, and clinical prognosis in inflammatory bowel disease

Blood was collected from control, UC, and CD patients through procedures established by UCLA. For cohort 1, IBD patients were recruited from a UCLA Gastroenterology clinic, and control samples were obtained from a UCLA Internal Medicine clinic. Serum samples from cohort 1 were prepared by UCLA Department of Pathology. Serum samples from cohort 2 were obtained from UCLA Center for Inflammatory Bowel Diseases Biobank. All blood samples were collected between 2012 and 2015. The blood samples were collected following indicated diagnostic procedures ordered by the physicians. Serum volume availability must be 100 μL or above. Specimen must be in good quality with no sign of hemolysis. For IBD patients, clinical data of follow-up visits at 6–18 months after the initial blood draw were collected for the analysis. The blood samples of cohort 1 and 2 do not overlap.

Inclusion criteria: Samples were obtained from male and female subjects (age 21–67 years). IBD-specific inclusion criteria: The IBD patients sought diagnosis and/or treatment for IBD. IBD diagnosis (UC and CD) was confirmed by board-certified gastroenterologists. Exclusion criteria: No pregnant women, prisoners, or minors under age 18 were included. Patients with concurrent acute infection (CMV, C. difficile, and tuberculosis) and malignant conditions were not included. IBD patients without follow-up visits at 6–18 months after the initial blood draw were not included. Dr. Koon has already included all qualified blood samples available at the time of ELISA experiments in 2015.

The Partial Mayo Score (PMS) and Harvey Bradshaw Index (HBI) were used to assess the clinical disease activity of UC and CD patients respectively [30, 31]. Evaluation of PMS and HBI was performed on the same day of blood sample collection for LL-37 ELISA, CRP, and other tests. The diagnosis of stricture was determined by imaging procedures or colonoscopy occurring within 1 month of blood sample collection. Mayo Endoscopic Subscore (MES) was used to assess the mucosal disease activity of the UC patients (range 0–3) [32]. All data were collected prospectively.

Measurement of LL-37 was performed using an ELISA kit (#HK321, Fisher Scientific, Pittsburgh, PA) at Dr. Koon’s laboratory according to manufacturer’s instructions as previously described [25]. CRP levels were determined by the UCLA Department of Pathology and their values were available at the CareConnect database.

Power analysis has shown that this study required at least 40 patients per group of control, UC, and CD patients to achieve a statistically significant difference of serum LL-37 levels between control (37 ng/ml), UC (56 ng/ml), and CD (58 ng/ml) patients with standard deviation = 21, alpha = 0.5, and power = 0.8. The combined data from two cohorts yielded 70 control, 80 UC, and 95 CD patients total that satisfied this requirement.

LL-37 protein concentration and CRP levels were arranged in order from low to high. Low LL-37 levels indicated moderate and severe disease, and high LL-37 levels indicated remission. High CRP levels indicated moderate and severe disease, and low CRP levels indicated remission. We determined the performance of multiple cut-off points of LL-37 and CRP levels at each disease parameter. The optimized universal cut-off points were shown in this study.

Prevalence of disease, sensitivity, specificity, positive predictive value (PPV), and negative predictive values (NPV), with their 95% confidence intervals (CI) were calculated using a clinical calculator website: http://​vassarstats.​net/​clin1.​html. The accuracy of a test was determined by Area under Curve (AUC) of Receiver Operating Characteristics (ROC) using: http://​www.​rad.​jhmi.​edu/​jeng/​javarad/​roc/​JROCFITi.​html. Odds ratio and relative risk were calculated using a clinical calculator website: https://​www.​medcalc.​org/​calc/​. Results were expressed as mean +/− SEM. Bar graphs and scatter plots were made using Microsoft Excel and GraphPad Prism (San Diego, CA). Unpaired Student’s t-tests were used for intergroup comparisons of continuous data and Fisher’s exact tests were used for intergroup comparisons of category data using GraphPad Prism. Significant p values are shown in each figure.

