Stool Interleukin-1β Differentiates Clostridioides difficile Infection (CDI) From Asymptomatic Carriage and Non-CDI Diarrhea

Abstract

Background

Despite advances in the understanding and diagnosis of Clostridioides difficile infection (CDI), clinical distinction within the colonization–infection continuum remains an unmet need.

Methods

By measuring stool cytokines and antitoxin antibodies in well-characterized cohorts of CDI (diarrhea, nucleic acid amplification test [NAAT] positive), non-CDI diarrhea (NCD; diarrhea, NAAT negative), asymptomatic carriers (ASC; no diarrhea, NAAT positive) and hospital controls (CON; no diarrhea, NAAT negative), we aim to discover novel biological markers to distinguish between these cohorts. We also explore the relationship of these stool cytokines and antitoxin antibody with stool toxin concentrations and disease severity.

Results

Stool interleukin (IL) 1β, stool immunoglobulin A (IgA), and immunoglobulin G (IgG) anti–toxin A had higher (P < .0001) concentrations in CDI (n = 120) vs ASC (n = 43), whereas toxins A, B, and fecal calprotectin did not. Areas under the receiver operating characteristic curve (ROC-AUCs) for IL-1β, IgA, and IgG anti–toxin A were 0.88, 0.83, and 0.83, respectively. A multipredictor model including IL-1β and IgA anti–toxin A achieved an ROC-AUC of 0.93. Stool IL-1β concentrations were higher in CDI compared to NCD (n = 75) (P < .0001) and NCD + ASC+ CON (CON, n = 75) (P < .0001), with ROC-AUCs of 0.83 and 0.86, respectively. Stool IL-1β had positive correlations with toxins A (ρ_A = +0.55) and B (ρ_B = +0.49) in CDI (P < .0001) but not in ASC (P > .05).

Conclusions

Stool concentrations of the inflammasome pathway, proinflammatory cytokine IL-1β, can accurately differentiate CDI from asymptomatic carriage and NCD, making it a promising biomarker for CDI diagnosis. Significant positive correlations exist between stool toxins and stool IL-1β in CDI but not in asymptomatic carriers.

Clostridioides difficile diagnostics cannot reliably distinguish infection from colonization. Stool IL-1β is >40-fold higher in CDI vs carriers and non-CDI diarrhea; correlating to stool toxin in CDI, not in carriers. Stool IL-1β, pivotal in CDI pathobiology, is a promising biomarker.

Clostridioides difficile is a leading nosocomial pathogen and a growing community-acquired infection worldwide [1, 2]. Despite its clinical importance, there is no broad consensus on the best approach to diagnosis [3]. Current guidelines support multiple approaches that can yield substantially different results [4, 5].

Multiple layers of data support that the pathobiology of Clostridioides difficile infection (CDI) is mediated by its main toxins A (tcdA) and B (tcdB). These activate the inflammasome via Toll-like receptor 9 (TLR9) signaling [6] and, via their glucosyl-transferase activity, permanently inactivate intracellular Rho-GTPases, leading to loss of cytoskeletal integrity and cell death [7].

The widespread adoption of nucleic acid amplification testing (NAAT), typically targeting the tcdB gene, led to improved sensitivity in detecting toxinogenic C. difficile. However, testing patients not meeting standard diarrhea definitions or having diarrhea for other reasons has led to incorrect classification of carriers as having CDI. Hence, approaches to arbitrate NAAT positivity by traditional enzyme immunoassays (EIAs) for toxin A/B—that can detect toxin levels ≥1000 pg/mL—gained popularity.

We previously reported that an ultrasensitive quantitative toxin immunoassay (USTIA) found no difference in median stool toxin A or B concentrations between NAAT-positive patients classified as CDI vs carriers [8]. Furthermore, a large proportion of asymptomatic carriers (ASC) had detectable toxin levels. Thus, the presence of C. difficile toxins, thought to be necessary for the CDI phenotype, is insufficient to define it in isolation. Our group hypothesized that the host immune response could distinguish between CDI and carriers. We found an elevation of serum granulocyte-colony stimulating factor (G-CSF) levels to be predictive of CDI vs carriage with an area under the receiver operating characteristic curve (ROC-AUC) of 0.842. Serum interleukin (IL) 4, IL-8, IL-10, IL-15, monocyte chemoattractant protein 1, and immunoglobulin G (IgG) anti–toxin A were significantly elevated in the serum of CDI patients compared to carriers [9].

While CDI can have prominent systemic manifestations, it remains primarily a colonic mucosal disorder. Therefore, studying immune markers in stool may more directly reflect the biological phenomena underlying CDI. Since current CDI diagnostics are stool-based, stool markers discriminating CDI from colonization could be easier to implement clinically.

