Non-Pathogenic Antibodies in HIT: Clustering for Clarity

Introduction

Antibodies against Platelet factor (PF)4/heparin complexes are found in patients who have heparin-induced thrombocytopenia (HIT). Patients undergoing cardiac surgery are at high risk of developing anti-PF4/heparin antibodies due to the release of PF4 from activated platelets and exposure to intravenous heparin. Reports on postoperative incidence vary from 22-61%. In the absence of symptoms of HIT, their clinical relevance remains as uncertain as the determinants for their pathogenicity.

We planned to examine the incidence and time course of anti-PF4/heparin antibodies in patients undergoing cardiac surgery using a commercially available IgG specific ELISA kit in common clinical use. After identifying serum containing anti-PF4/heparin antibodies, we planned to assess their reactivity against a panel of 16 mutant(m)PF4 proteins in order to map their epitopes and determine if any differences exist with clinically pathogenic HIT samples.

Methods

After obtaining informed consent from patients undergoing cardiac bypass surgery, we performed ELISA using: (1) commercial Diagnostica Stago Asserachrom IgG kit (Cat.Nr 00624); (2) in-house ELISA with platelet derived native(n)PF4; (3) in-house ELISA with recombinant wild type(wt) and mPF4, expressed in E. Coli BL21(DE3) via pET-11a expression vector. ELISA performed as per manufacturer instructions and published methods.

All samples were screened with the commercial kit on day 7, or the nearest available serum between day 4-10. Positive samples were then examined longitudinally to chart the onset of antibody formation and their duration.

All samples were screened again using nPF4, and the results compared with the commercial kit. Concordant positive samples were further investigated by assessing their reactivity against 16 mPF4s that we selected based on available crystallography of human PF4 (D7A, Q9A, K14A, S17A, Q18A, R20A, R22A, P34A, H35A, T38A, K46A, N47A, R49A, D54A, L55A and Q56A). Unsupervised agglomerative hierarchical clustering was performed with PearsonÕs correlations and McQuittyÕs linkage analysis using RStudio version 0.98.1103. Multiscale bootstrap resampling was performed on the clustering to obtain p-values that were significant if <0.05.

Results

Commercial ELISA identified 30/127 (23%) patients positive for anti-PF4/heparin antibodies following cardiac surgery. Four patients had antibodies pre-operatively. Thus, 26 patients had de novo antibodies: 21/26 (80%) were first detectable on days 5, 6 or 7; 11/26 (42%) had antibodies detected on only one occasion; 12/15 (80%) demonstrated increasing optical density after first detection; 3 patients with de novo antibodies lost their antibodies by day 8, 9 and 32.

In-house ELISA using nPF4 revealed only 10/127 (7%) positive cases: 8/30 (26%) patients identified by the commercial kit were also positive for anti-nPF4/heparin antibodies; another 2/97 (2%) were positive for anti-nPF4/heparin antibodies but negative on the commercial kit.

Cluster analysis of 10 non-pathogenic anti-PF4/heparin antibodies identified several candidate epitopes. Despite considerable overlap, distinct patterns of clustering emerged to differentiate HIT serum from non-pathogenic samples. See Figures 1 and 2.

Conclusion

23% of cardiac surgery patients are positive on commercial HIT screening. Using an in-house preparation of nPF4, this number falls to only 7%. In the absence of clinical symptoms for HIT, these cases represent true non-pathogenic antibodies to PF4/heparin complexes. The overlapping clusters identified reflect the polyclonal nature of anti-PF4/heparin antibodies. In addition, we postulate that some patients identified by HIT screening may be cross-reacting to PF4/nucleic acid complexes. Our study was limited by the small number of true non-pathogenic cases we could identify for mapping purposes. Future, collaborative studies are justified to confirm our findings and explore further determinants for PF4/heparin antibody pathogenicity.

Figure 1

Legend to Figure 1: Unsupervised cluster dendrogram of mPF4s: hierarchical cluster analysis using nonpathogenic antibody samples. Boxes highlight significant clusters with a p value <0.05.

Figure 2

 

Legend to Figure 2: Relative change in optical density of serum with mPF4 compared to wtPF4. ELISA with horseradish peroxidase conjugated polyclonal rabbit anti-human IgG.