The highs and lows of cyclic thrombocytopenia

Abstract

Cyclic thrombocytopenia (CTP) is characterized by periodic platelet oscillation with substantial amplitude. Most CTP cases have a thrombocytopenic background and are often misdiagnosed as immune thrombocytopenia with erratically-effective treatment choices. CTP also occurs during hydroxyurea treatment in patients with myeloproliferative diseases. While the etiology of CTP remains uncertain, here we evaluate historical, theoretical and clinical findings to provide a framework for understanding CTP pathophysiology. CTP retains the intrinsic oscillatory factors defined by the homeostatic regulation of platelet count, presenting as reciprocal platelet/thrombopoietin oscillations and stable oscillation periodicity. Moreover, CTP patients possess pathogenic factors destabilizing the platelet homeostatic system thereby creating opportunities for external perturbations to initiate and sustain the exaggerated platelet oscillations. Beyond humoral and cell-mediated autoimmunity, we propose recently uncovered germline and somatic genetic variants, such as those of MPL, STAT3, or DNMT3A, as pathogenic factors in thrombocytopenia-related CTP. Likewise, the JAK2 V617F or BCR::ABL1 translocation that drives underlying myeloproliferative diseases may also play a pathogenic role in hydroxyurea-induced CTP, where hydroxyurea treatment can serve as both a trigger and a pathogenic factor of platelet oscillation. Elucidating the pathogenic landscape of CTP provides an opportunity for targeted therapeutic approaches in the future.

Summary

Like other cyclic hematopoietic disorders, the cyclic thrombocytopenia (CTP) phenotype represents a pathologically exaggerated state of homeostasis. Yet, unlike cyclic neutropenia associated with ELANE mutations and managed by G-CSF treatment, CTP is poorly understood and managed (Table 1). Here we combine the historical background with recent advances in CTP pathophysiology to outline future directions in diagnosis and treatment of this disorder.

Graphical abstract

Normal thrombopoiesis is oscillatory upon perturbation due to thrombopoietin-mediated platelet homeostasis regulation. Cyclic thrombocytopenia (CTP) exhibits exaggerated reciprocal platelet/thrombopoietin oscillations. CTP occurs on either thrombocytopenia or thrombocytosis baseline. While the etiology of CTP remains uncertain, we propose a 2-hit model where CTP patients possess pathogenic risk factors destabilizing the platelet homeostatic system thereby creating opportunities for external perturbations to initiate CTP. The risk factors of thrombocytopenia-based CTP may include humoral and cell-mediated autoimmunity, and genetic variants in MPL or STAT3. Thrombocytosis-based CTP mostly occurs during hydroxyurea treatment in patients with myeloproliferative diseases with underlying JAK2 V617F or BCR:ABL1 translocation.

Clinical presentation and diagnosis of CTP

Case Presentation:

A 55 year old man presented with thrombocytopenia and bleeding. He was treated with a steroid pulse and taper, associated with an increase in platelet count; however 1 month later, he was found to be thrombocytopenic with mucosal bleeding. Twice weekly blood testing over 80 days revealed periodic platelet count fluctuations (1–409 × 10⁹/L) independent of therapeutic interventions, with 180 degrees out of phase cycling of plasma thrombopoietin (TPO, 6–2745 pg/mL). Further study revealed heterozygosity for a novel germline loss-of-function MPL variant, G404R; persistent T and B cell clonality and 2 subclonal pathogenic somatic gain-of-function variants in STAT3 (D661Y and Y640F).¹ He has had persistent cycling for 8 years and monitors his platelet counts weekly with his own CBC instrument. He receives occasional platelet transfusions for non-self-limited mucosal bleeding at platelet nadirs.