We measured a panel of innate and adaptive cytokines as well as antitoxin antibodies in stool samples from 4 established cohorts. These cohorts included patients with CDI, patients with non-CDI diarrhea (NCD), and ASC, as well as hospital controls (CON). We aimed to discover new stool markers to differentiate CDI from the other cohorts, and to explore their relationship with stool C. difficile toxin concentrations and disease severity.

METHODS

Patient Cohorts

All cohorts included patients ≥18 years of age at Beth Israel Deaconess Medical Center (Boston, Massachusetts) or Texas Medical Center (Houston, Texas), under institutional review board–approved protocols. The CDI cohort consisted of patients with acute (or acute on chronic) diarrhea or pseudomembranous colitis with C. difficile tcdB NAAT positive and a decision to treat by a clinician, as previously described [10]. The NCD cohort consisted of patients with acute (or acute on chronic) diarrhea but NAAT negative. The ASC cohort consisted of patients admitted to the hospital having received non-CDI–directed antibiotics who were not having diarrhea and with positive NAAT, as previously described [8]. The CON cohort consisted of similar patients to the ASC cohort but NAAT negative (see Supplementary Methods). Patients with known inflammatory bowel disease (IBD) were excluded from all cohorts.

Data Collection

Demographic, clinical, and laboratory data at the time of clinical stool testing (CDI, NCD) or enrollment (ASC, CON) were obtained from detailed electronic chart review as described previously [8]. Severity by the Infectious Diseases Society of America (IDSA) and European Society for Clinical Microbiology and Infectious Diseases (ESCMID) criteria [4, 5] was determined by 2 experienced clinicians (N. W., J. A. V. G.). Severe clinical outcomes (intensive care unit [ICU] stay, colectomy, or death) and CDI recurrence were prospectively recorded during the 40 days following CDI diagnosis [10]. Development of CDI was prospectively recorded during 40 days after enrollment in the ASC cohort. Attribution of severe outcomes was retrospectively, and independently, assessed by 2 experienced clinicians (C. D. A., J. A. V. G.). A CDI expert (C. P. K.) resolved discrepancies. N. W., C. D. A., J. A. V. G., and C. P. K. were blinded to the patient's biomarkers during severity and attribution determinations. Only severe outcomes attributed primarily to CDI are referred to as CDI-attributed severe outcomes. Measurements of stool toxin (USTIA) and fecal calprotectin (fCP) were made as previously described [8, 9].

Biological Assays

The stool was processed (see Supplementary Methods), aliquoted, frozen, and stored at –80°C. Freshly thawed aliquots were used. The measurements of cytokine concentrations (IL-1β, IL-4, IL-6, IL-8, IL-10, IL-15, G-CSF, and tumor necrosis factor alpha [TNF-α]) were performed using a Milliplex magnetic bead kit and Luminex analyzer (MAGPIX) (Millipore Sigma, Burlington, Massachusetts) as per the manufacturer's instructions. Stool samples were analyzed for concentrations of antibodies (IgG, immunoglobulin A [IgA]) to C. difficile toxins A and B by a semi-quantitative enzyme-linked immunosorbent assay as previously described [11–13]. NAP-1 strain status was determined as previously described [14].

Statistical Analysis

Based on an unpublished small pilot study, 50 patients were needed in each cohort to detect a 2-fold difference in the mean stool concentration of 2 most promising biological markers at a power of 80% and 5% significance level. Biomarker distributions skewed positively, requiring nonparametric analysis, so we increased the goal sample size by 20%. Stool sample availability was fixed for ASC (n = 43) and CON (n = 50). Approximately 50% of CDI patients had severe disease by IDSA criteria in the parent cohort [8, 10]; thus, we aimed to include 60 patients with nonsevere disease and 60 patients with severe disease. We included all patients with CDI-attributed severe outcomes with an available stool sample, after which we randomly selected the remaining severe patients.

Distributions were compared by the Wilcoxon, Kruskal–Wallis, χ², and Fisher exact tests as indicated. The adjusted significance level for multiple testing (q*) was computed via the Benjamini–Hochberg and Bonferroni methods. Markers meeting the Benjamini–Hochberg–adjusted significance threshold (BHST) for differences across all cohorts underwent pairwise testing. ROC-AUCs were determined for markers meeting the BHST. Stepwise logistic regression was used to build multipredictor models, including only markers with a lower bound of the 95% confidence interval (CI) of the ROC-AUC ≥0.65. A multipredictor model of clinical interest was defined as one increasing the ROC-AUC by ≥0.05 compared to the best single marker. The Spearman test was used to assess correlations. The significance level was not adjusted for multiple testing for correlation or severity analysis due to their exploratory nature. See Supplementary Methods for further detail.