Our results showed UC and CD patients had significantly increased serum cathelicidin levels with 56 ± 5 ng/ml and 58 ± 4 ng/ml (mean ± sem) respectively, compared to control patients averaging 37 ± 2 ng/ml (Fig. 1a). Cathelicidin levels were not associated with age at collection, age at diagnosis, duration of diseases, gender, or smoking habits in both UC and CD patients (data not shown). Serum cathelicidin levels were inversely proportional to PMS of UC patients, consistent with its known anti-inflammatory effect (Fig. 1b & d). As expected, serum CRP levels were directly proportional to PMS of UC patients (Fig. 1c-d).

We used two optimized universal cut-off points of LL-37 and CRP levels for indicating the clinical disease activity in UC patients from two cohorts. A combination of high LL-37 levels and low CRP levels was slightly more accurate than LL-37 alone or CRP alone in indicating UC clinical remission (Fig. 2a). Sensitivity and specificity of the LL-37 + CRP combined test in indicating UC clinical remission are 0.71 and 0.80 respectively (Additional file 1: Figure S1A). LL-37 alone was slightly more accurate than CRP alone in indicating moderate and severe UC clinical disease activity (Fig. 2b). Sensitivity and specificity of the LL-37 + CRP combined test in indicating moderate and severe UC clinical disease activity are 0.69 and 0.78 respectively (Additional file 1: Figure S1B).

To evaluate the future disease development of UC patients, we analyzed PMS values that were assessed by the physicians between 6 and 18 months after the initial blood draw for LL-37 and CRP measurement. Since some of the patients were initially in clinical remission while others began with moderate or severe disease activity, we evaluated these two groups of patients separately. The result shows that all patients in clinical remission initially (PMS = 0–1) remained in remission or slightly worsened to mild disease activity after 6–18 months (Fig. 2c). For the patients with moderate or severe disease initially, the high LL-37 group had significantly greater reduction in PMS values than the low LL-37 group (4.2 ± 0.5 versus 1.0 ± 0.6), suggesting improved future clinical activity (Fig. 2c). These patients with high LL-37 levels were significantly more likely to have a PMS reduction of 2 points or more in approximately 1 year, with an odds ratio equal to 39 (Fig. 2d). We found no association between LL-37 levels and use of medication from the day of blood draw (0 month) throughout monitoring period (6–18 months, average =12 months) (Additional file 2: Figure S2A). Among the steroid-free UC patients, the high LL-37 group still had significantly greater reduction of PMS values than the low LL-37 group (4.0 ± 0.7 versus 1.5 ± 0.5) (Additional file 2: Figure S2B). Both findings suggested that the improved future clinical activity of the high LL-37 group was not influenced by use of any specific group of medication.

On the other hand, there was no statistically significant difference in PMS reduction after 6–18 months between high and low CRP groups (Fig. 2e). Our study suggests that LL-37 is more accurate than CRP in indicating future clinical activity of moderate or severe UC patients.

LL-37 levels were inversely proportional to Mayo endoscopic scores (Fig. 3a). Although the difference of serum LL-37 levels between MES = 0 and MES = 3 was not statistically significant, we observed that all UC patients with severe mucosal disease activity (MES = 3) had LL-37 levels below 60 ng/ml while all UC patients with endoscopic remission (MES = 0) had LL-37 levels above 32 ng/ml (Fig. 3a). Therefore, we set up two different universal LL-37 cut-off points for indicating the mucosal disease activity in UC patients from two cohorts. CRP levels were directly proportional to Mayo endoscopic score (Fig. 3b). The difference of serum CRP levels between patients with MES = 0 and patients with MES = 3 was statistically significant (Fig. 3b).

In the evaluation of UC endoscopic remission, a combination of high LL-37 (above 32 ng/ml) levels and low CRP (0.4 mg/L or below) levels (AUC = 0.84) were more accurate than LL-37 alone (AUC = 0.76) or CRP alone (AUC = 0.71) (Fig. 3c). The LL-37 + CRP combined test (0.38) and LL-37 alone (0.38) had slightly better specificities than CRP alone (0.30) (Additional file 3: Figure S3A).