CTP is a rare, often misdiagnosed, clinical condition characterized by dramatic periodic fluctuations in circulating platelets numbers.^2,3 The incidence of CTP remains unknown but is almost certainly underestimated with only approximately 70 CTP cases reported in the literature. The oscillatory periods usually last between 3 to 5 weeks, but can vary between 14 to over 60 days, while the platelet number cycles range from severely thrombocytopenic nadirs to normal or typically thrombocytotic peaks.^2–4 The rarity of CTP and its diverse cycling parameters make the diagnosis a challenge. The diagnosis can be even more challenging due to association of CTP with a broad spectrum of pathogenic conditions, ranging from autoimmunity to other hematological diseases.^5–8 For example, CTP patients often present with thrombocytopenia and are initially misdiagnosed as immune thrombocytopenia (ITP). In many instances, a spontaneous platelet count increase in CTP patients is initially misinterpreted as a clinical response to ITP therapy, delaying diagnosis.^2,3 ITP patients refractory to standard therapies and displaying unusual platelet count changes not attributable to dose adjustment are clues that CTP should be considered.² In these situations, frequent monitoring of blood count every 3–7 days for about 8 weeks to cover 2 potential cycles is recommended, and an intrinsic oscillatory pattern of platelet count will confirm the diagnosis of CTP (Table 1).²

Table 1.

Basic features of cyclic thrombocytopenia and cyclic neutropenia

	Cyclic thrombocytopenia (CTP)	Cyclic neutropenia (CyN)
Characterization	Periodic platelet count oscillation with thrombocytopenic nadir	Periodic neutrophil count oscillation with neutropenic nadir, and cyclical changes of other blood cell lineages may occur.
Symptom	Usually mild bleeding at platelet nadir	Recurrent fever, aphthae, or gingivitis, etc. during neutropenia phase, sometime have serious infections
Cycling period	Usually 3–5 weeks, stable for individual patient and likely determined by the pathogenic background	About 21 days for patients with ELANE mutations
Homeostasis regulation	Present with reciprocal TPO/platelet count oscillations	Present with reciprocal G-CSF/neutrophil count oscillations
Pathogenesis	Unknown. Acquired disease, can be associated with risk factors such as autoantibody, clonal T-cells, or genetic mutations that associate with either a thrombocytopenic or thrombocytotic baseline of platelet homeostasis.	Usually a hereditary disease, most commonly with mutations in ELANE gene.
Diagnosis	Based on clinical history of platelet count changes, monitoring every 3–7 days for about 8 weeks may reveal cyclic platelet patterns. Thrombocytopenia-based CTP are often initially misdiagnosed as ITP.	Based on clinical history especially with recurrent symptoms, CyN may be confirmed by monitoring neutrophil count 2–3 times per week for 6 weeks. Genetic test on CyN-related genes may help.
Treatment	Thrombocytopenia-based CTP is usually refractory to ITP treatment, and currently no universally effective treatment available. Discontinuation of hydroxyurea treatment can stop hydroxyurea-induced CTP. New therapies based on a better understanding of pathophysiology may be possible in the future.	Generally managed by G-CSF administration, plus treatment of infection when needed.

Additionally, CTP can be induced by environmental or medical interventions, among which hydroxyurea treatment-induced CTP in patients with polycythemia vera (PV) is the most prominent example.^7,8 Since the platelet cycling periods are about 28.6 days in these cases,⁸ monthly clinical follow up may mask the cycling pattern. Adjusting follow up schedules may reveal platelet count fluctuations which then raise the diagnostic consideration of hydroxyurea-induced CTP.

Except for hydroxyurea-induced CTP, where platelet oscillations usually stop after discontinuation of hydroxyurea, clinical successes of CTP management are rare, and CTP pathophysiology remains poorly understood (Table 1).

Overview of CTP pathophysiology

The homeostatic regulation of thrombopoiesis, and hematopoiesis in general, is “wired” with negative feedback circuits containing substantial time delay, which normally results in small, attenuated oscillations around a set point. Such oscillatory feedback circuits also appear to be involved in CTP pathophysiology (Figure 1). CTP patients possess pathogenic alterations (“errors”) resulting in increased “gain” in the system which magnify the amplitude of platelet cycling. Additionally, an external perturbation may be needed to initiate CTP, reflecting a 2-hit model – where an inherited or acquired risk factor (“errors”) intersects with a subsequent environmental perturbation (drugs, infections, etc) (Figure 2).

Figure 1. The intrinsic oscillatory factors of CTP due to TPO-mediated platelet homeostasis.

