key: cord-0007211-2xlxjq0e
authors: Eisen, Damon P.; Dean, Melinda M.; Boermeester, Marja A.; Fidler, Katy J.; Gordon, Anthony C.; Kronborg, Gitte; Kun, Jürgen F. J.; Lau, Yu Lung; Payeras, Antonis; Valdimarsson, Helgi; Brett, Stephen J.; Ip, W. K. Eddie; Mila, Joan; Peters, Mark J.; Saevarsdottir, Saedis; van Till, J. W. Oliver; Hinds, Charles J.; McBryde, Emma S.
title: Low Serum Mannose-Binding Lectin Level Increases the Risk of Death due to Pneumococcal Infection
date: 2008-08-15
journal: Clin Infect Dis
DOI: 10.1086/590006
sha: 0223acccc42f22041b56648a386d0f043322bf3d
doc_id: 7211
cord_uid: 2xlxjq0e

Background. Previous studies have shown associations between low mannose-binding lectin (MBL) level or variant MBL2 genotype and sepsis susceptibility. However, MBL deficiency has not been rigorously defined, and associations with sepsis outcomes have not been subjected to multivariable analysis. Methods. We reanalyzed MBL results in a large cohort with use of individual data from 4 studies involving a total of 1642 healthy control subjects and systematically defined a reliable deficiency cutoff. Subsequently, data were reassessed to extend previous MBL and sepsis associations, with adjustment for known outcome predictors. We reanalyzed individual data from 675 patients from 5 adult studies and 1 pediatric study of MBL and severe bacterial infection. Results. XA/O and O/O MBL2 genotypes had the lowest median MBL concentrations. Receiver operating characteristic analysis revealed that an MBL cutoff value of 0.5 ≪g/mL was a reliable predictor of low-producing MBL2 genotypes (sensitivity, 82%; specificity, 82%; negative predictive value, 98%). MBL deficiency was associated with increased likelihood of death among patients with severe bacterial infection (odds ratio, 2.11; 95% confidence interval, 1.30–3.43). In intensive care unit–based studies, there was a trend toward increased risk of death among MBL-deficient patients (odds ratio, 1.58; 95% confidence interval, 0.90–2.77) after adjustment for Acute Physiology and Chronic Health Enquiry II score. The risk of death was increased among MBL-deficient patients with Streptococcus pneumoniae infection (odds ratio, 5.62; 95% confidence interval, 1.27–24.92) after adjustment for bacteremia, comorbidities, and age. Conclusions. We defined a serum level for MBL deficiency that can be used with confidence in future studies of MBL disease associations. The risk of death was increased among MBL-deficient patients with severe pneumococcal infection, highlighting the pathogenic significance of this innate immune defence protein.

molecule in the lectin complement pathway has been scrutinized for association with risks of infectious and inflammatory diseases, ischemic heart disease, and cancer. Serum MBL level and function varies substantially as a function of multiple polymorphisms of the MBL2 genes. Low MBL levels and impaired function are predominantly seen in the presence of structural gene polymorphism homozygosity and the LX promoter haplotype [1] . However, currently, there is no standard for MBL deficiency based on a direct functional assessment [7] genotypes alone to define MBL deficiency. Therefore, it is important to introduce more certainty into this field through a systematic assessment of MBL levels in a large, multinational population of subjects. MBL deficiency, variably defined, has been associated with increased susceptibility to many infectious diseases [8] . Invasive pneumococcal disease appears to be associated with low MBL levels [9, 10] , although there have been some conflicting results [6, 11] . Severe bacterial infection and sepsis have also been found to occur more frequently in MBL-deficient patients. The development of systemic inflammatory response syndrome in critically ill patients without infection has been shown to be more common in those with MBL deficiency [12] . Again, these studies have used differing criteria for the assessment of MBL deficiency, including presence of MBL2 polymorphism alone [6, 13] , MBL levels alone [14] , and both presence of MBL2 polymorphism and MBL levels [12, 15] ; other studies have included an assessment of MBL function [10] . In some of these studies of sepsis, univariate associations between low MBL levels [14, 16] or low-producing MBL2 haplotypes [15] and increased risk of death were present.

By analyzing MBL levels in healthy control populations from multiple studies, we accurately determined the range of MBL levels associated with each of the various MBL2 genotypes. Using this information, we determined the MBL level that most reliably predicts the presence of low-producing MBL2 genotypes. Furthermore, we applied this definition of MBL deficiency to a large group of patients with severe bacterial infection to assess the impact of this deficiency state on outcomes in patients with sepsis.

