Mortality in patients with Clostridium difficile infection correlates with host pro-inflammatory and humoral immune responses

Received 28 January 2013 Accepted 26 May 2013 School of Medicine and Medical Science, UCD Veterinary Sciences Centre, University College Dublin, Belfield, Dublin 4, Ireland Department of Medicine for the Older Person, Mater Misericordiae University Hospital, Eccles St, Dublin 7, Ireland Division of Gastroenterology, Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA 02215, USA Department of Clinical Microbiology, St Vincent’s University Hospital, Elm Park, Dublin 4, Ireland School of Public Health, Physiotherapy and Population Science, University College Dublin, Belfield, Dublin 4, Ireland


INTRODUCTION
C. difficile infection (CDI) is the most common cause of antibiotic-associated diarrhoea and colitis and is associated with a wide range of symptoms, from mild diarrhoea to fulminant colitis and death.Other risk factors, in addition to antimicrobial exposure, that are implicated in the development of CDI include advancing age (.65 years), prolonged hospital stay and severity of underlying disease (Cohen et al., 2010;Kyne et al., 2000;Pe ´pin et al., 2004;Thibault et al., 1991).The host anti-toxin immunoglobulin response has been shown to play an important role in mediating the outcome of colonization with C. difficile (Kyne et al., 2000), influencing the duration of disease (Warny et al., 1994) and determining the risk of recurrence (Kyne et al., 2001;Warny et al., 1994).High serum antitoxin IgG has demonstrated a degree of protection from severe, systemic CDI in animals (Johnson, 2012;Steele et al., 2012), but the direct relationship between the host anti-toxin immunoglobulin response in severe CDI and protection from fatal outcome has yet to be defined.
The association between ribotype 027 strain type and CDI severity under non-outbreak conditions is controversial (Cloud et al., 2009;Wilson et al., 2010); therefore, the relationship between strain type, disease severity and cytotoxin production remains to be determined.
We conducted a prospective cohort study of CDI cases to investigate host systemic IgG anti-toxin immune responses and the relationship to CDI disease severity and risk of mortality.We also investigated the role of in vitro cytotoxicity of the infecting C. difficile strain in influencing host inflammatory and immune responses.This study was approved by the local research ethics committees and consent was obtained from all participants.Patient demographic, clinical and laboratory data were collected, including maximum total peripheral WCC (610 9 l 21 ), serum CRP (mg l 21 ), minimum serum albumin (g l 21 ), peak serum creatinine level (mmol l 21 ), Charlson comorbidity scores and Horn's index of severity of admitting diagnosis (Charlson et al., 1987;Horn et al., 1983).The most extreme value for these markers in the 72 h following the positive stool C. difficile toxin test was used for analysis.