TPO-mediated feedback regulation of platelet production occurs in CTP and sustains reciprocal TPO/platelet oscillations at a stable cycling period (T). The orange arrows outline the sequential events of a cycle. Starting from a platelet nadir, where the MPL-mediated TPO clearance is around its lowest level, TPO level reaches its peak. Through TPO-mediated bone marrow platelet production, this TPO peak elicits a platelet production peak, which then constitutes a platelet count peak. Similarly, for the second half of the cycle, the platelet peak corresponding to TPO nadir, which delivers a platelet production nadir and platelet count nadir through the TPO-mediated platelet homeostasis mechanism. The platelet nadir then cause an instant TPO peak and a delayed platelet peak, and the cycles go on. Dark blue line: Platelet count; red line: TPO level; light blue line: Platelet gene expression level representing platelet production.

Figure 2. Diagram of CTP-associated risk factors and perturbations.

CTP can be induced by perturbations to platelet homeostasis on either thrombocytopenia or thrombocytosis baseline. For thrombocytopenia-based CTP, the baseline conditions involve compromised thrombopoiesis, accelerated platelet destruction, abnormal endocrine functions, or combinations of these. Humoral, cellular or genetic risk factors have been associated with thrombocytopenia-based CTP. Moreover, CTP phenotype occurs in patients with PV, CML or SCD, especially during hydroxyurea treatment. The green boxes in the three CTP platelet count diagrams outline the normal platelet count range.

Oscillatory nature of hematopoiesis and mathematical models

Homeostasis in hematopoiesis with its feedback regulation mechanisms and its entailed oscillatory nature were described in the 1960s. At that time, studies suggested that hematopoiesis originated from bone marrow hematopoietic stem cells (HSCs) and was generally under feedback regulation of lineage-specific humoral factors.⁹ For example, erythropoietin had been shown to be induced under anemic or anoxic stress to stimulate erythropoiesis, and thrombopoietin (TPO) was proposed to regulate thrombopoiesis, possibly through a feedback mechanism sensing platelet mass.^10–13 As it takes days or weeks to produce blood cells through lineage-specific differentiations in response to these and other humoral factors, analogous to electrical oscillators with nonlinear feedback cascades containing significant time delays, hematopoiesis is prone to “prolonged” (eg multi-day to multi-week) oscillation upon perturbations.¹⁴ Morley and colleagues outlined two features of these systems. First, “the period is equal or approximately equal to twice the time-delay.” Second, “the ultimate amplitude of oscillation depends on the loop gain.” In other words, whether a perturbation-triggered oscillation will be dampened, allowed to persist, or even exaggerated is determined by the particular status of the hematopoietic homeostasis system.¹⁴

As kinetic studies of hematopoiesis were pursued, mathematical models were developed to formulate hematopoiesis to explain oscillations under normal and diseased conditions. In 1968, Kirk and colleagues generated the first model, depicting erythropoiesis.^15–17 This model simulated the dynamic responses of erythropoiesis to perturbations, including transient overcompensations during recovery from radiation as well as sustained oscillations during the course of anti-erythrocyte iso-antibody administration.^15–18 In another early mathematical model, Morley and colleagues focused on granulopoiesis and simulated the normal neutrophil oscillations observed in healthy individuals.¹⁹ Moreover, it predicted that mild bone marrow deficiencies introduce high “loop gain” which could lead to substantial neutrophil oscillations, yet severe bone marrow deficiencies reduce the “loop gain” and abolish the oscillations.^14,19,20 A biological interpretation might be that systems with severe bone marrow deficiencies have lost their abilities to respond to homeostasis regulations and thus diminish their oscillatory tendencies. As a proof of concept, cyclical neutropenia was induced in dogs by cyclophosphamide administration at dose ranges causing mild bone marrow impairments.²¹ Interestingly, as revealed later, both cyclic neutropenia (CyN) and severe congenital neutropenia (SCN) can be caused by mutations of the same gene, ELANE, encoding neutrophil elastase, yet the mutational patterns associated with CyN versus SCN varied significantly.^22,23 For example, all the frame-shift mutations were associated with SCN, and patients with SCN had lower neutrophil counts compared to those with CyN.^22–24 Moreover, while patients with both CyN and SCN are clinically managed with granulocyte-colony stimulating factor (G-CSF) treatment, patients with CyN require a lower dose than those with SCN.^25,26 Furthermore, in several patients with SCN treated with high dose G-CSF, their “resolution” of severe neutropenia was to become CN.^27,28 These findings are generally in line with the mathematical modelling predictions. Later, models for CyN were further refined providing more insights to help understand the disease pathophysiology.^29,30

