Pre-existing self-reactive IgA antibodies associated with primary graft dysfunction after lung transplantation

Abstract

Background:

Primary graft Dysfunction (PGD) results in significant mortality and morbidity after lung transplantation (LT). The objective of this study was to evaluate if pre-existing antibodies to self-antigens in sera of LT recipients are associated with PGD.

Methods:

The serum profiles of IgG and IgA autoantibodies were analyzed using a customized proteomic microarray bearing 124 autoantigens. Autoantibodies were analyzed using Mann-Whitney U test or Fisher exact test. The association of the autoantibodies with clinical phenotypes and survival was analyzed by Kaplan-Meier Survival Analysis. Receiver operating curve characteristics (ROC) were calculated to evaluate the predictive value of the autoantibodies for PGD.

Results:

51 patients were included in this study. Autoantigen microarray analysis on the pre-transplantation samples identified 17 IgA and 3 IgG autoantibodies which were significantly higher in recipients who developed PGD compared to those who did not (adjusted p < .05 and fold change > 1.5). 6 IgA Abs were significantly associated with survival. Taken as a panel, an elevation of 6 IgA Abs had significant predictive value for PGD. Area under the curve value for the panel was 0.9413 for PGD with ROC analysis. Notably, 6 of the 17 IgA autoantigen targets are belong to proteoglycan family of extracellular matrix proteins.

Conclusion:

Pre-existing IgG and IgA autoantibodies in LT patients correlate with PGD and with survival in a single center, small cohort of lung transplant recipients. Further validation is needed to confirm the findings in the study.

1. Introduction

Lung transplantation (LT) is a treatment option for patients with advanced lung disease [1]. However, the median survival for bilateral LT is only 7.3 years, significantly shorter than for other solid organs [1–3]. Allograft injury can present as primary graft dysfunction (PGD) [4], acute cellular rejection (ACR), antibody mediated rejection (AMR), lymphocytic bronchiolitis, or chronic lung allograft dysfunction (CLAD) [1].

PGD is associated with significant early and late post-transplant mortality and morbidity. Grade 3 PGD, which occurs in about 30% of recipients, predicts a longer hospital length of stay, longer duration of mechanical ventilation and higher 90 day mortality [5]. It is also associated with early onset of chronic rejection (bronchiolitis obliterans) and early donor specific antibodies [6].

PGD is thought to result from several variables. Ischemia of the donor lung followed by reperfusion triggers multiple mechanisms resulting in epithelial and endothelial cell injury, activation of the innate immune system, and release of inflammatory cytokines. However, the specific factors which predispose some recipients to the development of lung reperfusion injury and thus PGD are not well understood. Consequently, treatment for PGD is largely supportive.

Allo-immunity with recognition of mismatched donor (histocompatibility antigens) HLA by the recipient’s immune system results in graft injury [7]. In addition to allo-immune mechanisms, recent reports also demonstrate a strong correlation between antibodies to self- antigens and development of graft dysfunction. Bharat et al. [8] describe a higher risk of PGD and BOS in patients with pre-transplant IgG antibodies to self-antigens (Collagen I, Collagen V and K-alpha tubulin).

There is some evidence that functionally active autoantibodies with specificities for self-antigens play an important role in the progression of chronic lung diseases such as chronic obstructive pulmonary disease and idiopathic pulmonary fibrosis [36–38]. Recently Patel and colleagues describe an interesting observation where they demonstrate that ischemia/reperfusion and antibody/complement deposition were increased in recipient mice who received donor lungs from smoke-exposed mice. This study adds to the evidence that autoantibodies, which are increased in patients with chronic lung disease may contribute to post transplant allograft injury [39].

2. Objective

Our objectives for this study were to establish a tractable system to evaluate levels of pre-transplant IgG and IgA antibodies (Abs) to self-antigens (SAgs) that correlate with PGD, and to potentially identify a specific panel of these Abs that are associated with subsequent survival.

