I. Hemodialysis patient plasma is proteolytically dysregulated and pro-fibrotic
HD patient plasma indicated broadly elevated proteolytic activity and regulation, a pro-fibrotic GF profile, cardiac stress, and elevated systemic inflammatory response compared to healthy plasma (Fig. 1). Concentrations of four matrix metalloproteinases (MMPs: MMP2, MMP3, MMP9, MMP14) and two tissue inhibitors of MMPs (TIMPs: TIMP1, TIMP2) were significantly elevated in dialysis patients both before and after HD compared to control values (Fig. 1A). In contrast, plasma levels of growth factors were more nuanced (Fig. 1B), with fibroblast growth factor two (FGF2) levels significantly below and insulin like growth factor 1 (IGF1) concentrations significantly above control levels. Platelet derived growth factor (PDGF) D concentrations pre-HD trended higher than control levels and were significantly elevated in post-HD plasma, while neither pre- nor post-HD PDGFB levels significantly differed from controls. In agreement with prior studies30–33, inflammation and cardiovascular stress measured uniformly higher in dialysis patients, shown by N-Terminal Prohormone of Brain Natriuretic Peptide (NTpBNP), C-reactive protein (CRP), tumor necrosis factor receptor 1 (TNFR1), and interleukin 1 beta (IL1β) as shown in Fig. 1C. Broad elevation of MMPs and TIMPs support enhanced tissue remodeling as an adaptive response to HD. Further, the widely sustained elevation of plasma MMPs and TIMPs post-treatment suggests a pathogenesis that is insufficiently cleared and potentially exacerbated by HD. Given prior findings that FGF2 reduces structural damage in CKD34, while elevated IGF1 correlates to systemic sclerosis35, these results suggest a pro-fibrotic phenotype in HD patient plasma. Further, given the key role PDGFs play in the genesis of vascular lesions, elevated chronic PDGFD and minor PDGF increases post-HD may contribute to a homeostatic shift toward tissue stiffening in HD patients36. This link between proteolytic dysregulation and fibrosis in HD patients is further supported by the mechanistic induction of TIMPs and attenuation of MMPs by PDGFs37. No differences were observed between dialysis patient and control plasma for the TGFβ precursor latency associated peptide (LAP), pro-platelet basic protein (PPBP)/ Beta-Thromboglobulin (βTG), Platelet-Activating Factor (PAF), Von Willebrand Factor (vWF), platelet factor 4 (PF4), Glyceraldehyde 3-Phosphate Dehydrogenase (GAPDH), Thrombopoietin (TPO), thromboxane metabolite (TXM) 11TXB2, and Interferon-γ (IFNγ). For vasodilator Prostaglandin I2 (Prostacyclin; PGI2) no difference was observed between pre-HD and healthy control levels, though post-HD levels were significantly suppressed (Figure SF1).
II. Intra-dialysis changes reveal reduced MMP and heightened TIMP1 plasma levels
The impacts of single HD sessions on circulatory biomarker accumulation were assessed by relative changes in paired plasma samples from dialysis patients pre- and post-treatment. One-sample t tests indicated plasma protein concentrations likely to change iatrogenically on average. Post-HD plasma indicated lower proteolytic activity, higher proteolytic inhibition, higher clot sensitivity, and lower glycolytic enzyme levels compared to paired pre-HD samples. Significant intra-dialytic changes were found in 2 MMPs (MMP3, MMP9), 2 TIMPs (TIMP1, TIMP2), no GFs, 1 clotting factor (vWF), 1 cardiovascular indicator (GAPDH), and no inflammatory indicators (Fig. 2). All measured proteases and proteolytic regulators decreased in a majority of HD patients, reflecting typical protein removal effects by mechanisms like membrane adsorption. However, TIMP1 significantly increased on average. Considering TIMP2, MMP3, and MMP9 significantly decrease on average, this isolates TIMP1 as a molecule that is acutely secreted into plasma during dialysis. In particular, an increased ratio of TIMP1 to MMP9 post-dialysis may link HD to increased tissue fibrosis and systemic stiffening characteristic of cardiorenal syndrome. In contrast, relative changes among measured growth factors were not statistically significant. FGF2, PDGFB, IGF1, and Connective Tissue Growth Factor (CTGF) increased in a majority of patients during HD sessions, while epidermal growth factor (EGF), LAP, PDGFD, and PPBP decreased in a majority of sessions (Figure SF2). Among measured clotting factors, vWF levels significantly increased on average. Among cardiovascular and inflammatory indicators, GAPDH significantly decreased on average. In parallel, 11TXB2 and CRP increased in a majority of sessions, while Thrombopoietin (TPO), NTpBNP, PGI2, TNFR1, TNFα, IFNγ, and IL1β decreased in a majority of sessions, with no significant paired average changes (Figure SF2).