Likewise, the feedback regulation mechanism of thrombopoiesis also lies in the center of mathematical assimilations of platelet count oscillations. However, different from CyN which has a clear association with ELANE mutations, CTP etiology remains unclear and apparently heterogeneous. Thus thrombopoiesis models are still evolving to incorporate and simulate the diverse clinical and biological findings of CTP.^30–34 Currently, these models have suggested connections of platelet oscillations with multiple pathogenic features, such as autoimmunity and differentiation rate of thrombopoiesis.^32,34 Nevertheless, the dynamic feedback regulation concepts and mathematical models of hematopoiesis provide a useful conceptual framework for sorting out the diverse clinical findings to obtain a picture of overall CTP pathophysiology.

Applying these theoretical concepts of hematopoietic oscillation will help us to dissect CTP pathophysiology into 2 aspects: one is intrinsic oscillatory factors that are derived from thrombopoiesis regulation and continue functioning in CTP (Figure 1); the other is the pathogenic background of CTP leading to an enhanced susceptibility to exaggerated and persistent platelet cycling (Figure 2).

Intrinsic oscillatory factors involved in CTP pathophysiology

Before considering the intrinsic oscillatory factors, it is necessary to remind us of the time scales involved in thrombopoiesis and platelet homeostasis. In the case of sudden life-threatening emergency or stress, immediate platelet rescue does not emanate from bone marrow platelet production. Rather, approximately one third of the body’s platelets are “on call” in the spleen available on literally a moment’s notice.^35,36 This “back up plan” had to evolve in order to allow humans to survive emergency situations and it serves to contrast the time delay of steady state platelet homeostasis, which usually takes about 2 weeks to deliver platelet compensations through thrombopoiesis. This process is primarily regulated by TPO which inversely sensing platelet count changes to adjust the scale of thrombopoiesis accordingly.³⁷ Based on the mathematical concept,¹⁴ this platelet homeostasis mechanism through TPO-mediated feedback regulation of thrombopoieses has oscillatory nature.

Over 35 years after being proposed as the humoral regulator for platelet production, TPO was purified and the gene identified in 1994 shortly after the identification of TPO receptor, MPL.^38–41 TPO is predominantly produced by the liver, and to a lesser extent, by the kidneys, from where it is then released into circulation.^42,43 MPL is expressed on the surface of HSCs and all megakaryocyte lineage cells including platelets. TPO/MPL interactions form the fulcrum point of TPO and platelet count balance. On one hand, TPO binds MPL to activate downstream signaling pathways, such as JAK2 and STAT3 pathways, which stimulate megakaryopoiesis at various developmental stages.⁴⁴ On the other hand, the TPO/MPL complex is internalized upon binding on platelets and megakaryocytes surfaces, removing TPO from extracellular environment.^45,46 As a result, TPO levels are inversely correlated with platelet and megakaryocyte mass, forming a negative feedback cascade regulating thrombopoiesis to maintain platelet homeostasis. For example, if platelet count decreases, TPO level increases almost instantly, promoting new platelet formation through megakaryopoiesis. As mentioned above, since the compensation is always significantly lagging behind the perturbation, the platelet count cycles during the adjustment process before it reaches equilibrium. And the cycling period is equal to twice the time delay involved in the homeostasis circuit.¹⁴ This homeostasis regulation explains the low amplitude platelet count fluctuations observed in healthy individuals which cycles every 3–5 weeks.^47,48

In CTP patients, evidence suggests that the TPO-mediated adjustments of thrombopoiesis also takes place constantly resulting in reciprocal platelet/TPO oscillations at a cycling period about twice the time delay involved in thrombopoiesis. Thus, we consider the TPO oscillation and cycling period that derived from this intrinsic platelet homeostasis process as intrinsic oscillatory factors of CTP (Figure 1).

