Longitudinal plasma proteomics in romiplostim‐treated patients with immune thrombocytopenia

To the Editor, Immune thrombocytopenia (ITP) is a heterogeneous autoimmune bleeding disorder with a complex and incompletely understood pathophysiology.1 Thrombopoietin receptor agonists (TPO-RAs) are a common second-line treatment for patients with ITP who fail or relapse after corticosteroid therapy.1, 2 TPO-RAs mimic the effects of endogenous TPO by binding to the TPO receptor on megakaryocytes and thus stimulating platelet production.1, 2 However, an immunomodulatory effect might also take place upon TPO-RA administration as a sustained remission has been observed in 10%–50% of patients after discontinuation.1, 2 The underlying mechanisms leading to a sustained remission after TPO-RAs are unclear and demand further biological studies.1, 2 Proteomics enables large-scale plasma protein analysis and has demonstrated potential for elucidating molecular mechanisms in ITP.3-6 So far, plasma proteomics in ITP has been used for the identification of response predictors to corticosteroids in children3 or to splenectomy in adults4 and to identify biomarkers for diagnosis and prognosis.5, 6 Here, we performed mass spectrometry (MS)-based plasma proteomics on longitudinally collected samples from the Stop TPO-RA in ITP Patients (STIP) study7 in which we investigated the sustained remission rate and clinical effects of romiplostim tapering, a type of TPO-RA, in patients with ITP. We evaluated plasma proteomic changes upon romiplostim treatment and tapering to better understand the mechanisms contributing to sustained remission in this cohort. Blood samples from 40 patients with ITP ≥3 months were collected at six time points: before romiplostim (T-1), during treatment at 1–3 months (T0), 6 months (T1) and 12 months (T2), and after tapering at 3 months (T3) and 12 months (T4) (Figure 1A). Methods and cohort characteristics of the STIP study have been described previously7 (Table S1). Romiplostim dose was cut in half every 2 weeks within 6 weeks. Routine laboratory screening was performed by the local laboratory (Table S2), and whole blood samples were transported within 3 h (T-1 and T4) or within 24 h (T0–T3) after blood draw (Figure 1A). A monoclonal antibody-specific immobilization of platelet antigen (MAIPA) assay was performed to assess anti-platelet antibody levels. Plasma was isolated and frozen (Methods S1). For this study, we compared patients with a sustained complete response off-treatment (SCROT, n = 4) to those with a non-sustained complete response (NSCR, n = 18) at 1 year after romiplostim tapering (Table S3). A SCROT was defined as platelet counts >100 × 109/L, no bleeding symptoms and no need for treatment. If any criteria were unmet, the outcome was classified as NSCR. Samples of healthy adults (n = 10) were used for comparisons to patients; these were transported within 24 h following the same protocol as samples from T0 to T3 (Table S1). For a detailed description of the proteomic methods, see Methods S1. Briefly, plasma proteins were denatured, reduced, alkylated and digested into peptides using trypsin as described previously.8 Thereafter, liquid chromatography tandem MS analysis was performed. Data were processed with DIA-NN (v1.8.1) and DIA-NN matrices were loaded into R (v4.1.1). Analytical performance was determined on coefficients of variations (CVs) enabled by repeated injections of the same quality control sample (QC). Analysis of longitudinal plasma samples with MS-based proteomics identified a median of 547 proteins per sample (Figure 1B). In total, 93% of proteins quantified in the QC samples had a CV of up to 30%, which highlights the robustness throughout our workflow (Figure S1). To further evaluate the relevance of our proteomic results, we checked for agreements between hospital laboratory measurements and proteins quantified with our MS-based plasma profiling approach. A strong correlation for C-reactive protein between protein measurements performed with conventional laboratory techniques and our MS data were found (ρ = 0.86) (Figure 1C). Additionally, strong correlations were found between alanine transaminase and fructose-bisphosphate aldolase B (ρ = 0.74) (Figure 1D), as well as between platelet counts and platelet glycoprotein-1b beta-chain (ρ = 0.78) (Figure 1E). Next, we explored global differences between plasma proteomes from all romiplostim-treated patients with ITP (n = 40) and healthy controls (n = 10) using principal component analysis (Figure S2A). This analysis pointed towards samples clustering in two main groups based on differences in transport time (Tm1 and T4 vs. T0–T3; Figure 1A). The separation between these groups was driven by granular cell signatures, including neutrophil- and platelet-associated proteins such as protein S100-A8 (S100A8) and pro-platelet basic protein (PPBP) (Figure S2B). For these proteins, the levels increased in samples with longer transport time (<24 vs. <3 h) (Figure S2C). For this reason, subsequent analysis based on longitudinal comparisons was only carried out between samples with the same transport time. Next, we investigated plasma proteomic levels characteristic of each treatment outcome group. These results showed significantly higher levels with complete data for two immunoglobulin variable regions (IGHV2-70D and IGLV2-8) at T-1 in the SCROT group compared to the NSCR group (Figure 2A,B). As variable regions have high interindividual variation, we evaluated the dynamics of main immunoglobulin classes as well as variable regions at the individual level. Despite the general trend of higher levels of immunoglobulin heavy gamma 1 in patients with an SCROT (Figure S3), these differences appeared to be mainly driven by inter-individual variation of the protein levels (Figure S4). Subsequently, we examined whether the immunoglobulin variable regions in our proteomic workflow correlated with the presence of anti-platelet antibodies determined by the MAIPA assay. No correlation between MAIPA results and plasma proteomic signatures was observed (Figure S5). Lastly, longitudinal changes within treatment outcome groups were explored through pairwise comparisons, taking into account differences in transport time (Figure 2D,E; Figure S6). These results showed that, with >50% data completeness (Methods