METHODS
The primary outcome was all-cause 30-day mortality rate.Secondary outcomes were recurrence ¡60 days after resolution of primary diarrhoeal episode or mortality directly attributable to CDI as determined from death certificate data.It was also noted where CDI was recorded as a contributory but not the leading cause of death.
Stool samples were collected at diagnosis and C. difficile was isolated after alcohol shock and anaerobic culture on cycloserine-cefoxitin fructose selective agar (BD Biosciences) at 37 uC for 48 h.
Venous blood (10 ml) was drawn on days 1, 3 and 12 post-diagnosis into an anticoagulant-free vacuum tube and immediately centrifuged at 3000 g for 5 min.Serum was stored at 280 uC.
C. difficile strain typing.Genomic DNA was purified from C. difficile after overnight anaerobic growth in brain-heart infusion (BHI) broth (Oxoid), supplemented with 1 % yeast extract and 0.1 % (w/v) L-cysteine (Sigma-Aldrich) at 37 uC.DNA was extracted using a Wizard Genomic DNA Purification kit (Promega) and assessed for concentration and purity using a Nanodrop 2000 (Thermo Scientific) before use in PCR.
Isolates were characterized by PCR ribotyping as described previously (Stubbs et al., 1999) and ribotypes were assigned by comparison with representatives from the Leiden University Medical Centre library, Leiden, The Netherlands.Isolates were toxinotyped as described previously (Rupnik et al., 1998).The presence of virulence factorspecific genes, encoding binary toxin (cdtB) and the 18/39 bp deletion in the regulatory gene tcdC, known to be associated with particular ribotypes, was also assessed by molecular methods, as described previously (Cohen et al., 2000).
Quantification of anti-TcdA and -TcdB IgG concentration.Levels of anti-TcdA and -TcdB IgG in patient sera were measured by direct ELISA using immobilized TcdA or TcdB proteins (provided by C. P. Kelly, Harvard Medical School, MA, USA) by a previously published method (Kyne et al., 2000).
The relative antibody concentration in arbitrary ELISA units (EU) for each sample was defined by extrapolation from an internal control standard curve (absorbance versus dilution factor) generated from pooled unused sera of arbitrary high, medium and low antibody levels.Peak anti-TcdA and -TcdB IgG concentrations within 12 days of CDI diagnosis were calculated for each patient.
Evaluation of bacterial cytotoxin titre.Bacterial isolates were grown overnight (16-18 h) in BHI broth supplemented with 1 % yeast extract and 0.1 % L-cysteine at 37 uC under anaerobic conditions to an optical density at 600 nm of 2.0.Broth culture supernatants were passed through a 0.2 mm filter and tenfold serially diluted (10 1 -10 8 -fold dilutions) in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10 % FBS (Sigma-Aldrich) and 1 % (v/v) L-glutamine/penicillin/streptomycin solution (final concentrations of 2 mM L-glutamine, 100 U penicillin ml 21 and 100 mg streptomycin ml 21 ; Sigma-Aldrich).The broth dilutions were then applied in triplicate to confluent Vero (African green monkey kidney) cell monolayers (LGC Standards) and incubated for 24 h at 37 uC in 5 % CO 2 .The Vero cell monolayers were then assessed by inverted light microscopy (Nikon; Unitech) for cell rounding.The cytotoxin titre was defined as the log 10 value of the broth supernatant dilution factor that was able to induce 50 % cell rounding (10 1 -10 8fold dilutions; cytotoxin titre 1-8) as described in a previously published method (Kamiya et al., 1989).
Broth supernatants that did not induce cell rounding were classified as cytotoxin titre 0. C. difficile toxin specificity was confirmed by neutralization with Clostridium sordellii anti-toxin (C.difficile Toxin/ Antitoxin kit; TechLab).
Statistical analysis.To assess whether the median anti-TcdA or -TcdB IgG titres were significantly different in cases of recurrence versus non-recurrence or high versus low bacterial cytotoxin titre, the Wilcoxon rank-sum test was used.Student's t-test was used to compare mean cytotoxin titres between ribotype 027 and nonribotype 027 isolates.Correlations between cytotoxin titre and clinical markers of inflammation were analysed using Pearson's correlation co-efficient (r).Correlations between cytotoxin titre and outcome frequencies were analysed using the point-biserial correlation coefficient (r pb ).
Univariate analysis of the risk factors for 30-day all-cause mortality were carried out using Student's t-test for parametric means, the Wilcoxon rank-sum test for non-parametric medians and a x 2 / Fisher's exact test for categorical outcome frequencies.

C. difficile isolation and molecular typing
Stool samples were collected from 128 of 150 patients and C. difficile was isolated from the stool and assigned a ribotype and toxinotype for 93 patients.A fully characterized C. difficile strain, full clinical data (co-morbidity scores, laboratory markers of inflammation and CDI outcome) and at least two serum samples within 12 days post-diagnosis were available for 86 patients (mean age 74.5, range 35-97 years), who comprised our final study population.
Ribotype 027 isolates were identified as either toxinotype IIIa (n56) or toxinotype IIIb (n52).Ribotype 078 and ribotype 017 isolates were identified as type V and type VIII, respectively.All other toxinotyped isolates were identified as type 0.