TPO oscillation of CTP

TPO oscillation is a prominent cyclic element in CTP. After the ELISA-based TPO assay became available in 1996, TPO oscillations that are reciprocal to platelet count patterns had been observed in a number of CTP patients and many other non-CTP forms of thrombocytopenia as well, such as chemotherapy-induced thrombocytopenia.^1,4,49–59 Studies of these CTP cases provided evidence that TPO oscillation caused cyclic platelet production leading to platelet oscillations. Using longitudinal whole blood transcriptome analysis, our group revealed TPO-mediated cycling of trilineage-specific gene expressions in both patients we studied, consistent with the effect of TPO on hematopoietic progenitors in addition to megakaryocyte lineage. The level of platelet-specific genes were in parallel with and preceded platelet count cycling, indicating cyclic platelet production accounting for platelet oscillations.¹ Similar to this, Zent and Connor observed oscillations of megakaryocyte or reticulated platelet numbers that preceded or correlated with platelet oscillations in their patients respectively.^51,57 Also, based on the platelet count and immature platelet fraction values reported by Yujiri and colleagues, the number of absolute immature platelets also exhibited a cyclic pattern correlating with platelet count in their patient.⁵³ The impacts of TPO on neutrophil or red cell reticulocyte production are mild and often not reflected in the cell counts, yet, there are several studies documenting fluctuations of neutrophil/reticulocyte counts that have the same periodicity with platelet count oscillations.^1,55–58,60 Notably, in about half of hydroxyurea-induced CTP cases in patients with polycythemia vera (PV), neutrophil count exhibited periodic oscillations in the same rhythm with platelet oscillation,^7,8 The mechanism is unknown but points to involvement of common myeloid progenitor cells, and we will discuss this in combination with the pathogenic factors of the hydroxyurea-induced CTP. Taken together, these data recapitulate the oscillatory circuit of TPO-mediated feedback regulation of thrombopoiesis illustrated in early mathematical models.¹⁴

Remarkably, the patients reported with similar reciprocal TPO/platelet oscillations displayed a variety of pathogenic associations ranging from clonal T cell populations,^1,4 megakaryocyte or megakaryocyte progenitor cell-specific autoantibody,⁵¹ autoimmune platelet destruction,⁵³ hormonal cycles,⁵⁷ other cytokine oscillations,^50,61 to background disease of PV, either with or without hydroxyurea treatment,^7,52,56 etc. Therefore, CTP with different pathogenic backgrounds appears to converge to the same TPO-mediated mechanism to sustain platelet oscillations.

Earlier studies explored cyclic thrombopoiesis with serial bone marrow examinations. Dan and colleagues observed sequential oscillations of megakaryocyte progenitor cells (CFU-Meg), megakaryocytes, and platelet counts correlating with the temporal order of thrombopoiesis.⁶² Nagasawa and colleagues performed another bone marrow study of 10 patients, 2 of whom showed similar oscillations of CFU-Meg and megakaryocyte, while the other patients had elevated CFU-Meg numbers throughout the platelet cycle but showed fluctuations at the megakaryocyte level, either with cyclic megakaryocyte number or the mean cytoplasmic area of megakaryocytes.⁶³ Megakaryocyte fluctuations have been shown in other studies as well.^5,64–70 However, TPO levels were not measured in these patients.

Cycling period of CTP

Based on the mathematical models of periodic hematopoiesis, the cycling period is a constant parameter of about twice the time delay involved in the feedback circuits underlying platelet oscillation. The majority of patients with CTP have a cycling period of 3–5 weeks, the same as the platelet cycling period observed in healthy individuals as well as in patients with PV or sickle cell disease (SCD), suggesting that a similar feedback circuit may underlie platelet homeostasis in healthy individuals and in CTP patients.^47,48,71,72 Furthermore, these cycling periods correlate well with the time delay of TPO-mediated platelet production. As observed in healthy individuals or in patients with cancer, a single dose of TPO administration results in platelet count increase after 5–7 days, with the induction peaking after 12–16 days and resolving after about 28 days.^48,73 These time delays also occur in TPO-receptor agonist stimulated platelet production in patients with ITP.^74,75 A higher dose of TPO resulted in a higher peak of platelet count without shortening the time delay.^73.76,77 These data further support a role of TPO-mediated feedback regulation of platelet homeostasis in the dynamics of platelet oscillation in CTP.