A combination of low LL-37 (below 60 ng/ml) levels and high CRP (above 0.5 mg/L) levels (AUC = 0.80) were more accurate than LL-37 alone (AUC = 0.59) or CRP alone (AUC = 0.71) in indicating severe endoscopic disease activity in UC patients (Fig. 3d). The LL-37 + CRP combined test (0.85) had better specificity than LL-37 alone (0.32) and CRP alone (0.77) (Additional file 3: Figure S3B). Co-evaluation of the LL-37 test and the CRP test can provide an accurate assessment of mucosal disease activity, especially the indication of severe UC endoscopic activity. We found no association between LL-37 levels, UC disease location (E1-3), and use of medication (Additional file 3: Figure S3C).

Serum cathelicidin levels were inversely proportional to HBI values of CD patients (Fig. 4a & c). There was no correlation between HBI values and CRP levels (Fig. 4b & c). We used a universal LL-37 cut-off point to differentiate levels of clinical disease activity in CD patients from both cohorts. For both indication of CD clinical remission and moderate/severe CD clinical disease activity, LL-37 alone was equally as accurate as CRP alone (Fig. 5a-b). Co-evaluation of both parameters did not further increase accuracy. The prevalence, sensitivity, specificity, PPV, NPV, and AUC data were shown in Additional file 4: Figure S4A-B. We found no association between LL-37 levels, age (A1-3), disease location (L1-4), disease behavior (B1-3/P), or use of medication at the time of blood draw (Additional file 4: Figure S4C). The mucosal disease activity of CD patients was assessed by simple endoscopic score (SES-CD); however, there was no statistically significant correlation between LL-37 and mucosal disease activity in CD patients (the high LL-37 group 3.11 ± 0.4 vs. the low LL-37 group 4.14 ± 1.6, mean ± sem).

To assess future disease development of CD patients, we analyzed their HBI values collected between 6 and 18 months after the initial blood draw for LL-37 and CRP measurement. We evaluated the patients with initial clinical remission and moderate/severe disease separately. The result shows that the patients initially in remission (HBI = 0–4) remained in remission after 6–18 months, regardless of LL-37 levels (Fig. 5c). For the patients with moderate or severe disease initially, the high LL-37 group had significantly reduced HBI values (−9.4 ± 0.7) at the second measurement compared to the low LL-37 group (−6.3 ± 0.8), suggesting better future clinical activity (Fig. 5c). These patients with high LL-37 levels were significantly likely to have HBI reduction of 8 points or more in 6–18 months, with an odds ratio equal to 31 (Fig. 5d). Most of these high LL-37 patients reached clinical remission. On the other hand, the same group of CD patients with CRP levels 0.4 mg/L or below tended to have similar future clinical activity as those with CRP above 0.4 mg/L, and the difference between groups was not statistically significant (Fig. 5e). LL-37 was better than CRP in indicating future clinical activity of moderate or severe CD patients. There was no association between initial disease activity, LL-37 levels, and use of medication from the day of blood draw (0 month) throughout monitoring period (6–18 months, average =12 months) (Additional file 5: Figure S5A). Among the steroid-free CD patients, the high LL-37 group still had significantly greater reduction of HBI values than the low LL-37 group (8.8 ± 0.48 versus 6.3 ± 0.80) (Additional file 5: Figure S5B). The improved future clinical activity of the CD patients in the high LL-37 group did not appear to be influenced by use of any specific group of medication.

Among CD patients, the low LL-37 group had significantly higher relative risk (RR = 1.8) than the high LL-37 group in developing intestinal stricture (Table 2a). Meanwhile, the relative risk of developing stricture in high CRP groups was not elevated in comparison to low CRP groups (Table 2b). LL-37 was superior to CRP in reflecting the relative risk of intestinal stricture in CD patients. The details of prevalence, sensitivity, specificity, PPV, and NPV data of the LL-37 test (below 40 ng/ml) are shown in Table 2c.