RESULTS

Characteristics of the Cohorts

The CDI cohort included 120 patients (59 met IDSA severity criteria). The remaining cohort sizes were as follows: NCD (n = 75), ASC (n = 43), CON (n = 50). There were no significant differences in age, sex, race, ethnicity, peak white blood cell count, and creatinine or albumin nadir across cohorts (Table 1). The NAP1/027 strain was observed more frequently in CDI vs ASC. Within 40 days of enrollment, 15 (12.5%) patients had a CDI-attributed severe outcome, 6 (5%) had a recurrent CDI episode, and 3 (6.9%) ASC patients developed CDI.

Table 1.

Demographic, Clinical, and Laboratory Characteristics of the 4 Study Cohorts

Characteristic	CDI (n = 120)	NCD (n = 75)	ASC (n = 43)	CON (n = 50)	P Value
Diarrhea	Yes	Yes	No	No
tcdB NAAT	Positive	Negative	Positive	Negative
Age, y, median (IQR)	65.5 (50.5–77)	65 (53–72)	64 (49–74)	63 (55–70)	.7933a
Female sex	64 (53.3)	32 (42.7)	22 (51.2)	17 (34)	.103b
Race					.093b
ȃWhite	85 (70.8)	41 (54.6)	30 (69.8)	36 (72)
ȃAfrican American	16 (13.2)	15 (20)	1 (2.3)	1 (2)
ȃAsian	7 (5.8)	5 (6.7)	2 (4.7)	1 (2)
ȃOther/mixed/unknown	12 (10)	14 (18.7)	10 (23.2)	12 (24)
Ethnicity					.185b
ȃLatino/Hispanic	9 (7.5)	8 (10.7)	1 (2.3)	1 (2)
Duration of diarrhea at diagnosis, d, median (IQR)	1 (1–3)	1 (.375–1.25)	…	…	.0001 a
Empiric CDI treatment in 48 h prior to stool sample	25 (20.8)	…	…	…
NAP1 027 strain	19 (19.6), n = 97	…	2 (4.7), n = 43	…	.022 b
Severe episode by IDSA criteria	59 (49.1)	…	…	…
Current episode is a recurrence	5 (4.2)	…	…	…
New CDI episode to occur within 40 d	6 (5)	…	3 (6.9)	…
CDI-attributed ICU stay, colectomy, or death within 40 d	15 (12.5)	…	…	…
WBC count, 109 cells/L, median (IQR)	11.4 (7–17.6)	8.8 (5.2–16.2)	11 (6.6–14.9)	12.5 (7.4–14.7)	.3146a
Creatinine, mg/dL, median (IQR)	1.1 (0.8–1.9)	1.2 (0.8–1.9)	1.2 (0.8–1.8)	1.2 (0.8–1.7)	.9609a
Albumin, g/dL, median (IQR)	3.2 (2.6–3.6)	3.4 (3–3.8)	3.4 (3.1–3.8)	3.3 (3–3.6)	.3015a

Data are presented as No. (%) unless otherwise indicated. Bolded P-values are significant (P < 0.05).

Abbreviations: ASC, asymptomatic carrier; CDI, Clostridioides difficile infection; CON, controls; ICU, intensive care unit; IDSA, Infectious Diseases Society of America; IQR, interquartile range; NAAT, nucleic acid amplification test; NCD, non–Clostridioides difficile diarrhea; WBC, white blood cell.

Kruskal–Wallis or Wilcoxon rank-rum tests.

Fisher exact test.

Differences in Stool Immune Markers Across Cohorts

Differences in the fecal immune markers’ distributions across all cohorts were observed (Table 2). Figure 1 shows stool IL-1β concentration distributions across cohorts; Supplementary Figures 2–4 display stool concentration distributions across study cohorts for IL-8, IgG anti–toxin A, and IgA anti–toxin A, respectively.

Table 2.