Host serum anti-TcdA and -TcdB IgG response
Of the patients infected with a toxinogenic C. difficile strain, nine (10.7 %) did not have detectable serum anti-TcdA IgG and four (4.8 %) did not have detectable anti-TcdB IgG within the 12 days after CDI diagnosis.
The relationship between patient age and level of anti-TcdA and -TcdB IgG response was assessed and no difference in mean age was found between patients exhibiting low and high serum anti-TcdA IgG titres (first quartile, 74.2 years, vs fourth quartile, 74.9 years) or between patients with low and high serum anti-TcdB IgG titres (first quartile, 74.7 years, and fourth quartile, 73.4 years).

Relationship between host serum anti-TcdA and -TcdB IgG response and outcome
The overall recurrence and 30-day all-cause mortality rates were 18.6 and 16.3 %, respectively.Of the 14 patients who died by day 30, death was directly attributable to CDI for seven patients (8.1 % overall) and was recorded as a contributory factor for five patients (5.8 % overall).The remaining two patients died of complications from an additional infection or after gastrointestinal surgery.
Patients who had a recurrence within 60 days of CDI diagnosis had no significant difference in median anti-TcdA IgG titre than patients who did not have a recurrence [31 EU (range 0.8-176.4),IQR 81.8, versus 16.1 EU (range 0-123.2),IQR 54].There was also no significant difference in anti-TcdB IgG titre between patients with and without recurrence [14.7 EU (range 4.6-107.1),IQR 62.3, versus 18.0 EU (range 0-159.3),IQR 55.7].The number of patients for whom death could be attributed to CDI was too low for statistical comparison.

Determination of bacterial cytotoxin titre
The majority of isolates in this study were toxigenic and induced a cytopathic effect, apart from two isolates (ribotype 010 and 001) that did not generate amplicons after toxinotyping PCR.
To analyse the association between bacterial cytotoxin titre and host anti-toxin immune responses, we classified patients into two categories according to whether they were infected with a strain demonstrating a low (titre ,3) or high (titre 4 or 5) cytotoxin titre.The majority of patients studied (n575; 87 %) were infected with a C. difficile strain exhibiting a low in vitro cytotoxin titre.

Correlation between bacterial cytotoxin titre and clinical inflammatory markers
To investigate whether there was a correlation between the in vitro cytotoxin titre of the infecting bacterial strain and the severity of the host inflammatory response during infection, patients were grouped according to the cytotoxin titre of the infecting isolate and mean clinical markers of inflammation were calculated (Table 1).
All CDI patients in this study demonstrated hypoalbuminaemia indicated by serum albumin levels below 3.9 g dl 21 .
There was a significant positive correlation between bacterial cytotoxin titre and CRP level (r50.28;P,0.05), peak WCC (r50.183;P,0.05) and minimum albumin concentration (r50.214;P,0.05), indicating that a rise in the in vitro cytotoxin titre of the infecting strain was associated with an increase in patient inflammatory responses.Patients in the highest cytotoxin titre group (cytotoxin titre 5) presented with inflammatory markers The mean (±SD) cytotoxin titre of ribotype 027 isolates was significantly higher than the mean of all other non-ribotype 027 isolates in this study (P,0.00001).The statistical difference between the mean cytotoxin titre of ribotype 027 isolates versus non-ribotype 027 isolates was calculated using Student's t-test.Twenty-one of the 86 patients (24.4 %) presented with a high serum creatinine level (.133 mmol l 21 , 1.5 mg dl 21 ), a marker of renal impairment.Serum creatinine levels did not, however, correlate significantly with cytotoxin titre.