3. Methods

3.1. Study population

This is a retrospective single center study including adult recipients who received LT from January 1, 2010 to December 31, 2015. Institutional review board (IRB) approval was obtained IRB # STU 072016–041. Study follow up was completed November 1, 2018. Inclusion criteria, include, patients transplanted in our center, samples available and retrievable from the HLA lab for patients included. Exclusion criteria include inadequate data, lack of follow up, or re-transplantation. Various demographic, clinical, outcome variables for the cohort were recorded from the electronic medical record. We evaluated PGD as defined by standard criteria [4]. Worst grade PGD after 24 h was used for analysis. Samples collected as standard clinical care and stored in HLA lab for patients included in the study were retrieved. Baseline sample was defined as the sample collected on the day of transplant prior to implantation. Clinical variables such as, demographics, grade of PGD, donor specific antibodies (DSA), cumulative acute rejection score (CAR), defined as cumulative of the acute rejections since transplant, time of death, cause of death, were recorded.

3.2. Transplant management protocol

All patients were screened for pre-formed HLA antibodies every 3 months prior to transplantation. Donor lungs were accepted only if virtual crossmatch was reviewed and approved by the director of HLA laboratory. Induction was not routinely used. Its use was limited to patients requiring Cardiopulmonary bypass (CPB), those with bleeding complications, coagulopathy, or hemodynamic instability where initiation of tacrolimus was delayed in the setting of renal injury. All other recipients received tacrolimus infusion at 30 micrograms/h initiated intraoperatively. The maintenance immunosuppression regimen consisted of tacrolimus, azathioprine, and prednisone.

3.3. Microarray analysis

Serum samples from recipients collected as per routine protocol as defined above were procured from histocompatibility lab after IRB approval. Sera of all subjects were aliquoted and stored at −80 °C. Autoantigen microarrays were manufactured in the microarray core facility of University of Texas Southwestern Medical Center, Dallas, TX, USA. A selection of 124 autoantigens was made based on published literature, prior known autoantibodies in various immune related disease, cancer, allergic disease etc. [34].

124 antigens were ordered from different vendors (see attached list of antigens and vendors in supplementary table 3) and autoantigen array chips are manufactured in Microarray core of UT Southwestern using an Nanoplotter Microarray priter. The antigens were selected based on review of the literature and have been tested for their specificity with corresponding antibodies on our autoantigen array system. Most of the antigens are recombinant proteins expressed in E coli, insect or mammalian cell expression systems. Antigens used in microarray are validated with ELISA or Western blot with good correlation in multiple prior published studies [42–44].

4 proteins (human IgG, human IgA, anti-human IgG and anti-human IgA) were also imprinted on the arrays as positive controls. Human serum samples were first treated with DNAse I to remove free-DNA and then applied onto antoantigen arrays with 1:50 dilution. The autoantibodies binding to the antigens on the array was detected with cy3-labeled anti-human IgG and cy5-labeled anti-human IgA, and the array slides were scanned with Genepix 4400A scanner with laser wavelengths 532 nm for cy3 and 635 nm for cy5 to generate Tiff images. Genepix Pro 7.0 software is used to analyze the image and generate the genepix report (GPR) files (Molecular Devices, Sunnyvale, California, USA). The net fluorescent intensity (NFI) of each antigen was generated by subtracting the local background and negative control (Phosphate buffered saline or simplified as PBS) signal. The signal-to-noise ratio (SNR = (Foreground Median-Background Median)/standard deviation (Background) was also generated for each antigen. SNR is used as a quantitative measure of the ability to resolve true signal from background noise. A higher SNR indicates higher signal over background noise. To avoid outliers in either NFI or SNR, autoantibody score (Ab-score) which is defined by log 2((NFI*SNR) + 1) were used for all downstream analysis.

