Clinical and prognostic associations of autoantibodies recognizing adrenergic/muscarinic receptors in patients with heart failure

Abstract Aims The importance of autoantibodies (AABs) against adrenergic/muscarinic receptors in heart failure (HF) is not well-understood. We investigated the prevalence and clinical/prognostic associations of four AABs recognizing the M2-muscarinic receptor or the β1-, β2-, or β3-adrenergic receptor in a large and well-characterized cohort of patients with HF. Methods and results Serum samples from 2256 patients with HF from the BIOSTAT-CHF cohort and 299 healthy controls were analysed using newly established chemiluminescence immunoassays. The primary outcome was a composite of all-cause mortality and HF rehospitalization at 2-year follow-up, and each outcome was also separately investigated. Collectively, 382 (16.9%) patients and 37 (12.4%) controls were seropositive for ≥1 AAB (P = 0.045). Seropositivity occurred more frequently only for anti-M2 AABs (P = 0.025). Amongst patients with HF, seropositivity was associated with the presence of comorbidities (renal disease, chronic obstructive pulmonary disease, stroke, and atrial fibrillation) and with medication use. Only anti-β1 AAB seropositivity was associated with the primary outcome [hazard ratio (95% confidence interval): 1.37 (1.04–1.81), P = 0.024] and HF rehospitalization [1.57 (1.13–2.19), P = 0.010] in univariable analyses but remained associated only with HF rehospitalization after multivariable adjustment for the BIOSTAT-CHF risk model [1.47 (1.05–2.07), P = 0.030]. Principal component analyses showed considerable overlap in B-lymphocyte activity between seropositive and seronegative patients, based on 31 circulating biomarkers related to B-lymphocyte function. Conclusions AAB seropositivity was not strongly associated with adverse outcomes in HF and was mostly related to the presence of comorbidities and medication use. Only anti-β1 AABs were independently associated with HF rehospitalization. The exact clinical value of AABs remains to be elucidated.


Introduction
Many studies have demonstrated the intricate relationship between immune activation and pathophysiological mechanisms in heart failure (HF). 1 Nevertheless, direct immunomodulatory interventions have as of yet no established role in the clinical management of HF.
Antibodies are soluble forms of the B-cell receptor that are secreted by activated plasma cells and constitute an important part of the adaptive immune response. 2 The generation of an effective immune response requires a broad antibody repertoire, which is achieved with impressive efficiency by recombination of immunoglobulin genes. 3 This permits the generation of antibodies with potentially infinite specificities. Unavoidably, this process also leads to the generation of antibodies against self-antigens. Specialized regulatory mechanisms have evolved to eliminate self-reactive lymphocytes; 4 nevertheless, these do not always function optimally. Accordingly, self-reactive lymphocytes and autoantibodies (AABs) feature prominently in numerous autoimmune rheumatic diseases 5 and other disease states. 6 A relationship between AABs against autonomic nervous system receptors (ANS-AABs) and HF has been described previously, particularly in patients with non-hereditary dilated cardiomyopathy (DCM) 7 and Chagas cardiomyopathy. 8 However, to our knowledge, most reports are limited to small cohorts, and no data from large, well-characterized, and diverse populations currently exist in the literature. We thus aimed to investigate the prevalence of ANS-AAB seropositivity in patients with HF and to determine its clinical associations and potential relationships with outcomes.

Patients
We performed a post hoc analysis of the BIOSTAT-CHF index cohort (n = 2516), which has been described previously. 9 Briefly, BIOSTAT-CHF was a multi-centre study enrolling patients from 11 European countries. Participants were aged ≥18 years and had symptoms of new-onset or worsening HF, combined with a left ventricular ejection fraction (LVEF) ≤ 40% or brain-type natriuretic peptide (BNP) and/or N-terminal pro-BNP (NT-proBNP) plasma levels > 400 or >2000 pg/mL, respectively. Participants either were not previously treated with angiotensinconverting enzyme inhibitors/angiotensin receptor blockers (ACEi/ARB) and/or β-adrenoreceptor blockers (BB) or were receiving ≤50% of guideline-recommended target doses and anticipated their initiation or uptitration. All patients were treated with loop diuretics. Participants could be enrolled as inpatients or outpatients. The primary outcome was a composite of all-cause mortality and rehospitalization for HF censored at 2-year follow-up, with each component separately constituting a secondary outcome. The study protocol was approved by local and national ethics committees (EudraCT 2010-020808-29; R&D Ref Number 2008-CA03; MREC Number 10/S1402/39), and all participants provided written informed consent. The study was performed according to the principles outlined in the Declaration of Helsinki.