Nonetheless, variability in CTP cycling periods among individual patients has been observed. Platelet cycling periods of 14d were observed in a few cases.^4,6 Our group recently described a patient with an extended cycling period of 39 days, who harbors a novel heterozygous loss-of-function MPL mutation.¹ The mutant allele is impaired in TPO-stimulated cell growth and cellular pathway activation.¹ This finding suggests that loss-of-function MPL mutations might impact the cycling period, by causing a longer time delay in the TPO-mediated feedback regulation of platelet production.¹ Other patients with cycling periods of about 40–50d have been reported.^{5,55,63,66–68} Moreover, long cycles of platelet oscillations with cycling periods of 50–70d have been reported in patients with PV or chronic myeloid leukemia (CML), and the oscillations occurred either spontaneously or with hydroxyurea treatment and were coupled with oscillations of granulocytes or white blood cells.^56,78,79 The underlying mechanism is unclear and may involve instability of HSC and effects of TPO signaling in HSC. If this were the case, given the longer time required for an HSC to become a platelet, it would not be surprising if the cycle length was correspondingly increased.

Underlying pathogenic elements associated with susceptibility to CTP

Why does platelet oscillation in CTP patients fail to return to baseline, whereas it does settle in healthy individuals? A mathematical explanation is that the ultimate amplitude depends on the particular status of the homeostasis system.¹⁴ This means that the compromised platelet homeostasis background shaped the oscillation amplitude in CTP.

CTP occurs on a variety of abnormal platelet homeostasis backgrounds ranging from thrombocytopenia to myeloproliferative diseases or sickle cell disease (Figure 2). In general, the base line platelet levels can fall out on either side of the normal range. For thrombocytopenia-based CTP, diverse pathogenic elements had been proposed, including autoantibody or cell-mediated autoimmunity against platelet or megakaryocyte/progenitor cells, and menstrual cycle synchronization.^2,3 Recent studies revealed new associations such as thyroid disorders, clonal T cells, as well as genetic mutations.^1,4 These abnormalities can be considered as risk factors of CTP. On top of these abnormal platelet homeostasis backgrounds, which under normal circumstances can be compensated for, a perturbation is likely needed to initiate platelet oscillations. Currently, only a limited number of reports of thrombocytopenia-based CTP suggested existence of a trigger prior to the onset of platelet oscillations. In contrast, for patients with thrombocytosis associated with myeloproliferative diseases or sickle cell disease, platelet oscillations were predominantly triggered by hydroxyurea treatment. Here we review pathogenic factors in the context of their related platelet homeostasis background (Figure 2).

Thrombocytopenia-based CTP

Humoral, cellular, and genetic factors, or often times, the combination of these factors are associated with pathogenic background in thrombocytopenia-based CTP, also referred to as idiopathic CTP. The primary humoral factor is the presence of autoantibodies targeting the platelet homeostasis system. Nearly 40% of CTP patients were reported to demonstrate anti-platelet autoantibodies.³ Sporadic CTP cases had autoantibodies specific for megakaryocyte or granulocyte-macrophage colony-stimulating factors.^51,66 In cases with serial measurements, platelet counts were often inversely correlated with the autoantibody titer that were associated with individual platelet or remained in plasma.^6,80,81 While recognizing the limitations in laboratory assessment of platelet autoantibodies, this evidence suggests these autoantibodies could contribute to the overall thrombocytopenic background in CTP or be a perturbation initiating CTP in susceptible patients.