III. Plasma proteolytic imbalance worsens with dialysis vintage and faster blood flow rate
Pre-treatment levels of all proteins measured were compared simultaneously by multivariate linear regression (MLR) against patient age, HD Vintage (commonly defined as the number of months of regular HD treatment), average HD blood flow rate (BFR), sex, and weight to determine potential targets and patient predictors of cyclical dialytic stress (Table ST6). Plasma levels of PAF correlated uniquely to BFR against all patient predictors (p = 0.017) (Fig. 1D), elevations of which may contribute to the development of microvascular dysfunction38 and amplification of leukocyte-induced microvascular alterations through enhanced platelet recruitment39. Separately, MMP3 correlated uniquely to BFR (p = 0.025) and MMP14 correlated uniquely to HD Vintage (p = 0.038) against all patient predictors, indicating increasingly disturbed proteolytic balance.
Relative plasma changes intra-HD compared to treatment parameters indicated effects and predictors of protein removal and accumulation (Table ST6). Among simultaneously compared patient predictors, BFR correlated negatively against ΔPAF (p = 0.018) and ΔMMP2 (p = 0.058) (Fig. 2B), while ΔMMP3 was predicted by both BFR (p = 0.020) and weight (p = 0.046) (Figure SF3), reflecting acute effects of increasing prescribed HD speed. HD Vintage uniquely predicted positive ΔIFNγ (p = 0.043), Δ11TXB2 (p = 0.032), and ΔTIMP1 (p = 0.027), reflecting both acute changes during dialysis and longer-term chronic adaptations to repeated dialysis exposure (Fig. 2C). Positive correlations between BFR and pre-treatment PAF and MMPs suggest increasing BFR may aid short-term clearance while also contributing to long-term dysfunction. However, increasing IFNγ and decreasing 11TXB2 with vintage imply chronically enhanced inflammatory sensitivity and platelet dysfunction in response to prolonged HD treatment. The combination of chronically elevated plasma TIMP1 levels in HD patients (Fig. 1A), further acute elevation in response to dialysis (Fig. 2A), and positive correlation to dialysis vintage suggests both acute and chronic adaptations that enhance MMP inhibition and fibrotic ECM deposition40.
IV. Intra-dialysis platelet transcriptome changes reflect long-term drivers of dysfunction
Platelet lysate (PL) from pre- and post-HD blood samples were assessed for relative RNA changes to examine chronic effects on platelet function between treatment sessions. High variance accounted for the lack of statistical significance, yet PCR analysis revealed greater average RNA content in post-HD platelets for all secreted biomarkers relative to generic platelet Purinergic Receptor P2Y12 (P2RY12) (Fig. 3A). This is consistent with expected platelet turnover, increasing overall RNA reserves. However, disproportionate intra-HD changes between markers reveal selective degradation following translation41,42 and upregulation in megakaryocytes prior to fragmentation25, neglecting endogenous uptake mechanisms43. Significant increases were found in 2 growth factors (EGF, PPBP), 2 clotting factors (PF4, vWF), and 1 metabolic indicator (GAPDH). Among growth factor RNAs, increased EGF and PPBP may signify enhanced regenerative and fibrotic potential, while the majority decreases observed for FGF2, PDGFB, PDGFD, and TGFβ1 signify stimulated depletion without compensatory upregulation in new platelets. Among clotting factor RNAs, significant increases measured for PF4 and vWF indicate platelet hypersensitization, while the majority decreases observed for PAF RNA may presage a pathway to long-term dysfunction. While no significant intra-HD changes were found in MMPs or TIMPs, MMP1 decreased while all 3 measured TIMP transcripts increased in a majority of patients, indicating preferential TIMP upregulation. The multi-modal distribution of TIMP1 changes reveals patient subsets exhibiting either depleted or highly upregulated TIMP1 RNA. Because of patient differences in rates of platelet renewal, this could signify that HD stimulates both TIMP1 translation in existing platelets and upregulation in subsequent platelet production. The parallel significant increase in GAPDH represents the increased metabolic capacity of platelets upon renewal. When correlated simultaneously against patient predictors by multivariate linear regression, Vintage uniquely predicted ΔPAFAH2 RNA (p = 1.9E-4) (Table ST6). Increasing intra-dialytic PAF upregulation (R = 0.861, p = 9.2E-6) (Fig. 3B) supports altered platelet translative potential leading to clotting dysfunction with long-term HD. These results imply that beyond RNA renewal effects of chronically shortened platelet life cycles, megakaryocytes may respond to iatrogenic stress by shifting platelet protective and translative behaviors.