Association between bacterial cytotoxin and outcome
Recurrence, 30-day all-cause mortality and attributable death did not correlate significantly with cytotoxin titre (r pb 50.12, 20.08 and 20.03, respectively).When patients were categorized depending on whether they were infected with low or high cytotoxin strains, 12 out of 16 patients (75 %) that exhibited a recurrence of CDI were infected with a strain of low cytotoxin titre.Of the patients who died by day 30, 13 out of 14 (93 %) were infected with a strain of low cytotoxin titre; however; there was no significant association between cytotoxin titre and recurrence or 30-day all-cause mortality overall.All patients for whom death was directly attributable to CDI were infected with a strain of low cytotoxin titre; however, this was also not statistically significant.
Multivariable analysis, controlling for age and underlying disease severity (Horns index .3),was carried out on risk factors that were significantly associated with 30-day allcause mortality by univariate analysis.

DISCUSSION
The host anti-toxin immunoglobulin response has been shown to play an important role in influencing the outcome of C. difficile colonization (Kyne et al., 2000), Host immune response to C. difficile infection duration of CDI disease (Warny et al., 1994) and risk of recurrence (Kyne et al., 2001;Warny et al., 1994).Cases of severe CDI are increasing and the role of the host antitoxin immunoglobulin response in severe CDI and protection from mortality has yet to be defined.
We have shown for the first time that patients exhibiting low serum levels of IgG antibody to TcdA were significantly more likely to die within 30 days of CDI diagnosis.Patients who died within 30 days of CDI diagnosis also exhibited lower anti-TcdB IgG levels.The importance of high serum anti-toxin IgG levels in protecting from mortality due to severe CDI disease has been demonstrated in animals and protection was shown to be mediated by a reduction in systemically disseminated toxins in the serum (Johnson, 2012;Steele et al., 2012).A role for systemic toxaemia in severe human CDI has not yet been established.Future studies are needed to investigate toxaemia in severe CDI using freshly collected sera and highly sensitive toxin assays (He et al., 2009;Steele et al., 2012).This will allow an evaluation of whether the protection associated with high serum anti-toxin IgG levels is due to a reduction in systemic toxaemia (in addition to neutralization of colonic luminal or mucosal toxins).Quantification of toxaemia may also determine whether low anti-TcdA levels are more closely associated with mortality than anti-TcdB IgG levels due to differences in serum concentrations of the respective toxins.
It is important to note that the anti-toxin IgG response mounted by CDI patients was independent of age and co-morbidity.The phenomenon of 'immunosenescence' suggests that elderly patients are less able to mount an immune response to infection, which was not evident in this study, although the majority of patients studied were elderly, which may bias this finding.In addition, we did not assess the neutralizing ability of anti-TcdA and -TcdB IgG from the patients in this cohort in an in vitro cytotoxicity assay.Although this study was limited by an assessment of IgG quantity but not of functional neutralizing activity, we have shown a significant relationship between IgG levels and protection from mortality.
We also investigated the role of in vitro cytotoxicity of the infecting C. difficile strain in influencing host inflammatory and immune responses.Isolates of ribotype 027 in our study population exhibited a significantly higher in vitro cytotoxin titre than all other ribotypes, which may support claims of the 'hypervirulent' status of this strain with respect to cytotoxin production (Kuehne et al., 2010;Thelestam & Chaves-Olarte, 2000).However, some isolates of ribotypes 014 and 001 also exhibited high cytotoxin titres and induced acute inflammatory responses.The TcdC negative regulator has previously been considered responsible for hyper-production of toxins; however, these isolates were negative for mutations/deletions in the tcdC gene and strains of ribotype 078 that contained a 39 bp deletion in tcdC exhibited comparatively low cytotoxin titres.The data presented here therefore support the suggestions of others that the role of TcdC in influencing toxin production is called into question (Bakker et al., 2012;Cartman et al., 2012).
We found that patients infected with strains that exhibited a high cytotoxin titre were significantly more likely to have higher anti-TcdA and -TcdB IgG levels and were more likely to have clinical markers of severe disease, including peak WCC .20610 9 l 21 and CRP concentrations .230mg l 21 .This suggests that both the host adaptive and innate immune systems respond sensitively to C. difficile toxin antigens and initiate a response that is reflective of the toxin challenge posed by the infecting strain.This may explain why the majority of patients who were infected with a strain of high cytotoxin titre and exhibited clinical markers of severe disease were still alive 30 days after CDI diagnosis, as these patients were most likely to have a high protective anti-toxin IgG response.
In our study, factors that were significantly associated with an increased risk of mortality after CDI diagnosis were host-associated, including underlying disease severity (Horn's index), host clinical markers of inflammation and host anti-toxin IgG levels.Of the three non-toxin bacterial-associated factors investigated here (ribotype 027, tcdC deletion and binary toxin), none was significantly associated with mortality.This suggests that laboratory assays that specifically detect the presence of these markers may be of limited use in identifying patients at risk of a fatal outcome of CDI.The outcome of CDI may also be influenced by the host response to other C. difficile antigens, including flagellin, surface-layer proteins and cysteine proteases; however, these were not investigated in this study.
High cytotoxin titre was not a risk factor for 30-day mortality, and the majority of CDI patients overall (87 %) were infected with a strain exhibiting a low cytotoxin titre.This suggests that patients are highly sensitive even to low toxin titres.It is also possible that other non-humoral aspects of the host immune response to C. difficile, including inflammatory cytokine release, vascular leakage and neutrophil migration, may be contributory risk factors for 30-day mortality.
On multivariate analysis, three independent predictors of mortality were identified: high peak WCC, high serum creatinine and low anti-TcdA IgG titre.The data in our study therefore suggest that prediction of the outcome of CDI should be host orientated, by monitoring laboratory markers of inflammation and clinical assessment of patient health.The association between high anti-toxin IgG titre and protection from mortality as shown here provides additional supportive evidence for the development and therapeutic use of active and passive immunotherapies for the management of CDI.
Patient clinical details and sample collection.One hundred and fifty consecutive patients with CDI [positive stool test using Meridian ImmunoCard Toxins A/B (Meridian BioScience) and diarrhoea not attributable to any other cause] were recruited during 2008 and 2009 at two hospitals in Dublin, Ireland.