Besides the few cases in which autoantibodies directly affected platelet homeostasis, Steinbrecher and colleagues revealed thyroid gland disorders in all the 8 females of their total 9 patients.⁴ Several other cases revealed thyroid diseases in CTP patients.^53,54,82 Although thyroid diseases commonly have an autoimmune mechanism and often coexist with ITP, the pathogenic connection with CTP remains to be further elucidated.^4,83,84 Interestingly, Nema and colleagues reported a CTP patient with a secondary hypothyroidism due to hypopituitarism whose CTP responded to thyroxine and corticosteroid replacement.⁸⁵

Another interesting observation involving CTP and its relation to the endocrine system comes in the form of the female menstrual cycle.^2,81,86–88 Helleberg et al. described a CTP patient who presented with symptomatic, treatment-refractory thrombocytopenia that occurred during her menses, and had GPIIb/IIIa antibodies inversely correlated with platelet numbers.⁸⁷ The patient was started on low dose hormonal therapy, which was not specified in the report, resulted in a lasting stabilization of her circulating platelet numbers and resolution of her symptoms.⁸⁷ In another patient, Menakuru et al. started a patient on danazol, potentially due to its ability to reduce estrogen production, which resulted in a prolonged platelet response.⁸⁸ The effect of danazol may also be partially explained by it being a sex steroid analogue like progesterone, which can downregulate the expression of Fc receptors in monocytes, thereby decreasing immune mediated clearance of platelets.^89,90

Recent studies revealed associations of clonal T cell populations with CTP. Füreder and Fogarty reported the first 2 cases with clonal T-cell receptor (TCR) rearrangement.^5,58 Steinbrecher observed clonal TCR rearrangement in 6 of 8 patients in their CTP cohort.⁴ Our group performed next generation sequencing-based clonality analysis on 2 patients with stable platelet oscillations and revealed clonal T cell populations in both patients.¹ Collectively, these data suggest that clonal T cells appears to be a common feature among CTP patients. Moreover, the clonal populations in both of our patients had a steady, non-oscillatory longitudinal pattern throughout platelet count cycles, suggesting that the T cell clones are likely a stable factor contributing to the CTP pathogenic background¹ unlike those T cell clones recently described in several patients with ITP which were expanded when platelet counts were low.⁹¹ In line with the prediction that the amplitude of the oscillation system depends on the overall status of the hematopoietic system, we believe that a stable pathogenic background sustains stable platelet oscillations.¹⁴ In fact, our patients and many other reported CTP cases had predictable platelet oscillations that lasted for years.

The pathogenic role of the clonal T cells in CTP remains unclear. Our study raised a possible connection of these clonal T cells with somatic STAT3 mutations and large granular lymphocyte (LGL) populations that potentially contribute to the thrombocytopenia background. Both of our patients possessed somatic pathogenic gain-of-function STAT3 variants (Y640F and D661Y) with low allele frequencies in peripheral blood.¹ These variants were first identified and characterized in the CD8+ T-cell LGL leukemia and are associated with immune cytopenias, including neutropenia or pure red cell aplasia among patients with T-LGL leukemia.^92,93 Moreover, these STAT3-mutated LGL populations, often times with low allele frequency and subclinical levels of LGL, were also found in blood samples of a subset of patients with pure red cell aplasia,⁹⁴ aplastic anemia or myelodysplastic syndromes.⁹⁵ In bone marrow cells from a patient with immune-mediated aplastic anemia, STAT3 Y640F mutation was located to a CD8+ T cell clone which expressed aberrant cytotoxic genes that was distinguishable from other lymphocyte populations. The clonal size decreased with response to treatment.⁹⁶ Background (subclinical) LGL populations were observed in both of our CTP patients, and T-LGL lymphoproliferation has been reported in 2 patients with CTP.^4,5 Interestingly, increased numbers of circulating LGLs was also closely associated with adult-onset of cyclic neutropenia, which is different from the inherited form of cyclic neutropenia.⁹⁷

Besides STAT3, one of our 2 CTP patients possesses other pathogenic somatic mutations of PPM1D (R458X and L484X), which also relates to clonal hematopoiesis.^1,98 Another CTP patient had clonal T cells and somatic pathogenic DNMT3A W314X mutation (unpublished data). Collectively, the prevalence and mechanisms of the pathogenic somatic mutations and their relation to clonal hematopoiesis in CTP warrant further investigation.