Autoantibody measurements
Measurements of AAB titres against β1-, β2-, and β3-adrenergic receptors, as well as M2-muscarinic receptors, were performed using chemiluminescence immunoassays (ImmunometriX, Berlin, Germany) on blood serum samples from 2256/2516 (89.7%) patients from the BIOSTAT-CHF index cohort and in 299 controls that self-identified as healthy [acquired commercially from in.vent Diagnostica (Berlin, Germany)]. Assay development and calculation of binding indices are described in the Supplementary material online, Methods. For quality control of the anti-β1 AAB assay, one negative sample and one positive sample from the control group (binding index below the median and higher than 10, respectively) were measured in duplicate in all microtitre plates during analysis (n = 26 duplicates each). As there is no prior knowledge on the threshold for relevant AAB concentrations, the 99th percentile value of negative controls (1.3883) was used to define a cut-off under which an assay would be considered definitively negative (henceforth 'seronegative'), while the 1st percentile value of positive controls (7.3818) was similarly used to define a cut-off above which an assay would be considered definitively positive (henceforth 'seropositive') (see Supplementary material online, Figure S1). Patients with assay results in between the two cut-off values were labelled 'intermediate'. The same cut-offs were also applied for defining subgroups in the remaining AAB assays.

Laboratory indices
The estimated glomerular filtration rate (eGFR) was calculated using the modification of diet in renal disease (MDRD) formula. NT-proBNP and hs-cTnT were measured using sandwich immunoassays (Roche Inc.), and CRP was measured using competitive immunoassays on a Luminex platform (Alere Inc.). Anaemia was defined according to the World Health Organization definition.

Biomarker measurements
Plasma biomarkers were determined using the proximity extension assay technology (Olink Proteomics Inc.), as part of four biomarker panels involving 92 biomarker measurements each (Cardiovascular-II, Cardiovascular-III, and Immune Response and Oncology II panels), thus 368 biomarkers in total. The complete list of available biomarkers has been reported previously. 10 Overlapping biomarkers between the four panels were amphiregulin, c-kit ligand, and tissue factor pathway inhibitor-2 (measured twice) and interleukin-6 (IL-6) (measured three times). For overlapping biomarkers, the mean of all measurements was used, leaving a total of 363 unique biomarkers. Additionally, eight biomarkers were excluded from analyses because >10% of measurements were outside the assay's limit of detection. Thus, in total, 355 biomarkers were available for further analysis.

Pathway over-representation analysis
Pathway over-representation analysis of the 355 biomarkers was performed using the 'GProfiler' pathway analyser (version e104_eg51_p15_ 3922dba). 11 The results of the analysis were classified based on the gene ontology (GO) classification of biological processes (annotation 2022-01-13). 12 Corrections for multiple testing were performed using the built-in g:SCS algorithm, using a false discovery rate threshold of 5%. Only significantly over-represented processes that included ≥5 of their constituents were included in the final selection. The biomarker corneodesmosin could not be analysed by GProfiler, and since the biomarkers BNP and NT-proBNP share the same protein designation, the effective number of analysed biomarkers was eventually 353.

Selection of GO biological processes related to B-lymphocytes and antibodies
The pathway over-representation analysis of the 353 biomarkers yielded 768 over-represented biological processes. Of these, 13 contained the terms 'B cell' and 'immunoglobulin'. To eliminate potential data redundancies due to member overlap between processes, 10 the 13 selected processes were visualized in a directed acyclic graph, and the most distal non-redundant processes were selected ( Figure 1A, Figure 1B, Supplementary material online, Figure S2). This yielded five processes related to B-lymphocytes and antibodies, represented by different combinations of 31 biomarkers ( Figure 1C).