V. Platelet secretomes shift with dialysis vintage
To determine intra-dialytic effects on platelet activity, blood was re-sampled from 12 HD patients for platelet proteomic analysis (Fig. 3C). Among measured MMPs, TIMPs, and GFs for paired pre- and post-HD samples, one significant decrease was observed in connective tissue growth factor (CTGF), signifying a case in which granule-mediated secretion outpaces production. While relative platelet lysate protein changes generally showed high variability, MMP3 correlated uniquely to positive BFR (p = 0.046) (Table ST6), signifying net synthesis accompanying platelet renewal (Fig. 3D). TIMP2 correlated uniquely to negative Vintage (p = 0.012) (Table ST6), suggesting depletion after long-term HD, while PPBP correlated uniquely to positive Vintage (p = 0.021) (Table ST6), revealing enhanced production of fibrotic molecules (Fig. 3E). Collectively, these suggest that faster BFR increases platelet proteolytic capabilities, and that long-term HD patients develop pro-fibrotic platelets.
VI. Elevated activation and inflammatory state indicate platelet dysfunction
Plasma measurements revealed chronic accumulation of MMPs and TIMPs and acutely heightened TIMP1, while disproportionate changes in the platelet secretome revealed altered protein synthesis between HD sessions. However, further examination of platelet secretome correlations to plasma clotting, inflammatory, and cardiovascular regulators revealed fundamental behavioral adaptations to HD (Fig. 4). In aggregate, these findings reveal accelerated platelet turnover with enhanced translative capacity, augmented proteolytic enzyme translation and secretion, elevated platelet inflammatory and regenerative response, platelet dysfunction in patients with attenuated native platelet inhibition, and further evidence of worsening dysfunction with faster BFR and longer Vintage.
i. Chronic elevation of platelet activators accelerates platelet turnover and enhances overall translative capacity, while acute elevation preferentially augments MMP/TIMP translation and secretion
Platelet secretome correlations to plasma measurements of clotting factors connect platelet activation with both long- and short-term changes in transcriptomic potential and secretory behavior (Fig. 4A). Multiple positive correlations between RNA and pre-HD plasma clotting factor levels reveal selective growth factor, MMP1, and TIMP upregulation upon platelet renewal for patients with chronic elevation of PAF and TXM. In parallel, negative RNA correlations to acute plasma clotting factor changes show selective translation and degradation of MMP1 RNA with iatrogenic clotting. Negative platelet lysate IGF1 and positive MMP14 correlations to pre-HD plasma PAF reveal heighted platelet sensitization to granule secretion and membrane-MMP presentation. Finally, correlations between the change in platelet lysate (ΔPL) and plasma (Δplasma) clotting factors similarly show MMP and TIMP secretion and MMP14 presentation with acute clotting. When collectively applied as predictors of platelet activity, multivariate linear regression reveals that pre-HD clotting factor elevation of PAF and TXM indicate MMP1 upregulation (R2 = 0.91, p = 0.027), while pre-HD elevation and intra-HD clearance of PAF and TXM indicate TIMP2 upregulation (R2 = 0.98, p = 0.044) (Table ST7).