3.4. Statistical analysis

All group comparisons were performed with a non-parametric Mann-Whitney U test (or welch t-test) and a Fisher exact test for continuous and categorical data, respectively, using graphical user interface for R (The R Foundation for Statistical Computing). Criteria used to define differential expression in group comparisons is to meet both fold change > 1.5 and Benjamini-Hochberg adjusted p < .05. IgG and IgA autoAbs were evaluated for their predictive value for PGD using receiver operating characteristic (ROC) curve characteristic analysis. This analysis also determined optimal cut-off values for dichotomizing the values of autoAb reactivity (in Ab-score). Survival were estimated using Kaplan-Meier curves and survival curves between groups were compared by 2-sided log-rank test. All p values were 2 sided, and p < .05 was considered statistically significant. Significant IgA autoantibodies were then used to evaluate whether selected autoantibodies will correctly classify LT patients with PGD (PGD+) from those without PGD (PGD−) by ROC analysis. Support vector machines (SVM) were employed in the study [35]. We began by choosing radial kernel and tune the optimal model parameters (i.e., cost and gamma) to achieve the best diagonal performance on hold-on-one-out cross validation test. The SVM is trained using data from all but one of the samples. The sample not used in training is then assigned a class by the SVM. A single SVM experiment consists of a series of hold-one-out experiments, each sample being held out and tested exactly once. The e1071 R package is used for implement the SVM analysis.

4. Results

4.1. Baseline characteristics

Fifty-one LT patients met predefined criteria were selected for this study. Among them, 21 had PGD and 30 had no PGD. The mean age for LT was 54.86 years with 30 men (58.8%) and 21 women (41.2%). There was no difference in mean age in the groups with and without PGD. Majority in the group had PGD grade 2 or higher (76%). The most common indication for lung transplantation was pulmonary fibrosis other (non- IPF) (Table 1). Five recipients in the pulmonary fibrosis non IPF group had connective disease related interstitial lung disease. (Scleroderma: 2, Rheumatoid Arthritis: 2, Polymyositis: 1). Recipients in PGD group had higher CAR score compared with PGD negative group, but did not meet statistical significance (p = .20). PGD group had trend to worse survival but was not statistically significant, (p = .27) (Supplementary Fig. 1). No difference was observed in the presence of DSA between the two groups (p = .74).

Table 1.

Characteristics of 51 patients who did or did not develop PGD in baseline.

Characteristic	ALL (n = 51)	No PGD (n = 30)	PGD 1–3 (n = 21)	p value
Age, year (mean ± SD)	54.86 ± 15.57	56 ± 16.75	53.23 ± 13.95	0.64a
Gender				1.00b
Male	30 (58.8%)	18 (60%)	12 (57.1%)
Female	21 (41.2%)	12 (40%)	9 (42.9%)
Pathology				0.09b
COPD	9	5	4
IPF	11	8	3
Pulmonary Fibrosis non IPF!	18	11	7
Cystic fibrosis	7	5	2
PAH	5	0	5
Sarcoidosis	1	1	0
Race				0.59b
Caucasian	35	19	16
Black	6	5	1
Hispanic	9	5	4
Asian	1	1	0
DSA
Present, n (%)	40 (78.4%)	24 (80%)	16 (76.2%)	0.74b
Not Present, n (%)	11 (21.6%)	6 (20%)	5 (23.8%)
Acute rejection
CAR score (mean ± SD)	2.98 ± 2.65	2.57 ± 2.40	3.57 ± 2.94	0.20c
Standardized CAR (mean ± SD)	0.46 ± 0.35	0.41 ± 0.36	0.52 ± 0.32	0.28c
# total infection (mean ± SD)	1.71 ± 1.69	1.8 ± 1.86	1.55 ± 1.47	0.18c

4.2. Pre-existing IgA and IgG Autoantibodies associated with PGD

Samples collected from 51 patients at baseline (before transplantation) were analyzed by autoantigen microarray for the levels of IgG and IgA antibodies against 124 potential autoantigens. To assess the differences between PGD group and PGD negative group, we performed group comparisons using the 124 IgG and IgA autoantibodies, respectively. The criteria to declare differentially recognized autoantigens is to meet both fold change > 1.5 and Benjamini-Hochberg adjusted p < .05. Volcano plot as in Fig. 1, identified 3 IgG autoantibodies at baseline that were significantly higher in the PGD group compared to PGD negative group. These are IgG antibodies to Periplakin, Acetylcholine receptor (AchR3) and Angiotensin II receptor type-1(Fig. 1). Seventeen IgA autoantibodies were observed to be elevated significantly in PGD group at baseline compared with PGD negative group (Fig. 1).