Definition of MBL deficiency with use of healthy control subjects. The published literature was searched with use of the PubMed database to identify studies in which MBL levels had been measured by ELISA and in which MBL2 genotypes (promoter and structural haplotype) had been determined. Investigators identified by this process were invited to participate in a collaborative study. Individual patient data were collated from 1642 healthy subjects from 4 studies (table 1) [1, [17] [18] [19] . Three other healthy control populations including 501 subjects with fully characterized MBL2 genotype and MBL levels are described elsewhere [21] [22] [23] . Data from these other studies were not made available for this analysis. MBL levels were grouped according to MBL2 genotypes and compared with use of the Kruskal-Wallis test. Receiver operating characteristic analysis was then performed using MBL levels as a predictor of the lowest-producing MBL2 genotypes. The suggested cutoff value for MBL deficiency was found by maximizing the sensitivity and specificity of the MBL value for predicting low-producing MBL2 genotypes. Comparison of the frequency of MBL deficiency based on MBL level and MBL2 

Published studies of the association between MBL deficiency and bacterial infection or sepsis were identified by search of the PubMed database, and investigators were invited to share individual patient data. These studies [6, 10-12, 16, 20] included 675 patients for whom MBL levels and/or MBL2 genotypes had been determined. Specifically, MBL levels were measured in 558 patients (table 1) . Blood samples for MBL assays were obtained on the day of diagnosis of sepsis [6, 10, 20] or within 48 h [11, 12, 16] after the diagnosis. MBL levels were determined by double-sandwich ELISA in all studies other than the Dutch peritonitis series [20] in which mannan-binding ELISA was used. The patients in this study of combined data were classified as being MBL deficient or replete with use of the MBL serum level cutoff derived from the preceding analysis of healthy control subjects. Data on the specific organism causing sepsis were present for 482 patients; 287 of these patients had bacteremia. Data on the presence or absence of comorbidities were recorded for a total of 326 patients. These included information on the presence of chronic obstructive pulmonary disease, hepatopathy, cancer, diabetes, and HIV infection in 264 patients. An additional 62 patients had comorbidities classified on the basis of the McCabe score [24] . The quality of the genetic associations reported by these studies was assessed on the basis of published criteria [25] . Data for 975 patients included in 3 other studies of MBL and sepsis [13] [14] [15] were not made available for this study.

Categorical variables were compared using the x 2 test. Student's t test and the Mann-Whitney log-rank test were used to compare normally and nonnormally distributed variables, re-spectively. Meta-analysis and binary logistic regression was used for multivariable analysis to determine the association between MBL deficiency and death. Statistical analyses were performed in Minitab, release 14, and Stata, version 9.

Although there were statistically significant differences in MBL serum levels among all of the MBL2 genotype groupings ( , by Kruskal-Wallis log-rank test), the most markedly P p .02 decreased levels and the least intragroup variation were seen in patients with genotypes XA/O and O/O (figure 1). The MBL levels in patients with the YA/O genotype were clearly intermediate between these obviously low-producing MBL2 genotypes and the high-producing MBL groups of YA/YA, YA/XA, and XA/XA. The XA/O and O/O groups were chosen to definitively represent the MBL2-deficient genotypes. Receiver operating characteristic analysis revealed that MBL level was a reliable predictor of these MBL2 genotypes (area under the receiver operating characteristic curve, 0.90) (figure 2). An MBL serum level of 0.50 mg/mL maximized sensitivity and specificity for predicting low-producing MBL2 genotypes (both were 82%). This cutoff had a positive predictive value of 31% and a negative predictive value of 98%. This level is proposed as a useful indicator of MBL deficiency, particularly for screening purposes, because of its high negative predictive value.

Using this MBL cutoff level, we determined the prevalence of MBL deficiency in the control populations studied and compared this with the prevalence as indicated by low-producing MBL2 genotypes (table 1) . For all of the cohorts of healthy control subjects that were included in this study, the rates of Among the studies of MBL and severe bacterial infection that were incorporated in the combined analysis, the median genetic association study quality scores [25] were 7 in 2 casecontrolled studies and 6 in the 3 cohort studies; these scores were assessed as being of moderate-to-high quality. These median quality scores are higher than those found in a systematic review of studies of genetic association and sepsis [25] . The most common deficiencies of the studies used in the review were the absence of power calculations and Hardy-Weinberg equilibrium analysis. A power calculation for the present study was performed assuming a 25% mortality rate and a frequency of MBL deficiency of 30%. This mortality rate was chosen because it is midway between the reported mortalities for pneumococcal bacteremia [26] and severe sepsis [27] , and the majority of this cohort consists of patient groups with these conditions. Assuming a relative risk of death of 2 among MBL-deficient patients, 168 patients would be required to achieve 80% power to detect a difference between groups at a P value of .05. We were able to evaluate survival among 477 patients who had their MBL levels determined; this is substantially more than the number required to power this observation. Overall, there were 123 deaths among 591 patients (mortality, 21%) for whom information on in-hospital survival relating to the episode of severe infection was available.