S1), patients with SCROT had significantly higher levels of PPBP and thrombospondin 1 (THBS1), and lower levels of fibulin 5 (FBLN5) at T4 compared to T-1 (Figure 2C). Decreased FBLN5 was also observed compared to NSCR at T4 (Figure S7). For patients with NSCR, levels of proteins such as cofilin 1 (CFL1), osteonectin (SPARC) and platelet factor 4 (PF4) were significantly higher at T4 compared to T-1 (Figure 2D). This is, to the best of our knowledge, the first study to characterize plasma proteomic signatures in a cohort of patients with ITP who were longitudinally followed after romiplostim tapering. Previous proteomic ITP research has mainly focused on identifying diagnostic or prognostic markers cross-sectionally, highlighting proteins such as THBS1 and CFL1 as diagnostic and splenectomy response markers respectively.3-6 Differences in study designs and cohorts, along with the inherent heterogeneity among ITP patients themselves, complicate generalization and interpretation of these studies. Longitudinal studies provide a way to address these issues, but standardization of pre-analytical handling is essential to enable longitudinal comparisons.9 This study underscores the importance of pre-analytical handling, as we showed that transport time affects the plasma levels of neutrophil- and platelet-associated proteins. Therefore, we restricted our comparisons to time points with the same pre-analytical time. However, we cannot exclude further potential effects of pre-analytical variability. In our cross-sectional analysis, we found increased levels of immunoglobulin variable regions (IGHV2-70D and IGLV2-8) before or at the start of romiplostim (T-1) in patients who achieved a sustained remission after romiplostim tapering. This seems attributable to inter-individual variability and requires further validation. Moreover, no association between these immunoglobulin variable regions and the presence of anti-platelet antibodies was found. Nonetheless, given the established role of antibodies in ITP10 and observed higher levels in all patients with SCROT, future studies should investigate levels and specificity of immunoglobulins in ITP patients stratified by treatment response. In our longitudinal comparison, most of the proteins that increased after romiplostim are platelet-associated proteins like PPBP, SPARC, THBS1 and PF4. THBS1 has been previously proposed to be a diagnostic marker as decreased levels were observed in paediatric ITP patients.5 A proteomics analysis of bone marrow reported a downregulation of PPBP and SPARC in newly diagnosed ITP patients.11 Interestingly, both patient outcome groups, with and without complete remission, showed a trend of increasing abundance levels of platelet-associated proteins during romiplostim treatment (T0 until T2), but normalization to abundance levels of healthy controls occurred almost exclusively in patients with complete remission. These results likely reflect the increment in platelet mass and any functional involvement of these platelet-associated proteins remains to be elucidated. Literature is lacking, but THBS1 may modulate T cells and macrophages by activating secreted latent Transforming Growth Factor β (TGF-β).12 Besides platelet activation, PF4 has been shown to affect immune responses in a context-dependent way.13, 14 Additionally, reduced FBLN5 after successful tapering might indicate a restored coagulant state, given its link to platelet aggregation, von Willebrand factor activity and vascular wall remodelling.15 However, further validation in larger cohorts and functional studies is needed given our small cohort, with limited SCROT cases, and the explorative nature of our study. In this study, different pre-analytical transport times restricted the opportunity to analyse this longitudinal cohort at its fullest. Despite this, we observed changes in several platelet-associated proteins and FBLN5 during romiplostim treatment and tapering within the treatment outcome groups. In rare diseases like ITP, we highly recommend longitudinal study designs with standardized pre-analytical handling to enable robust biomarker discovery with unbiased plasma proteomics. V.S.N. and E.R.S. equally contributed to this work and are co-first authors. E.R.S. and D.M.S. performed the sample preparation, MS analysis and data processing. V.S.N., E.R.S. and D.M.S. performed the proteomics data analysis, wrote the manuscript and designed the figures. V.S.N. coordinated the STIP study and collected the material and data. M.R.S. designed and supervised the STIP study. M.d.H. supervised the plasma processing and collection at Sanquin. M.v.d.B. supervised the proteomics experimental workflow. R.K., M.d.H., M.R.S., M.v.d.B. and D.M.S. supervised the research, took part in work discussions and critically reviewed the manuscript. All authors reviewed and edited the original draft. Sanquin grant (number: PPOC22-03/ L2699, to Rick Kapur) was used to fund this work. All authors have declared that there are no conflicts of interest associated with this manuscript. The STIP study trial was registered at https://onderzoekmetmensen.nl/en/trial/22827 (the Dutch Trial Register) as #NL6605. The STIP study was approved by the Medical Ethics Committee Leiden-Den Haag-Delft (approval number: 58678.098.16) and conducted in accordance with the Declaration of Helsinki. Written informed consent was collected prior to inclusion in the study. For original data, please contact one of the corresponding authors. Figure S1. Quality control and assessment of experimental workflow. Figure S2. Delivery time-dependent changes in plasma protein profiles. Figure S3. Plasma protein levels of immunoglobulin heavy constant gamma subtypes. Figure S4. Plasma protein levels of significantly different immunoglobulin light regions between treatment outcome groups. Figure S5. Correlation of protein levels to monoclonal antibody-specific immobilization of platelet antigen (MAIPA) results. Figure S6. Longitudinal comparisons within treatment response groups. Table S1. Baseline characteristics of STIP study cohort and healthy controls. Table S2. Laboratory parameters measured by diagnostic testing. Table S3. Characteristics of patients with a sustained complete remission off-treatment (SCROT) or non-sustained complete response (NSCR). Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.