Stool Toxin and Immune Markers in the 4 Cohorts

Stool	CDI	NCD	ASC	CON	P Valuea	BH	BF
	Median (IQR), No.	Median (IQR), No.	Median (IQR), No.	Median (IQR), No.		(q* = .0447)	(q* = .0026)
Toxin A, pg/mL	101.1 (0–2974.5), 120	0 (0–0), 75	61.2 (0–450.7), 43	0 (0–0), 50	.0001	SIG	SIG
Toxin B, pg/mL	232.3 (0–6250.9), 120	0 (0–0), 75	43.3 (0–499), 43	0 (0–0), 50	.0001	SIG	SIG
IgA anti–toxin A, EU/mL	54.6 (17.1–104.5), 120	14.2 (5–95.3), 48	4.8 (1.8–17.3), 43	6.3 (2.1–33.9), 50	.0001	SIG	SIG
IgG anti–toxin A, EU/mL	5.9 (1.4–16.8), 120	1.2 (0.3–4.9), 48	0.7 (0.4–1.1), 43	0.9 (0.5–1.5), 50	.0001	SIG	SIG
IgA anti–toxin B, EU/mL	6.6 (2.6–17.1), 120	3.5 (1.7–9.5), 48	1.5 (0.7–5), 43	1.9 (0.9–6.3), 50	.0001	SIG	SIG
IgG anti–toxin B, EU/mL	0.3 (0–3.5), 120	0 (0–0), 48	0 (0–0.3), 43	0.1 (0–1.2), 50	.0015	SIG	SIG
G-CSF, pg/mL	2.4 (0.3–15.2), 120	0.3 (0.3–0.3), 75	0.3 (0.3–5), 43	0.3 (0.3–1.6), 50	.0001	SIG	SIG
IL-10, pg/mL	0 (0–1.4), 120	0 (0–0), 75	0.5 (0–1.5), 43	0.5 (0–2.5), 50	.0001	SIG	SIG
IL-15, pg/mL	0.9 (0.1–2.9), 120	0.1 (0.1–0.1), 75	0.1 (0.1–4.4), 43	0.1 (0.1–2.3), 50	.0001	SIG	SIG
IL-1β, pg/mL	26.6 (1.9–927.4), 120	0.6 (0.2–2.1), 75	0 (0–1.8), 43	0 (0–2.2), 50	.0001	SIG	SIG
IL-4, pg/mL	0.1 (0.1–3.2), 120	0.1 (0.1–0.1), 75	5.6 (0.8–12.7), 43	1.9 (0.8–7.9), 50	.0001	SIG	SIG
IL-6, pg/mL	0 (0–0.9), 120	0 (0–0), 75	0 (0–2.7), 43	0 (0–0), 50	.0003	SIG	SIG
IL-8, pg/mL	6.9 (0–102.8), 120	0 (0–0), 75	0 (0–2.4), 43	0.6 (0–2.9), 50	.0001	SIG	SIG
TNF-α, pg/mL	0.5 (0–4.7), 120	0 (0–0), 75	0 (0–0), 43	0 (0–0), 50	.0001	SIG	SIG
Fecal calprotectin, µg/g	185 (60–851.5), 43	75.8 (22.8–169.5), 26	168.4 (75.3–406.8), 42	171 (75.5–400.5), 28	.0177	SIG	NS

Abbreviations: ASC, asymptomatic carrier; BF, Bonferroni method; BH, Benjamini–Hochberg method; CDI, Clostridioides difficile infection; CON, controls; EU, enzyme-linked immunosorbent assay unit; G-CSF, granulocyte colony-stimulating factor; IgA, immunoglobulin A; IgG, immunoglobulin G; IL, interleukin; IQR, interquartile range; NCD, non–Clostridioides difficile diarrhea; NS, not statistically significant; q*, adjusted significance threshold; SIG, statistically significant; TNF-α, tumor necrosis factor alpha.

Kruskal–Wallis test.

Figure 1.

Stool concentrations of interleukin 1β across the 4 study cohorts. The y-axis is shown in logarithmic scale. *CDI vs NCD, P < .0001; **CDI vs ASC, P < .0001; ***CDI vs (NCD + ASC + CON), P < .0001. Abbreviations: ASC, asymptomatic carrier; CDI, Clostridioides difficile infection; CON, control; IL-1β, interleukin 1β; IQR, interquartile range; NCD, non–Clostridioides difficile diarrhea.

Stool Immune Markers Differentiate CDI From Asymptomatic Carriers

Median stool concentrations of IL-1β, IL-8, IL-15, TNF-α, G-CSF, IgA anti–toxin A, IgG anti–toxin A, and IgA anti–toxin B were higher in CDI than ASC; conversely, median stool concentrations of IL-4 were higher in ASC than CDI (Table 3). Toxins A and B and fCP concentrations did not differ significantly between the 2 cohorts, as previously described [9]. Every marker with significant differences was controlled for NAP1/027 strain positivity when strain typing was available (CDI = 97/120, ASC= 43/43). NAP1 strain was a covariable contributing to the significant differences between CDI and ASC groups in the stool concentrations of IgG anti–toxin A (P_NAP1 = .041), IL-4 (P_NAP1 = .044), IgA anti–toxin B (P_NAP1 = .043, P_{NAP1*(IgA anti–toxin B)} = .032), and G-CSF(P_NAP1 = .035). Differences between groups remained significant after accounting for the contribution of NAP1 strain.

Table 3.