Statistical analysis
Statistical analyses were conducted using R-studio (R version 4.2.0). Normality of continuous variables was determined by visual inspection of histograms and Q-Q plots. Baseline characteristics were compared for each of the four investigated AABs between seronegative, intermediate, and seropositive patients. Normally distributed variables were compared using one-way analysis of variance (ANOVA), non-normally distributed continuous variables using Kruskal-Wallis tests, and categorical/binary variables using chi-square tests. In case of a significant ANOVA result, pairwise post hoc testing with independent t-tests using Bonferroni correction was performed. Chemiluminescence assay results for each AAB were compared between patients with HF and healthy controls when considering the corresponding binding index as continuous variable using the Mann-Whitney U test and when classified based on the previously defined groups by using chi-square tests. Associations between AAB status and the primary outcome and all-cause mortality were investigated using Kaplan-Meier curves, while associations with hospitalization were investigated using cumulative incidence function curves while considering all-cause mortality a competing risk. Univariable associations were investigated with log-rank tests, while multivariable corrections for previously published risk models for this cohort 9 were performed using Cox regression, when statistical assumptions were met. These included, amongst others, age, comorbidities, measures of congestion, NT-proBNP, and renal function. 9 Statistical testing in survival analyses was performed using the log-rank test. Statistical significance was considered as P ≤ 0.05, considering the exploratory character of this study.
Significant findings from survival analyses were further verified with propensity score matching (MatchIt package v. 4.5.0) based on the aforementioned variables of the BIOSTAT-CHF risk model. Seropositive patients were matched 1 : 1 to seronegative patients. A nearest-neighbour matching algorithm with a caliper of 0.2 and without replacement, taking into account the region of common support, was used for this purpose. 13 Unmatched observations were discarded. Statistical and graphical balance diagnostics tools were performed in order to evaluate the matching process.
To investigate the relative state of B-lymphocyte function in seronegative and seropositive patients, principal component analyses (PCA) were conducted for each AAB, by excluding patients that were categorized as intermediate. Subsequently, the results were plotted on biplots with concentration ellipses for each group.

Comparison of assay-binding indices in healthy controls and patients with heart failure
Assay-binding indices for each AAB are presented as continuous variables using half-violin plots with superimposed boxplots and individual data points in Figure 2A. There were no statistically significant differences between healthy controls and patients with HF in the median binding index for anti-

Classification of autoantibody status
The relative prevalence of seronegative, intermediate, and seropositive status amongst healthy controls and patients with HF is presented in Figure 2B. The prevalence of seropositivity amongst patients with HF in increasing order was 1.8% (n = 41) for anti-β3, 5.1% (n = 114) for anti-β1, 5.9% (n = 133) for anti-β2, and 7.6% (n = 172) for anti-M2 AABs. The proportion of groups was significantly different between patients with HF compared with healthy controls only for anti-M2 AABs, driven by a higher proportion of seropositive patients (P = 0.025). In total, 382 (16.9%) patients with HF were classified as seropositive for at least one AAB, compared with 37 (12.4%) of the healthy controls (P = 0.045). Summary statistics for assay-binding indices in total and stratified by AAB status are presented in Supplementary material online, Tables S1 and S2, respectively. The co-occurrence of either intermediate or seropositive status between the different AABs is presented in Figure 2C, while the cooccurrence of seropositive status only is presented in Figure 2D. Amongst the 382 patients that were seropositive for at least one AABs, co-occurrence of seropositivity for any of the remaining AABs had a low prevalence, with the highest value being observed for seropositivity for all four AABs in 15/382 (3.9%) patients. Collectively, 340/382 (89.0%) patients were seropositive for only one of the studied AABs.

Anti-β1 autoantibodies
Anti-β1 AAB seropositivity was associated with a lower prevalence of a pri-

Anti-β2 autoantibodies
Anti-β2 AAB seropositivity was associated with a history of stroke [129 Seropositive diabetics were also less likely to be using oral hypoglycaemic agents (P = 0.042).