Inflammatory elevation accompanies amplified platelet inflammatory and regenerative response
Inflammatory cytokines influence the platelet secretome in dialysis patients (Fig. 4B). Platelet EGF and IL1β RNA increased with plasma IL1β, while EGF RNA was translated and degraded with plasma ΔIFNγ, showing regenerative and inflammatory contributions with systemic inflammation. Conversely, positive PDGF-DD, TIMP2, and thromboxane (TX) RNA correlations to acute plasma ΔCRP suggest that acute inflammation suppresses translation and degradation of pro-fibrotic RNA (R2 = 0.87, p = 0.0007) (Table ST8). Whereas clotting factors have greater emphasis on platelet MMP/TIMP production and secretion, inflammatory stimulation leads to amplification and increased production of growth factors. Elevated plasma TNFα accompanied reduced ΔMMP24 and ΔTIMP3, suggesting reduced MMP/TIMP production. Degranulation with acute inflammation is observed by presentation of MMP14 and depletion of PPBP.
ii. Patients with higher native platelet inhibition show lower platelet dysfunction
Platelet correlations to cardiovascular markers validate known regulatory and dialytic effects (Fig. 5). Platelet dysfunction is observed by increasing PAF RNA with plasma NTpBNP. TIMP2 RNA upregulation and PL TIMP1 depletion with plasma ΔTPO shows selective TIMP renewal with platelet turnover. Positive RNA correlations to plasma PGI2 and ΔPGI2 validate platelet inhibition of RNA translation and degradation, though minimal improvement in adjusted R2 is observed by multiplexing RNA predictors against plasma PGI2 (Table ST9). Finally, platelet MMP1 RNA, MMP3 and CTGF production are attenuated by higher plasma PGI2. Thus, normal restorative platelet inhibition indicates lower platelet dysfunction.
iii. Faster BFR and longer HD Vintage exacerbate platelet dysfunction
Comparisons of platelet RNA and protein levels against clotting factors indicate broader effects of BFR and Vintage on long-term platelet translation and secretion. Faster BFR increasingly stimulates platelet dysfunction, shown by comparison of pre-HD plasma PAF to BFR (Fig. 1D) by multivariate linear regression. Similarly, stimulation of platelet RNA renewal and degranulation with plasma clotting factor elevation depict a mechanism of increasingly stimulated platelet dysfunction with BFR (Fig. 4A). Decreasing plasma ΔPAF and ΔTXM with both higher BFR and Vintage, respectively (Figs. 2B&C), and increasing ΔMMP1 RNA and PL ΔMMPs and ΔTIMPs with decreasing ΔPAF and ΔTXM (Fig. 4A) show preferential augmentation of platelet MMPs and TIMPs with increasing BFR and Vintage. Further, platelet-driven mechanisms causing tissue dysregulation appear to extend between HD sessions, shown by the significant relative increases in clotting factor RNA post-HD. Thus, in addition to directly enhancing platelet PAF RNA (Fig. 3B), faster BFR and longer Vintage may cause platelet behavioral adaptations that reduce regenerative contributions and disrupt proteolytic balance.
VII. Hemodialysis shifts platelet function from regenerative to pro-fibrotic
Platelet correlations to circulating clotting, inflammatory, and cardiovascular factors indicate worsening dysfunction with accelerated platelet turnover, subdued acute phase response, cardiac stress markers, and attenuated platelet inhibition. In parallel, secretome correlations to fibrotic and regenerative factors reveal distinct modes of platelet behavior and suggest a chronic shift toward a pro-fibrotic mileu. Increasingly pro-fibrotic platelets with proteolytic circulation and decreasingly pro-fibrotic platelets with regenerative circulation show two sides of the same coin. Regenerative feedback responses to fibrotic activity and vice versa show short-term mode-switching between HD sessions. Inverse proportionalities between alpha-granule proteins and lysosome proteases illustrate a long-term mechanism accounting for shifts in production.