Stool Immune Markers in the Clostridioides difficile Infection Cohort Compared With the Asymptomatic Carrier Cohort, Ordered From Highest to Lowest Area Under the Receiver Operating Characteristic Curve

Stool	CDI	ASC	P Valuea	BH	BF	AUC-ROC (95% CI)
	Median (IQR), No.	Median (IQR), No.		(q* = 0.03)	(q* = 0.0033)
IL-1βb, pg/mL	26.6 (1.9–927.4), 120	0 (0–1.8), 43	.0001	SIG	SIG	0.88 (.82–.92)
IgA anti–toxin Ab, EU/mL	54.6 (17.1–104.5), 120	4.8 (1.8–17.3), 43	.0001	SIG	SIG	0.83 (.76–.88)
IgG anti–toxin Ab, EU/mL	5.9 (1.4–16.8), 120	0.7 (0.4–1.1), 43	.0001	SIG	SIG	0.83 (.76–.88)
IL-4b, pg/mL	0.1 (0.1–3.2), 120	5.6 (0.8–12.7), 43	.0001	SIG	SIG	0.73 (.66–.80)
IgA anti–toxin Bb, EU/mL	6.6 (2.6–17.1), 120	1.5 (0.7–5), 43	.0001	SIG	SIG	0.72 (.65–.79)
IL-8b, pg/mL	6.9 (0–102.8), 120	0 (0–2.4), 43	.0001	SIG	SIG	0.71 (.64–.78)
TNF-αb, pg/mL	0.5 (0–4.7), 120	0 (0–0), 43	.0004	SIG	SIG	0.67 (.59–.74)
G-CSFb, pg/mL	2.4 (0.3–15.2), 120	0.3 (0.3–5), 43	.0114	SIG	NS	0.62 (.55–.70)
IL-15b, pg/mL	0.9 (0.1–2.9), 120	0.1 (0.1–4.4), 43	.0188	SIG	NS	0.61 (.53–.69)
Toxin A, pg/mL	101.1 (0–2974.5), 120	61.2 (0–450.7), 43	.5294	NS	NS	…
Toxin B, pg/mL	232.3 (0–6250.9), 120	43.3 (0–499), 43	.0712	NS	NS	…
IgG anti–toxin B, EU/mL	0.3 (0–3.5), 120	0 (0–0.3), 43	.2669	NS	NS	…
IL-10, pg/mL	0 (0–1.4), 120	0.5 (0–1.5), 43	.0408	NS	NS	…
IL-6, pg/mL	0 (0–0.9), 120	0 (0–2.7), 43	.7572	NS	NS	…
Fecal calprotectin, µg/g	185 (60–851.5), 43	168.4 (75.3–406.8), 42	.3629	NS	NS	…

Abbreviations: ASC, asymptomatic carrier; AUC-ROC, area under the receiver operating characteristic curve; BF, Bonferroni method; BH, Benjamini–Hochberg method; CDI, Clostridioides difficile infection; CI, confidence interval; EU, enzyme-linked immunosorbent assay unit; G-CSF, granulocyte colony-stimulating factor; IgA, immunoglobulin A; IgG, immunoglobulin G; IL, interleukin; IQR, interquartile range; NS, not statistically significant; q*, adjusted significance threshold; SIG, statistically significant; TNF-α, tumor necrosis factor alpha.

Wilcoxon rank-sum test.

ROC-AUC lower bound of 95% CI of .65.

IL-1β, IgA anti–toxin A, IL-4, IgG anti–toxin A, and IgA anti–toxin B met the threshold for inclusion into multipredictor models. Model iterations are in Supplementary Table 3. One multipredictor model met the prespecified threshold for clinical utility. This model includes IL-1β, IgA anti–toxin A, and their interaction term with ROC-AUC of 0.93 (95% CI, .87–.96) (Supplementary Figure 1).

Stool Immune Markers Differentiate CDI From NCD

All markers had distributions that differed significantly between the CDI and NCD cohorts (Supplementary Table 1). IL-1β, IL-4, IL-8, and TNF-α met the threshold for inclusion into multiple predictor model building. No multiple predictor model met the clinical interest threshold (Supplementary Table 3).

Stool Immune Markers Differentiate CDI From All Other Cohorts Combined

We tested the stool markers ability to distinguish CDI from the remaining 3 cohorts, to replicate the heterogeneity of patients tested for CDI in real-world clinical scenarios. IL-1β, IL-6, IL-8, IL-15, TNF-α, G-CSF, IgA anti–toxin A, IgG anti–toxin A, and IgA anti–toxin B differed significantly between the CDI cohort and the remaining cohorts (Supplementary Table 2). IL-1β, IgG anti–toxin A, IgA anti–toxin A and, IL-8 met the threshold for inclusion in multiple predictor models. No multipredictor model met the prespecified clinical interest threshold (Supplementary Table 3).