Anti-β3 autoantibodies
Anti-β3 AAB seropositivity was associated with a higher prevalence of at-

Survival analysis
Kaplan-Meier survival curves are presented for each AAB in Figure 3A for the combined outcome and all-cause mortality alone ( Figure 3B) at 2-year follow-up. Cumulative incidence function curves for the competing risk analysis of HF rehospitalization at 2-year follow-up with all-cause mortality as a competing risk are presented in Figure 3C. Regarding the primary outcome, anti-β1 seropositive patients trended towards a poorer prognosis, without reaching statistical significance (P = 0.068). Anti-β3 status was associated with the primary outcome, with seropositive patients having an overall better prognosis (P = 0.016). Anti-β2 and anti-M2 status were not significantly associated with the primary outcome. There were no significant associations between AAB status and all-cause mortality. Competing risks analysis revealed only a significant association between   Continued anti-β1 seropositivity and a higher probability of HF rehospitalization (P = 0.029). A sensitivity analysis was performed after merging seronegative and intermediate patients because these groups showed comparable results in previous survival analyses. This revealed that the primary outcome and HF rehospitalization alone were more frequent only in patients who were seropositive for anti-β1 AABs (P = 0.023 and 0.008, respectively) ( Figure 4).
Since the two significant comparisons for anti-β1 AABs in the sensitivity analysis also met the criteria for the proportionality of hazards assumption, additional multivariable analyses were performed (Cox regression for the combined outcome and competing risk regression for rehospitalization).  Heart Association; BMI, body mass index; HR, heart rate; SBP/DBP, systolic/diastolic blood pressure; 6MWT, 6-min walk test; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro-brain natriuretic peptide; CRP, C-reactive protein; Hb, haemoglobin; BB, β-adrenoreceptor antagonist; ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; MRA, mineralocorticoid receptor antagonist. *P ≤ 0.05; **only for patients with diabetes mellitus; †no significant between-group differences in post hoc testing.
A B C Figure 3 Kaplan-Meier curves for the combined outcome (all-cause mortality and rehospitalization for HF) censored at 2 years (A) and for all-cause mortality alone (B), as well as cumulative incidence function curves for rehospitalization for HF only with all-cause mortality as a competing risk (C). From left to right, results of univariable analyses for anti-β1, anti-β2, anti-β3, and anti-M2 autoantibodies. HF, heart failure. Additional validation was performed by means of propensity score matching as described in the Methods section. All patients that were anti-β1 seropositive except one (n = 113) were matched 1 : 1 to seronegative patients, forming well-balanced groups (standardized mean differences < 0.1, variance ratios < 2). Seropositive patients had significantly higher risk for HF rehospitalization at 2-year follow-up (standardized HR (95% CI) 1.83 (1.08-3.10), P = 0.024). Because patients with intermediate AAB status seemed to have a different prognosis, particularly in the case of anti-β3 AABs (Figure 3), additional analyses were performed in patients with intermediate AAB status for all AABs, in order to investigate whether the AAB titre as a continuous variable was associated with any of the examined outcomes (see Supplementary material online, Table S3). Intermediate patients for anti-β3 AABs showed an inverse association between anti-β3 AAB titre and the combined outcome [HR (95% CI): 0.82 (0.71-0.94) for each one-unit increase, P = 0.006], which remained significant after adjustment for the corresponding risk model [HR (95% CI): 0.83 (0.72-0.95), P = 0.009]. In patients with intermediate anti-β2 status, AAB titres were associated with reduced all-cause mortality [HR (95% CI): 0.83 (0.72-0.95), P = 0.008]. However, this was no longer significant after multivariable correction for the corresponding risk model (P = 0.196). Patients with intermediate status for anti-β1/anti-M2 AABs did not show associations between their corresponding AAB titres and any examined outcomes.