i. Perturbed plasma proteolytic balance is associated with pro-fibrotic platelets with enhanced MMP/TIMP production and secretion
Increasing granule MMPs and TIMPs with increasing pro-fibrotic RNA identify platelet populations with enhanced fibrotic contributions (Fig. 6A). Lysosome MMPs and granule TIMPs increased with fibrotic GF, MMP, and TIMP RNA, while membrane-type MMPs increased with clotting factor RNA. Platelet TGFβ1, TIMP1, and PPBP upregulation predicted stored MMP1 (R2 = 0.88, p = 0.043) and MMP2 (R2 = 0.92, p = 0.0033) concentrations (Table ST10). Increasingly pro-fibrotic platelet transcriptomes are also identified by acute plasma MMP and TIMP increases (Fig. 6B). Fibrotic GF, TIMP, and TX transcripts increased with acute increases in plasma TIMPs (R2 = 0.45, p = 0.025) (Table ST11), while PAF RNA increased with acute increases in plasma MMPs (R2 = 0.60, p = 0.026) (Table ST12). The self-amplifying cycle of fibrotic transcript upregulation and protein synthesis with acute proteolytic imbalance is further supported by observations of enhanced platelet MMP/TIMP production and secretion (Fig. 6C), shown by increases in platelet lysate MMP/TIMP with chronic plasma TIMP elevation (R2 = 0.79, p = 0.043) and decreases in platelet lysate MMP/TIMP with acute plasma TIMP elevation (R2 = 0.79, p = 0.043) (Table ST13).
ii. Higher plasma levels of fibro-protective GFs correlate to lower fibrotic and higher regenerative contributions by platelets
Higher plasma levels of regenerative growth factors indicated platelets with decreasing fibrotic contributions (Fig. 7A). Plasma FGF2 correlated to reduced RNA for inflammatory IL1β and fibrotic TGFβ1.Platelet production and secretion of fibrotic CTGF and TIMP1 decreased with plasma FGF2 and IGF1, indicated by negative correlations to pre-HD levels and positive correlations to intra-HD changes. Using MLR, TGFβ1 RNA retention and reduced PL CTGF and TIMP1 predicted plasma FGF2 levels (R2 = 1.0, p = 0.0021) (Table ST14). Further, platelet TIMP1 was predicted by plasma IGF1 depression pre-HD, accumulation intra-HD, and FGF2 clearance (R2 = 1.0, p = 0.036) (Table ST14). However, unlike FGF2, plasma IGF1 correlated to increased platelet MMP1 RNA. Finally, clotting factor translation was reduced with ΔEGF, shown by TX RNA preservation.
iii. Acute elevation in fibrotic activity and PDGFs accompany a regenerative platelet response between treatment sessions
Acute enhancements in platelet secretome growth factors to acute plasma increases in fibrotic activity suggest short-term behavioral adaptations between HD sessions (Fig. 7B). Increased fibro-protective RNA (FGF2 and EGF) with fibrotic plasma (MMP9, ΔPPBP, and ΔTIMP3) shows a regenerative shift post-HD. This is supported by platelet IGF1 and PDGF retention with plasma ΔMMPs and ΔTIMPs. FGF2 upregulation with ΔPDGFD and retention with pre-HD PDGFD demonstrate a case of fibrotic enhancement of regenerative potential. In fact, multivariate linear regression revealed that plasma MMP9, PDGFD, and PPBP positively predict FGF2 upregulation (R2 = 0.76, p = 0.015) (Table ST15). Further feedback activity is observed by secretome correlations to PDGFs (Fig. 7C). As expected, PDGFB RNA decreases as plasma levels increase because of translation followed by degradation. Other platelet lysate correlations to plasma PDGFs show platelet depletion of PF4, retention of MMP3 and TIMP3, and acutely stimulated production of membrane-type MMPs.
iv. Secretome correlations support pro-fibrotic physiological adaptations in platelets
Platelet behavioral adaptations may be explained in part by selective differences in granule activity (Fig. 7D). Decreasing platelet lysate MMP9 and increasing platelet lysate PPBP with increasing EGF RNA show an inverse trade-off between producing lysosomal proteases and alpha granule proteins. Further, plasma PPBP was accompanied by reduced lysosomal secretion, observed by platelet lysate MMP2 retention, and increased alpha-granule production, observed by TIMP3 and PDGFB synthesis. Thus, increasingly regenerative platelet phenotypes may be characterized by increasing alpha-granule production and lysosome attenuation. Conversely, increasingly fibrotic platelets may present with increasing lysosome production and alpha granules production with pro-fibrotic disproportion.