The association between serum MBL level deficiency and death in infected patients was determined by x 2 test on crude, pooled data. There was an increase in the rate of MBL deficiency among patients with severe bacterial infection who died (45 of 94 patients who died were MBL deficient), compared with those who survived (116 of 383 patients who survived were MBL deficient; OR, 2.11; 95% CI, 1.30-3.43;

). Meta-anal-P p .001 ysis, which adjusted for varying prevalences of serum MBL level deficiency and death in the individual studies, showed a similar effect, but the 95% CI did not indicate statistical significance (OR, 1.44; 95% CI, 0.91-2.3; ) (figure 3). A funnel plot P p .12 indicated that there was no reporting bias in the studies of MBL infection outcome ( figure 4 ). This funnel plot involved only the studies included in this reanalysis, because the frequency of serum MBL level deficiency could not be ascertained from the other studies of MBL and sepsis in which individual patient data were not made available [13] [14] [15] . Multivariable analysis of defined risk factors was performed for the 454 patients from studies in which APACHE II scores were calculated [10, 16, 20] . After adjustment for APACHE II score, age, and sex, serum MBL level deficiency was associated with a trend toward increased likelihood of death among infected patients (OR, 1.58; 95% CI, 0.90-2.77) (table 2) .

Because the largest subset of patients with severe bacterial infection was the subset with pneumococcal disease (186 of 675 patients), this was analyzed separately. The majority of these patients (171 of 186 patients) came from 2 studies that concentrated exclusively on S. pneumoniae infection [6, 11] but in which APACHE II scores were not calculated. In addition, although Fine scores [28] were calculated for 99 of the 186 patients with pneumococcal disease, this disease severity index could not be used in the multivariable analysis of factors influencing death, because it was calculated for an insufficient number of patients who died ( ) to be incorporated. n p 5 Therefore, the presence of bacteremia and patient comorbidities were used instead to assess independent risk factors for death. Multivariable analysis, adjusted for comorbidity, bacteremia, age, and sex, revealed that, in patients with pneumococcal infection, MBL deficiency increased the risk of death (table 2) . 

We studied a large cohort of healthy subjects with individual data drawn from 4 studies in which both MBL level and promoter and/or structural MBL2 genotype were characterized. The size of the cohort ( ) made it possible to generalize n p 1642 with far greater certainty than with previous studies about the MBL levels to be expected in those with the lowest-producing MBL2 genotypes. Assessment of the MBL levels found in each of the MBL2 genotype groups revealed that the interquartile ranges in patients with XA/O and O/O genotypes are substantially lower than 0.50 mg/mL and that this level coincides with the median in patients with the YA/O genotype. Furthermore, the lower quartile (0.28 mg/mL) in patients with the YA/O approximates the upper quartile (0.39 mg/mL) in patients with the XA/O genotype. The patients with the other MBL2 genotypes (YA/YA, YA/XA, and XA/XA) had median MBL levels that were clearly higher than this cutoff. On the basis of these findings, we propose that patients with the XA/O or O/O genotype can be considered to be clearly MBL deficient. It would be inappropriate to use the O/O group alone to determine MBL deficiency because of the very substantial overlap with patients with the XA/O genotype. Subsequently, receiver operating characteristic analysis confirmed that an MBL level !0.50 mg/mL is a valid indicator of the presence of MBL2-deficient genotypes.

It is important to appreciate the large variation in MBL serum levels seen in each MBL2 genotype group. These nonparametrically distributed values all ranged to 0 mg/mL. This is the obvious explanation for the observation that MBL2 genotype alone is a poor indicator of the presence of low MBL levels and, therefore, of MBL deficiency. Although our study revealed that a serum MBL level !0.5 mg/mL is a valid indicator of the presence of the low-producing MBL2 genotypes XA/O and O/O, such low levels have also been found in a significant number of persons with alternative genotypes. This accounts for the discrepancy in the frequency of MBL deficiency between the MBL level and genotype definitions explored in this and other studies.

The association between MBL deficiency and severe bacterial infection convincingly demonstrated in this reanalysis of a large group of patients from previous studies confirms previous observations [10, 12, 13, 15, 16] . A trend that approached statistical significance was found between MBL deficiency and death in patients with severe bacterial infection, with use of APACHE II score to adjust for disease severity and with the incorporation of patient risk factors, such as age and sex in multivariable analysis. This extends previous findings of univariate associations between MBL deficiency and death due to sepsis [14] [15] [16] .