Analysis of circulating B-lymphocyte-associated markers in relation to autoantibody seropositivity
PCA was used on the 31 B-lymphocyte-related biomarkers ( Figure 5). Concentration ellipses were plotted for seropositive and seronegative individuals per AAB. All sub-analyses demonstrated considerable overlap A B C Figure 4 Sensitivity analyses of the results presented in Figure 3, by stratifying categories as seronegative/intermediate and seropositive. Kaplan-Meier curves for the combined outcome (all-cause mortality and rehospitalization for HF) censored at 2 years (A) and for all-cause mortality alone (B), as well as cumulative incidence function curves for rehospitalization for HF only with all-cause mortality as a competing risk (C ). For anti-β1 autoantibodies, Cox regression analyses for the combined outcome and competing risks analyses for HF-related rehospitalization (both univariable and multivariable) are presented in the Results section. HF, heart failure. between seropositive and seronegative patients. The relative contribution of the 31 biomarkers to the first and second principal components is presented in Supplementary material online, Figure S3.

Discussion
In this post hoc analysis of a large and heterogeneous cohort of patients with HF, 16.9% of patients were seropositive for at least one ANS-AAB. Overall, the prevalence of seropositivity did not significantly differ from that of healthy controls, with the exception of a significantly higher prevalence of seropositivity for anti-M2 AABs. Seropositivity for ANS-AABs was associated with the presence of comorbidities (renal disease, COPD, AF, and stroke) or ARB/ MRA use. Only seropositivity for anti-β1 AABs was significantly associated with a higher probability of HF rehospitalization at 2-year follow-up. PCA revealed considerable overlap of B-lymphocyte activity between seropositive and seronegative patients with HF, based on 31 circulating biomarkers.
Most studies evaluating ANS-AABs in patients with HF have been conducted in patients with Chagas cardiomyopathy and non-hereditary DCM, mostly focusing on anti-β1 and anti-M2 AABs. Approximately a third of patients with asymptomatic Chagas disease have detectable anti-β1, anti-β2, and anti-M2 AABs, while almost all who develop cardiomyopathy are seropositive for all aforementioned AABs. 8 Amongst patients with non-hereditary DCM, anti-β1 and anti-M2 AABs were detected in 31% and 36-39% of patients, respectively. 14 Interestingly, immunoglobulin subclasses may also be relevant, as in patients with DCM, the presence of IgG3-anti-β1 AABs was associated with a better LVEF during follow-up compared with seronegative patients, or those with non-IgG3-anti-β1 AABs. 15 Furthermore, an increased prevalence of ANS-AAB seropositivity has been reported in women with peripartum cardiomyopathy (PPCM) compared with healthy pregnant controls. 16 In patients with non-hereditary DCM, anti-β1 AABs are also associated with an increased risk of ventricular tachycardia and sudden cardiac death, 17 while anti-M2 AABs predict rehospitalization but not all-cause mortality in PPCM. 18 In our study, we only identified an independent association between seropositivity for anti-β1 AABs with HF rehospitalization at 2-year follow-up, similar to a previous study in a smaller cohort. 19 This might be explained by counterbalance of anti-β1 AAB actions by anti-β2 AABs, 20 although their cooccurrence was limited. Conversely, seropositivity for anti-β3 AABs trended towards an association with better prognosis. However, the low seroprevalence of anti-β3 AABs made additional analyses difficult.  Continued ANS-AABs can potentially influence cardiac function by causing inappropriate positive or negative chronotropic or inotropic effects, by inducing cardiomyocyte apoptosis, or by activating the complement cascade. 14,[21][22][23] Other authors reported that anti-β1 AABs from patients with DCM enhanced T-cell proliferation via PKA/MAPK-signalling, which was prevented by metoprolol. 24 They also reduced interferon-γ production, while increasing IL-4 production by T-cells. 24 Anti-β1 AABs also impaired endothelial function in Wistar rats by negatively affecting NO signaling. 25 In contrast, anti-β3 AABs from patients with HF protected rats with abdominal aortic banding from developing LV dysfunction and dilation. 26 Anti-β1 and anti-M2 AABs have been found to induce HF in rodents, which could be reversed upon their removal. 27,28 This suggests a drug-like function on their corresponding target receptors. 28 However, the pharmacology governing these interactions is as of yet unclear 28 and the agonist or antagonist characteristics of anti-β1 AABs can also be of interest with regard to whether their removal would lead to clinical benefit. 27 Extensive studies in the setting of autoimmunity have led to the identification of numerous mechanisms underlying the generation of AABs. One such mechanism has been termed 'molecular mimicry', which describes the process of antibody generation against microbial proteins, which are also structurally homologous to self-peptides. 29 A search of the literature did not reveal the presence of any structural homologues of ANS receptors; still, structural similarities between microbial peptides and ANS receptors cannot be ruled out. A different mechanism involves the generation of AABs against strictly intracellular antigens, as often occurs in autoimmune diseases (e.g. anti-nuclear antibodies and anti-cytoplasmic antibodies). 30 The latter is less likely in this case, considering that ANS receptors are transported to the cell membrane. An additional explanation could be that B-lymphocytes expressing B-cell receptors that recognize self-antigens with high affinity are efficiently depleted or functionally silenced, while B-lymphocytes with medium-or low-affinity receptors may escape tolerance mechanisms and give rise to antibody-producing plasma cells. 31 Collectively, circulating markers of B-lymphocyte function did not differ significantly between seropositive and seronegative patients in our PCA analyses, suggesting that ANS-AAB generation was not an acute process.
Autoreactive B-lymphocytes and plasma cells also occur in the general population, and seropositivity for AABs associated with thyroid disorders or autoimmune diseases does not per se lead to clinical manifestations. 14,32,33 Yet, both within the field of cardiovascular disease and in other disease states, circulating ANS-AABs have been implicated in the pathophysiology of various diseases. The reasons behind the differential effects of circulating AABs in the general population and in specific disease processes remain unknown. Natural autoreactive immunoglobulin-M (IgM) antibodies may protect from autoimmune disease, 34 while the subtype of circulating IgG antibodies as well as the glycosylation and sialylation status of their Fc-segments regulates higher affinity binding to inhibitory Fcγ receptors (FcγRIIA, FcγRIIB). 31 Interestingly, animal studies have shown that the generation of IgG antibodies with pro-inflammatory profiles requires a T-lymphocyte-dependent immune response, 35 which is in agreement with previous findings by our group underlying the important prognostic role of T-lymphocyte activation in HF. 10 Furthermore, anti-β1 AABs from patients with HF lead to cardiac fibrosis, cardiomyocyte apoptosis, LV dilatation, and increases in LV mass in wild-type mice, which was not observed in mice lacking T-lymphocytes. 36 The same study showed that blockade of IL-6 production using small interfering RNA in wild-type mice ameliorated cardiomyocyte apoptosis and that anti-β1 AABs upregulated IL-6 production in T-lymphocytes from patients with HF. 36 These findings illustrate the interplay between T-lymphocytes and ANS-AABs and are further supported by the prognostic value of IL-6 in HF. 37