Toxins A and B Correlate With Select Stool Immune Markers in CDI but Not in Asymptomatic Carriers

Since toxins have been described to activate inflammatory pathways in animal and human ex vivo models, we proposed to study the correlations between toxins and immune markers in stool by cohort. Stool levels of IL-1β showed significant (P < .0001) positive correlations with stool levels of toxin A (ρ = +0.55) and toxin B (ρ = +0.49) in CDI, but this was not the case in in the ASC cohort (P_A = .98, P_B = .91), in spite of similar toxin concentrations (Figure 2).

Figure 2.

Correlations of stool biomarkers with stool toxins A and B by study cohort. Black cells: no significant correlation. Cell shading gradient: degree ± direction of correlation. Cell values: Spearman correlation coefficient (ρ). Abbreviations: ASC, asymptomatic carrier; CDI, Clostridioides difficile infection; G-CSF, granulocyte colony-stimulating factor; IgA, immunoglobulin A; IgG, immunoglobulin G; IL, interleukin; TNF, tumor necrosis factor.

Stool Immune Markers and Disease Severity

To test the correlation between CDI severity with stool immune markers, IL-1β, IgA, and IgG anti–toxin A were selected (highest ROC-AUCs for the CDI vs NCD + ASC+ CON). We also chose a marker under clinical use for severity prediction (fCP), and stool toxins A + B as determined by USTIA, which we have shown to correlate with severity and severe outcomes [10]. Stool IL-1β did not have a statistically significant association with initial severity by IDSA or ESCMID criteria. Median IL-1β stool concentrations were significantly higher, by 20-fold, in patients with severe CDI-attributed outcomes compared to those without them (P = .04) (Table 4). Stool IgG anti–toxin A was not associated with initial severity per IDSA criteria but was by ESCMID criteria as well as with CDI-attributed severe outcomes. Severity by IDSA and ESCMID criteria was significantly associated with higher median concentration of toxin A + B, as we previously described [10]; however, no significant association was observed between toxin A + B and CDI-attributed severe outcomes, unlike in the larger parent dataset. fCP did not show significant associations with severity or severe CDI-attributed outcomes.

Table 4.

Severity and Clostridioides difficile Infection–Attributed Severe Outcomes by Most Promising Stool Immune Markers (Interleukin 1β, Immunoglobulin A, and Immunoglobulin G Against Toxin A)

Stool Markers	Nonsevere/Absent, Median (IQR), No.	Severe/Present, Median (IQR), No.	P Valuea
Severity at diagnosis by IDSA	Nonsevere	Severe
ȃIL-1β, pg/mL	24.9 (1.9–1218.9), 61	31.1 (1.7–807.2), 59	.7688
ȃIgA anti–toxin A, U/L	40.9 (14.4–106), 61	69.8 (24.3–104), 59	.539
ȃIgG anti–toxin A, U/L	4.5 (1.1–17.3), 61	7.0 (2.5–16.4), 59	.3353
ȃCalprotectin, µg/g	182.5 (63.9–851.5), 26	218.5 (59.9–369.5), 17	.7091
ȃToxin A + B, pg/mL	76.2 (0.0–3057.5), 61	1374.7 (66.5–66 464.9), 59	.0007
Severity at diagnosis by ESCMID	Nonsevere	Severe
ȃIL-1β, pg/mL	7.0 (1.8–495.6), 63	250.8 (3.7–1370.4), 57	.0702
ȃIgA anti–toxin A, U/L	46.2 (14.1–104), 63	69.8 (26.4–105), 57	.2595
ȃIgG anti–toxin A, U/L	3.1 (1.1–11.2), 63	9.7 (2.9–24.4), 57	.0149
ȃCalprotectin, µg/g	177 (60–709.7), 29	232.7 (69.6–1001.1), 14	.5769
ȃToxin A + B, pg/mL	66.5 (0.0–4613.1), 63	1277.8 (154.1–19 127.8), 57	.0042
CDI primarily attributed severe outcomes (ICU, death, colectomy) within 40 d	Absent	Present
ȃIL-1β, pg/mL	22.6 (1.7–692.9), 105	454.7 (8.0–2368.3), 15	.0443
ȃIgA anti–toxin A, U/L	53.8 (16.2–104), 105	89.8 (24.3–106), 15	.4437
ȃIgG anti–toxin A, U/L	5.2 (1.2–12.9), 105	11.7 (5.4–101), 15	.0422
ȃCalprotectin, µg/g	180 (60–709.7), 41	653.4 (246.8–1059.9), 2	.2986
ȃToxin A + B, pg/mL	407.3 (0.0–7726.8), 105	1117.1 (180.6–129 192.6), 15	.0784

P values in bold are statistically significant.