Conflicting reports on the association between invasive pneumococcal disease [6, 9] and MBL deficiency also seem to be resolved by this combined analysis. For the first time (to our knowledge), MBL deficiency was shown to increase the risk of death among patients with pneumococcal infection. Although APACHE II and Fine scores could not be used to adjust for pneumococcal disease severity, bacteremia-which has a major impact on mortality [29] -and comorbidity could be incorporated in the multivariable analysis. This finding suggests that patients with pneumococcal sepsis may be an important group in which to investigate the potential benefits of MBL therapy.

MBL binds to microbial cell surface sugars arranged in dense spatial arrays referred to as pathogen-associated molecular patterns. MBL binds to pneumococcus [30] , but this may be abrogated in the presence of a polysaccharide capsule [31] . It must be recognized that the MBL pathway is a minor contributor to complement-mediated defence against S. pneumoniae, compared with the classic pathway [32] . MBL does, however, also contribute to bacterial killing by noncomplement mechanisms, such as opsonophagocytosis. Pathogens other than S. pneumoniae were isolated in samples from fewer patients in our study than was S. pneumoniae. Specifically, the smaller numbers of patients with S. aureus infection may explain the absence of an apparent significant association between MBL deficiency and death in these patients because of insufficient statistical power (multivariable analysis data not shown). MBL appears to play a role in the response to S. aureus infection, binding to the organism and precipitating C4 deposition with increased neutrophil phagocytosis [33] . Compelling evidence from a MBL knockout mouse model revealed reduced survival in S. aureus-infected MBL-null animals [34] . MBL deficiency appears to predispose to meningococcal sepsis [35] , another severe infection caused by an encapsulated organism. This clinical association is also supported by investigations of the ability of MBL to bind to and facilitate killing of Neisseria meningitidis [36] . There was only a small number of patients with N. meningitidis infection in the current study, because patients from a previous large study did not have MBL levels or MBL2 promoter haplotypes determined [35] .

The most important limitation of the present study is the heterogeneity of both the control and infected populations. Nonetheless, the large sample collection from multiple racial and geographic sources has provided an opportunity to demonstrate robust associations between MBL and predisposition to severe infection. Differences in serum MBL level testing methodologies may interfere with comparison of results. The basic laboratory methodology used for all results from the healthy control population that were collated here was the same methodology used with the double anti-MBL sandwich antibody ELISAs performed. Finally, differences may exist in the nature of the serum standards used to produce standard curves and to derive results.

The nature of the acute-phase MBL response was studied as part of a number of the sepsis studies included here [11, 16, 37] . These studies reveal that, at most, the variation in MBL levels is of a factor of 2-3 fold, with one study showing the same proportions of septic patients having increased, decreased, and stable MBL levels [37] . MBL did not behave as an acutephase reactant when it was studied in the context of pneumococcal infection [11] . In the context of surgical injury, MBL levels were not shown to alter substantially, in marked comparison to the known acute-phase reactants CRP and IL-6 [38] . It is unlikely, therefore, that a substantial decrease in MBL levels as a result of consumption from sepsis is the cause of the association between low MBL level, sepsis, and poor outcome. This is particularly pertinent to the association demonstrated between MBL deficiency and death in patients with pneumococcal sepsis. Although changes in MBL have been correlated with the type of organism causing sepsis, it has been shown that MBL is consumed in sepsis due to gram-negative, but not gram-positive, organisms [39] .

MBL may be extracted from plasma or produced by recombinant means for clinical use in deficient or septic patients. Phase I studies of MBL infusion have been performed [40] . MBL has a long serum half-life, and there is limited experience of its use for cystic fibrosis [41] and other chronic infections. Recombinant human MBL replacement therapy is now being studied in clinical trials involving patients receiving allogeneic stem cell transplantation and liver transplantation. The data from the large patient cohort that we report suggest a serum level that reliably indicates MBL deficiency. It is clear from this data that low MBL level is a more reliable indicator of deficiency than is the presence of low-producing MBL2 genotypes. This rationally derived MBL serum level cutoff for deficiency could be used to guide replacement therapy with MBL. Because death in the context of severe bacterial infection, especially when due to pneumococcus, was associated with MBL deficiency after adjustment for bacteremia, age, and comorbidity, severe bacterial pneumonia could be an important future target for study of MBL therapy. Mortality associated with bacteremic pneumococcal infection has remained largely unchanged, despite 50 years of advancements in supportive care. Treatment with an innate, immune-based therapy may provide a new opportunity to reduce this substantial disease burden.