Limitations
The post hoc character of this study and the relatively low prevalence of AAB seropositivity suggest that some analyses may have been underpowered, particularly in the case of anti-β3 AABs. Additional characteristics for the healthy control population were not available and could thus not be accounted for in statistical analyses. Additionally, the control population was significantly different from the examined patients (younger, more women), and our findings could suggest potential age dependency of AAB formation. Extensive antibody repertoire mapping was not performed, which might have led to the identification of AABs with greater prognostic implications. In addition, our study did not include any patients with Chagas cardiomyopathy or PPCM and lacked the necessary data to determine the presence of non-hereditary DCM. The BIOSTAT-CHF cohort included patients that were suboptimally treated for HF and might thus not be representative of the general HF population. Furthermore, it could be hypothesized that the uptitration of HF treatments after initial blood sampling could have masked or attenuated the effects of AABs on outcomes. Lastly, we did not evaluate the functional characteristics of measured AABs (e.g. inhibitory and stimulatory), and the specific isotype/class, subclass, and allotype of AABs were not determined.

Conclusion
In a large and heterogeneous population of patients with HF, seropositivity for ANS-AABs was not strongly associated with adverse outcomes and was mostly related to existing comorbidities and medication use, while only seropositivity for anti-β1 AABs was independently associated with rehospitalization for HF. Further studies are necessary to better elucidate the potential prognostic significance and clinical associations of ANS-AABs in patients with non-hereditary DCM, Chagas cardiomyopathy, or PPCM.