Abbreviations: CDI, Clostridioides difficile infection; ESCMID, European Society for Clinical Microbiology and Infectious Diseases; ICU, intensive care unit; IDSA, Infectious Diseases Society of America; IgA, immunoglobulin A; IgG, immunoglobulin G; IL, interleukin; IQR, interquartile range.

P values obtained by the Wilcoxon rank-sum test.

DISCUSSION

Stool IL-1β Is a Promising Biomarker for Diagnosis of CDI

This study finds that patients with CDI have substantially higher median stool IL-1β concentrations (>40-fold) than patients with NCD and asymptomatic carriers. Furthermore, median IL-1β fecal concentrations at CDI diagnosis are 20-fold higher in patients who are to develop severe outcomes (colectomy, death, ICU admission) attributed to CDI within 40 days of diagnosis. This, combined with ROC-AUCs ranging from 0.83 to 0.88 to differentiate CDI from carriers, NCD, and a combination of both with antibiotic-exposed hospital controls, makes stool IL-1β a highly promising stool biomarker in CDI diagnosis and perhaps even outcome prognostication. Furthermore, IL-1β can be measured noninvasively in the same stool sample used clinically to determine the presence of C. difficile (via NAAT or USTIA).

A study in 1997 found increased fecal concentrations of IL-1β in severe CDI compared with mild CDI (n = 18, P = .0045) [15]. Another study in 2014 found a higher median fold change in stool levels of IL-1β of CDI patients vs NCD patients by a semi-quantitative proteome array method; however, differences were not statistically significant. This could be explained by a lower sensitivity of the assay. A similar, but statistically significant difference was found for IL-8, which is replicated in our study [16]. We are not aware of other clinical studies of fecal IL-1β in CDI. Studies of circulating IL-1β concentrations in CDI have had mixed results. A previous study from our own group using a subset from the same parent cohort did not find significant differences in serum IL-1β between the CDI, ASC, and NCD cohorts [9]. Another study found no difference in serum IL-1β between patients with CDI, non-CDI inpatients with diarrhea, or healthy outpatients [17], while a subsequent one found CDI and non-CDI patients to have higher serum IL-1β than healthy outpatients but no difference among each other [18]. Later publications found elevations in serum IL-1β in CDI as compared to healthy controls, but no difference between severe and nonsevere CDI [19]; while another study did find an approximately 2-fold elevation of serum IL-1β in severe vs nonsevere CDI [20]. Such heterogeneity could be explained by variability in the definitions of CDI, control groups, and severity as well as the sensitivity of the assays utilized.

Some studies have proposed using fCP in the diagnosis of CDI. fCP concentrations are reported to be approximately 2-fold higher in CDI vs NCD (consistent with our findings), but this is of limited utility since NAAT and toxin detection outperform fCP for this distinction and the fCP concentrations in CDI and ASC patients do not differ significantly [21]. Studies suggesting higher fCP concentrations in severe CDI compared to mild disease have had conflicting results [22, 23]. A more recent study found a 13-fold higher concentration of fCP in CDI inpatients compared to healthy outpatients classified as ASC by positive Glutamate Dehydrogenase EIA, which is inconsistent with our findings [24]. However, only 14% were NAAT positive. Hence, this “carrier” cohort is noncomparable to ASC.

IL-1β Is a Crucial Mediator of C. difficile Toxin–Induced Colonic Inflammation

IL-1β is a proinflammatory cytokine that is one of the downstream effectors of the TLR9-NFκB-NLRP3-ASC inflammasome cascade [25, 26]. Clostridioides difficile toxins tcdA and tcdB have been shown to induce IL-1β secretion, as well as intestinal inflammation and damage, in a murine model. Knockout of inflammasome ASC protein (apoptosis-associated speck-like protein containing a caspase recruitment domain) abolished tcdA- and tcdB-induced IL-1β secretion and decreased intestinal inflammation and damage. Similarly, pretreatment with the IL-1 receptor antagonist anakinra in tcdA and tcdB-challenged wild-type mice decreased intestinal inflammation and damage [27]. Incubation of human colonic explants with tcdA and tcdB led to increased IL-1β levels in the supernatant as compared to incubation with phosphate-buffered saline [28]. Our findings of elevated median IL-1β concentrations in CDI patients vs other cohorts and of correlations between stool tcdA and tcdB concentrations and stool IL-1β concentrations in CDI patients clinically validate the experiments above. Notably, despite having similar stool toxin concentrations to CDI patients, ASCs have lower median stool IL-1β concentrations and lack correlation between stool toxin and IL-1β concentrations. This raises the question of which mechanism could underlie resistance to toxin-induced inflammasome-mediated IL-1β release and colonic inflammation. Previous murine experiments showed that Saccharomyces boulardii culture supernatant inhibited tcdA-induced IL-1β secretion via inhibition of the Erk1/2 mitogen-activated protein kinase [29]; regulatory signals could come from the microbial and metabolic environment. Further study into regulatory mechanisms, particularly within the inflammasome pathways, may improve our understanding of the differences underlying CDI and symptomless carriage.