Fig. 1.

Self-reactive IgG and IgA autoantibodies differentially expressed between PGD+ and PGD− at baseline level. [A] Volcano plot illustrating differences in IgG autoantibody expression between PGD− and PGD+. [B] Volcano plot illustrating differences in IgA autoantibody expression between PGD− and PGD+.

These 17 elevated IgA antibodies were directed on antigens such as nuclear proteins, extracellular matrix proteins and cytosolic proteins. Interestingly, six (Aggrecan, Heparan Sulfate, Laminin, Proteoglycan, rhHSPG2, SDC1) among the seventeen antibodies were associated with proteoglycan antigens (Fig. 5).

Fig. 5.

Among the 17 significant IgA autoAbs, 6 of them (Aggrecan, Heparan sulfate, Laminin, Proteoglycan, rhHSPG2, and SDC1) were associated with proteoglycans. The levels of the 6 autoAbs were highly correlated, indicating IgA autoAbs against these proteins may play a role in PGD of LT. Upper triangle represent Pearson correlation coefficients, violin plots in the diagonal showed differences between PGD− and PGD+; lower triangle are scatter plots.

4.3. IgA autoantibodies and survival of LT patients

ROC analysis was used to determine the optimal cutoff values for dichotomizing autoantibody reactivity (antibody score, Supplemental Table 2). These cutoff values were then used to perform Kaplan-Meier analyses (KMA) to correlate Ab score with survival. 6 IgA Abs to SAgs were significantly associated with worse survival when present at higher than cutoff antibody scores (Fig. 3). In addition, a combined analysis of all six IgA Abs was repeated with KMA. We found that the group with a low score (0–3 Ab above cutoff) (n = 24) had significantly better survival (p = .006) compared to the high score (4–6 Ab above cutoff) group (Fig. 4A). In contrast, survival was not significantly different for the 3 IgG Abs to SAgs or the remaining 11 IgA Abs to SAgs (supplemental Figs. 4, 5). Another repetitive analysis was done using univariate cox proportional analysis for survival with similar results (Fig. 2).

Fig. 3.

Kaplan–Meier survival analysis of IgA autoantibodies between PGD+ and PGD−. 6 IgA autoantibodies significantly associated with survival were showed only, the other 11 IgA and 3 IgG autoantibodies were showed in the supplementary Figures. The optimal cut-off values for dichotomizing the values of autoantibody reactivity was determined by ROC (receiver operating characteristic) analysis. [A] Survival analysis of Filaggrin (HR: 1.32; CI: 1.11–1.56). [B] Survival analysis of Factor P (HR: 1.47; CI: 1.12–1.92. [C] Survival analysis of CRP antigen (HR: 1.19; CI: 1.02–1.41). [D] Survival analysis of Laminin (HR: 1.28; CI: 1.06–1.56). [E] Survival analysis of RSV antigen (HR: 1.08; CI: 1.03–1.33) [F] Survival analysis of Heparan sulfate (HR: 1.25; CI: 1.06–1.45). HR: hazard ratio; CI: confidence interval.

Fig. 4.

Combined analysis of 6 selected IgA autoantibodies. If Antibody score of IgA autoantibody is greater than cut-off value, it will be scored as 1, otherwise 0. The maximal score that could be attained was 6.0. [A] Kaplan–Meier survival among patients with low score (0–3) or high score (4–6).[B] ROC analysis by support vector machine model. ROC = receiver operating characteristic. AUC = area under the curve, CI = confidence interval.

Fig. 2.

Hazard ratios and 95% CI from univariate Cox proportional regression analysis for survival with 17 IgA antibodies. Red box represents statistically very significant (p < .01), orange box represents statistically significant (p < .05) and blue box indicate not statistically significant (p > .05). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4.4. IgA autoantibodies as biomarkers for prediction of PGD

Six IgA Abs to SAgs that correlated with survival were used to evaluate whether selected autoantibodies will correctly classify PGD positive from PGD negative groups. ROC analysis performed using the panel of 6 IgA Abs, correlated strongly with PGD with area under the curve (AUC) 0.94 and confidence interval 0.87–0.96 (Fig. 4B). Taken as a panel, an elevation in these 6 IgA autoantibodies had very favorable operating characteristics with significant AUC suggesting that measuring these autoantibodies may constitute a clinically useful predictive panel for the occurrence of PGD.