Analysis of the relationship between mannose-binding lectin (MBL) genotype, MBL levels and function in an Australian blood donor population

The role of mannose-binding lectin (MBL) in paediatric oncology patients with febrile neutropenia

Variable mannose-binding lectin expression during postoperative acute-phase response

Messias-Reason I. Mannan-binding lectin plasma levels in leprosy: deficiency confers protection against the lepromatous but not the tuberculoid forms

Mannose-binding lectin plasma levels and gene polymorphisms in Plasmodium falciparum malaria

Variant mannose-binding lectin alleles are not associated with susceptibility to or outcome of invasive pneumococcal infection in randomly included patients

Susceptibility to HIV infection and progression of AIDS in relation to variant alleles of mannosebinding lectin

Impact of mannose-binding lectin on susceptibility to infectious diseases

MBL genotype and risk of invasive pneumococcal disease: a case-control study

Low mannose-binding lectin function is associated with sepsis in adult patients

Mannose-binding lectin does not act as an acute-phase reactant in adults with community-acquired pneumococcal pneumonia

Increased incidence and severity of the systemic inflammatory response syndrome in patients deficient in mannose-binding lectin

Polymorphisms in CD14, mannose-binding lectin, and Toll-like receptor-2 are associated with increased prevalence of infection in critically ill adults

Van den Berghe G. Intensive insulin therapy exerts antiinflammatory effects in critically ill patients and counteracts the adverse effect of low mannose-binding lectin levels

Association of mannose-binding lectin polymorphisms with sepsis and fatal outcome, in patients with systemic inflammatory response syndrome

Mannose-binding lectin polymorphisms in severe sepsis: relationship to levels, incidence, and outcome

Mannose-binding lectin in severe acute respiratory syndrome coronavirus infection

Association of a new mannosebinding lectin variant with severe malaria in Gabonese children

Mannan-binding lectin and complement C4A in Icelandic multicase families with systemic lupus erythematosus

Mannose-binding lectin deficiency facilitates abdominal yeast infection in patients with secondary peritonitis

Association of mannose-binding lectin gene (MBL2) polymorphisms with rheumatoid arthritis in an Indian cohort of case-control samples

Different molecular events result in low protein levels of mannan-binding lectin in populations from southeast Africa and South America

Mannose binding lectin (MBL) genotype distributions with relation to serum levels in UK Caucasoids

Gram-negative bacteraemia. I. Etiology and ecology

A systematic review of the quality of genetic association studies in human sepsis

Early predictors of mortality in pneumococcal bacteraemia

Adultpopulation incidence of severe sepsis in Australian and New Zealand intensive care units

A prediction rule to identify lowrisk patients with community-acquired pneumonia

Re-evaluation of the therapy of severe pneumonia caused by Streptococcus pneumoniae

Binding of mannan-binding protein to various bacterial pathogens of meningitis

Effect of capsulation of opportunistic pathogenic bacteria on binding of the pattern recognition molecules mannan-binding lectin, L-ficolin, and H-ficolin

The classical pathway is the dominant complement pathway required for innate immunity to Streptococcus pneumoniae infection in mice

Enhancement of complement activation and opsonophagocytosis by complexes of mannose-binding lectin with mannose-binding lectin-associated serine protease after binding to Staphylococcus aureus

Mannose-binding lectin-deficient mice are susceptible to infection with Staphylococcus aureus

Association of variants of the gene for mannose-binding lectin with susceptibility to meningococcal disease. Meningococcal Research Group

Mannose-binding lectin enhances phagocytosis and killing of Neisseria meningitidis by human macrophages

Mannose binding lectin acute phase activity in patients with severe infection

Influence of major surgery on the mannan-binding lectin pathway of innate immunity

Involvement of complement pathways in patients with bacterial septicemia

Human plasma-derived mannose-binding lectin: a phase I safety and pharmacokinetic study

Mannose-binding lectin (MBL) therapy in an MBL-deficient patient with severe cystic fibrosis lung disease

We thank the UK National Institute of Health Research Biomedical Research Centre Funding Scheme.Potential conflicts of interest. D.P.E. is the inventor on the use patent application "A medical protocol" for hypersupplementation with mannosebinding lectin for the treatment and prevention of sepsis. M.M.D. is listed as an inventor on the patent WO/2003/090774 "Mannose-binding lectin and uses thereof" but has not received any financial interest from this patent. A.C.G. holds stock options in Sirius Genomics. All other authors: no conflicts.