Stool Antitoxin Antibodies Remain of Interest in CDI

Median stool IgA and IgG against toxin A were elevated 6- to 11-fold in CDI compared to ASC or ASC + NCD + CON with ROC-AUCs ranging from 0.77 to 0.83. This supports the importance of the humoral immune response to tcdA as an immune biological feature of early CDI. While IgA is secreted into the colonic lumen via transcytosis, the transport of IgG antitoxin antibodies is facilitated by toxin-induced mucosal damage, as shown in murine and ex vivo human intestinal cell layer models, via a paracellular FcRn (neonatal Fc receptor)–independent manner [30]. Thus, higher stool IgG anti–toxin A may only reflect the degree of colonic inflammation, leading to permeability to large proteins in general.

The absence of a significantly higher median stool concentration of anti–toxin B IgG in CDI compared to ASC is puzzling, considering the findings for anti–toxin A IgA and IgG and anti–toxin B IgA. This apparent discrepancy is inconsistent with the current paradigm considering the humoral response to tcdB of clinical importance [31, 32], albeit better studied in the context of recurrent CDI. We hypothesize this may be due to analytical limitations in our anti–toxin B assay, in particular at the level of the capture antigen. The unclear specificity of antitoxin IgGs as markers of pathogen presence as opposed to markers of colonic permeability may limit their diagnostic role.

Strengths and Limitations

The use of a prospectively enrolled, well-characterized, stringently defined yet clinically relevant cohort is a major strength of our study. We made use of a novel USTIA with a clinical cutoff of 20 pg/mL for each toxin, approximately 5-fold lower than classic cytotoxicity assays and approximately 50-fold lower than standard EIA. The USTIA has a 5-log range of quantification, permitting previously impossible analyses and correlations. Strict adjustment of the significance threshold for multiple testing for each immune marker renders the possibility of biomarker false discovery low. Our findings’ consistency with previously published murine and in vitro human data also reinforces their validity.

Some limitations of our study should be considered. First, we excluded patients with IBD after noting that a subset of IBD patients in an unpublished pilot study had different immune marker elevation profiles compared to patients without IBD. While the distinction of C. difficile carriage from infection is an unmet need in IBD patients, our results cannot be generalized to them. We plan to investigate that question in a separate study. Second, our ASC cohort has inclusion and exclusion criteria carefully crafted to reflect patients at risk of CDI who did not develop the infection (inpatients, antibiotic-exposed, no treatment with CDI antibiotics). Nonetheless, the critical diagnostic challenge lies in distinguishing patients colonized with C. difficile who may have diarrhea from a different cause (ie, medications, other infections, dysmotility). Currently, there is no gold standard to classify such patients; therefore, we believe our cohort is the closest achievable proxy. From the analytical standpoint, the Milliplex kit and Luminex analyzer are not designed specifically for stool. However, similar technologies are used in stool to study other enteropathogens [33], cow’s milk protein allergy [34], and C. difficile itself [35].

The next steps include further validation of these findings in a larger cohort. Simultaneously, developing a sensitive multiplex assay (eg, fecal toxins and IL-1β) that can run on standard clinical platforms is vital to clinical adoption. It is also of interest to further explore the inflammasome pathway’s regulatory mechanisms in CDI and C. difficile colonization.

CONCLUSIONS

Stool concentrations of the inflammasome pathway cytokine IL-1β can accurately differentiate CDI from asymptomatic carriage and NCD, making it a promising biomarker for CDI diagnosis. Positive correlations between stool concentrations of toxins and IL-1β in CDI clinically validate previous preclinical data that demonstrated toxin-induced inflammasome-dependent secretion of IL-1β mediating colonic inflammation. The absence of this correlation in asymptomatic carriers despite stool toxin concentrations similar to CDI patients suggests that regulatory processes in the inflammasome pathway may be essential in determining the disease phenotype.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors acknowledge the patients participating in this study as well as all the nursing and laboratory staff at both Beth Israel Deaconess Medical Center and Texas Medical Center, who were essential to this effort.

Financial support. This work was supported by the National Institute of Allergy and Infectious Diseases (grant number R01-AI116596 to N. R. P. and C. P. K. and National Institutes of Health Loan Repayment Funding to C. D. A.); and the National Institute of Diabetes and Digestive and Kidney Diseases (award number T32-DK007760 to C. P. K. and J. A. V. G.). Ultrasensitive quantitative toxin immunoassays were provided as an in-kind service by bioMérieux.