5. Discussion

5.1. End stage chronic lung disease and autoimmunity

In chronic lung diseases, recurrent damage to cells and tissues exposes the adaptive immune system to numerous self-antigens. A variety of antibodies to such antigens have been detected in individuals with these diseases, consistent with immune conditioning and the development of autoimmunity over time. Mechanisms proposed for this phenomenon include release of sequestered self-antigens due to numerous injury-repair cycles in patients with end stage lung disease. Self-antigens are normally sequestered, preventing activation of self-reactive lymphocytes. Injury mechanisms that reveal sequestered self-antigen to host immune system combined with loss of regulatory T cells may lead to further activation of self-reactive lymphocytes and development of tissue-restricted autoimmunity [9]. It is still unclear if autoimmunity accelerates underlying chronic lung disease and very little is known about its influence on allograft injury patterns after LT.

In Cystic Fibrosis, for example, autoantibodies to PAD4, BPI, CCP, and pancreatic beta cells increase and are thought to worsen disease severity [21–25]. Similarly, autoantibodies are described in several prior studies in emphysema [26–30] and IPF [31–33], although their pathogenic significance is not clear. Because CF, emphysema, and IPF are some of the most common lung diseases that lead to LT, recent studies have suggested that such preexisting autoantibodies may also influence the immune response to lung allografts following transplantation [10,11]. The identification of specific autoantibodies to self antigens can provide important clues that potentially link the complex immune dysregulation leading to end stage lung disease with allograft injury patterns after lung transplantation.

5.2. Implications for panel of autoantibodies in the study

To expand our understanding of the role of such autoantibodies to lung transplant outcomes, we performed a comprehensive Ab microarray analysis of serum samples of 51 recipients who received LT at our center. Three IgG autoantibodies were identified that had significant > 2-fold elevations associated with the subsequent development of PGD, and these were against Periplakin, Muscarinic acetylcholine receptor type 3, and Angiotensin II receptor type 1. Of note, AGTR2 receptor activating antibodies have been found in kidney transplant recipients who had refractory vascular rejection [13]. Similarly, another three-center study [14] among lung transplant recipients revealed less freedom from antibody-mediated rejection for those recipients with strong/intermediate binding antibodies to AGTR2. Several reports in kidney transplant recipients [16], heart transplantation [17], lung transplantation [14], and liver transplantation [18] suggest detrimental effects resulting in graft loss with antibodies to AGTR2, and endothelin type A receptor. IgG autoantibodies to human periplakin have not been linked to transplant rejection, but anti-periplakin Abs have been reported in serum and bronchoalveolar lavage in patients with IPF (40%) [20]. Autoantibodies against AChR3 have not previously been reported and are novel.

Given their important role in mucosal host defense, we also screened for IgA antibodies, and found a striking global increase in IgA autoantibodies in patients who subsequently developed PGD, seen in the marked rightward shift of Ab levels in Fig. 1B. The reason for this generalized increase in serum IgA autoantibodies is unclear, but mucosal immunity is known to be stimulated in chronic lung diseases such as CF, emphysema, and interstitial lung diseases, and this increase in IgA autoantibody levels preceding PGD may reflect an overall hyperactivity in mucosal immunity. We focused on 17 IgA autoantibodies that were significantly elevated in pre-transplant patients that subsequently developed PGD. Serum IgA is known to interact with Fc-receptor expressed on several immune effector cells such as macrophages, natural killer cells and neutrophils. This leads to antibody dependent cytotoxicity (ADCC), degranulation of eosinophils, basophils, phagocytosis by macrophages and triggering burst activity of polymorph nuclear leukocytes results in cell death [40,41]. Similar mechanisms could possibly lead to injury patterns described in PGD.