Fig. 1 .
Fig.1.C. difficile ribotype 027 isolates exhibit significantly higher in vitro cytotoxin titres than isolates of other ribotypes.(a) The in vitro cytotoxin titre was calculated for each clinical isolate by log 10 dilution of bacterial supernatant able to induce 50 % rounding of Vero cells after 24 h incubation at 37 6C.The mean (±SD) was calculated for the most prevalent ribotypes (n¢3) isolated from patients in this study.(b) The mean (±SD) cytotoxin titre of ribotype 027 isolates was significantly higher than the mean of all other non-ribotype 027 isolates in this study (P,0.00001).The statistical difference between the mean cytotoxin titre of ribotype 027 isolates versus non-ribotype 027 isolates was calculated using Student's t-test.

Fig. 2 .
Fig. 2. Box-and-whisker plot showing that patients infected with C. difficile isolates that exhibit high in vitro cytotoxin titres have significantly higher anti-toxin IgG responses.Patients were categorized into two groups based on whether they were infected with C. difficile isolates of high (4 or 5) or low (0-3) in vitro cytotoxin titre.The median (horizontal line) ±upper quartile and lower quartile (upper and lower boxes) anti-TcdA (open boxes) and -TcdB (shaded boxes) IgG titres were calculated for each group.The whiskers represent maximum and minimum values and outlying data points are indicated by 6.The statistical difference between the median anti-TcdA or -TcdB IgG titre for patients infected with an isolate of high versus low in vitro cytotoxin was calculated using the Wilcoxon rank-sum test.

Table 1 .
Correlation between in vitro bacterial cytotoxin titre of the infecting strain and host inflammatory markers Cytotoxin titres are shown as mean values, with SD in parentheses.For ribotype, the number of isolates for each ribotype is shown in parentheses.The number of isolates is shown for each titre value