Another interesting application described recently [40] are IgG and IgA cancer therapeutic antibodies resulting in antibody dependent cytotoxicity (ADCC) resulting in tumor cell lysis. When neutrophils are used as effector cells, IgA therapeutic antibodies induced significantly higher levels of ADCC compared to IgG variants of the same monoclonal antibody [40,41].

A second striking feature of these IgA autoantibodies is that 4 of the top 17 autoantigen targets are proteoglycan-family extracellular matrix proteins, and two others are biochemical preparations of heparan sulfate and proteoglycan. Proteoglycans are critical members of the extracellular matrix, particularly of the basement membrane, which is known to be attacked in most chronic lung diseases such as emphysema and IPF. SDC1 encodes syndecan 1, known to mediate cell adhesion and cell signaling, and HSPG2 encodes perlecan, another proteoglycan that physically interacts with laminin in the basement membrane. Aggrecan is also a proteoglycan, though it is found largely in articular cartilage. Proteoglycans have specific biophysical functions, including their enormous capacity to retain water molecules required to maintain structural integrity of tissues. Importantly, proteoglycans also serve as a large reservoir of growth factors and are therefore required for endothelial growth and vascular regeneration, a critical process following organ transplantation and graft maintenance.

While cause-effect relationships between elevations in autoantibodies and PGD are not clear, the strong correlation between IgA autoantibody elevations and survival suggests a pathogenic link. In addition, a thematic specificity towards proteoglycans found in the basement membrane, a structure known to be compromised in chronic lung diseases as well as allograft rejection, also raises a number of possible mechanisms whereby host systemic immunity originating from the lung mucosa is primed to focus an immediate inflammatory attack on basement membrane-associated structures of the allograft such as airways and blood vessels, with subsequent inflammation-related lung edema that is the hallmark of PGD. Such a mechanism would prompt exploration of options for immune modulation in addition to the traditional immunosuppression protocols currently in use. As an example, an intervention study by Bharat et al. [12] evaluated the response of extracorporeal photopheresis (ECP) on levels of alloantibodies against K alpha 1 T, collagen I, collagen V, and circulating levels of pro-inflammatory cytokines, and found that ECP is associated with reduction in levels autoantibodies and alloantibodies. Irrespective of the mechanistic implications, these results reinforce the importance of systematically measuring non-HLA antibodies.

5.3. Limitations in the study

Several limitations exist in our study. This is small cohort looking at autoantibodies for risk stratification. While we claim that 6 IgA antibodies correlate with PGD and survival, other confounding variables, donor, recipient and intraoperative factors could have contributed to the presence or absence of PGD. Larger sample size is needed to confirm the preliminary findings.

Furthermore, another limitation is that the autoantigen array chips are selected based on prior literature. Their post translation modification is not clearly known. Antibodies in our study were not validated with commercially available tests such as ELISA or Western blot analysis. However, antigens used in microarray are validated with ELISA or Western blot with good correlation in multiple prior published studies [42–44].

In addition, the IgG and IgA autoantibodies could be epiphenomena of the underlying injured lung rather than factors leading to PGD. Our small sample size limits the ability to stratify the autoantibodies based on underlying disease.

Further analysis to evaluate IgG and IgA autoantibodies after lung transplantation needs to be done. Tissue staining of the explant lung and one month post-transplant biopsies with this panel of autoantibodies or perhaps considering to study the effect of antibody panel in an animal model are the next steps. Further, prior to consideration of animal model, we would need to replicate the findings of our small cohort in a larger sample size with internal and external validation cohorts. This would include verification with ELISA or western blot. We still struggle with the aftermath of PGD leading to increased duration of mechanical ventilation, hospital stay and early chronic lung allograft dysfunction in some cases. Our study points to an important aspect of pre-existing autoimmunity in patients with end stage lung disease that can impact allograft injury.

5.4. Conclusion

In conclusion, we report that novel antibodies to self- antigens (IgG and IgA) measured prior to transplantation correlate with the incidence of PGD and also with patient survival. To our knowledge, this is also the first report of pre-transplant IgA autoantibodies holding predictive value in transplant outcomes. Further validation is needed for this panel, including verification with commercially available tests along with larger sample size, as well as mechanistic studies into the role of mucosal autoimmunity in early allograft dysfunction.