In addition, several germline mutations have been described in association with CTP. Aranda and colleagues described a family with inherited cyclic thrombocytopenia of unknown genetic cause. The father and 4 of the 9 siblings had cyclic platelet oscillations.⁹⁹ Additionally, we identified a novel germline loss-of-function MPL G404R mutation in one of our patients.¹ This MPL mutation is insufficient to cause CTP, but it is part of the abnormal platelet homeostasis background and appears to shape the cycling parameters of this patient.¹ For example, this patient had very high TPO peak levels without rebound thrombocytosis, which may relate to its defect in TPO-stimulated growth.¹

Several reports suggested that infections or treatment can trigger the onset of CTP.

Wasastjerna described a patient who had history of easy bruising and developed acute CTP after mild infection of the upper respiratory tract. The cycling amplitude kept increasing after each cycle - the bleeding worsened and the subsequent platelet peaks increased; the patient died from cerebral hemorrhage during the 5th thrombocytopenic period.¹⁰⁰ In patients previously diagnosed with ITP, platelet oscillation was reported to occur after cyclosporine treatment or during their remission process, suggesting that perturbations due to treatment-induced transient platelet response and/or an alleviation in pathogenic platelet destruction may contribute to the cycling phenotype in these cases.^63,65It is worth noting that in a number of CTP cases of different types, treatment of one facet of the pathophysiology appears to lead to a partial correction of the CTP emphasizing the at least dual nature of the pathophysiology.

CTP associated with myeloproliferative diseases and sickle cell disease

CTP occurs in patients with myeloproliferative diseases, such as PV and CML.^{7,8,52,56,79,101,102} These patients have thrombocytotic base line platelet counts. In addition, substantial platelet oscillations are observed in patients with sickle cell disease (SCD), where the base line platelet count can be either normal or thrombocytotic.⁷¹ Three common aspects can be drawn related to CTP in patients with these disease backgrounds. First, their baseline diseases are caused by genetic mutations affecting HSCs: most patients with PV or CML had acquired JAK2 V617F mutation or BCR::ABL1 translocation in HSCs respectively, and patients with SCD had germline HBB mutation.^103–105 Second, spontaneous mild periodic platelet oscillations had been observed in patients with these hematological diseases and CTP cycling periods are about the same as their spontaneous cycling periods.^71,72,78 Third, except for one reported case,⁵⁶ CTP is frequently induced by hydroxyurea treatment, leading to substantial platelet oscillations.^{7,8,52,79,101,102} All these three factors could contribute to the hematopoietic landscape and the CTP phenotype developed. It remains unclear whether there are additional pathogenic factors which would delineate subgroups of patients with higher susceptibility to hydroxyurea-induced CTP.

Most reports on hydroxyurea-induced CTP were in the context of PV. Tauscher and colleagues characterized 12 cases of CTP in patients with PV, all of whom had the JAK2 V617F mutation.⁸ Hydroxyurea-induced platelet oscillation in PV had inverse TPO/platelet patterns with cycling periods of about 28 days, the same as that of spontaneous oscillations.^7,8,72 These suggest that TPO-mediated delayed feedback regulation of platelet production underlies both spontaneous and hydroxyurea-induced platelet oscillations. Moreover, nearly half of the CTP cases showed white blood cell oscillations in the same rhythm with platelet oscillations,⁸ which is different from the 15 day cycles of the spontaneous white blood cell oscillation in PV.⁷² This alludes to the instability of overall myelopoiesis and the possibility that the TPO fluctuation also impact the concurrent white blood cell oscillations perhaps via the HSCs or the common myeloid progenitor cells.

Hydroxyurea contributes to the oscillation in multiple ways. As a trigger, hydroxyurea introduced perturbations that may kickoff the oscillations or enhance the existing platelet oscillations. Hydroxyurea inhibits DNA synthesis and induces cell death in S phase, causing platelet count reduction leading to TPO accumulation.^7,106 Subsequently, the concentration-dependent TPO stimulation of megakaryopoiesis may overcome the marrow suppression of hydroxyurea and drive a platelet count rebound, triggering platelet oscillation. Similarly, hydroxyurea may enhance existing platelet oscillation if the effect of hydroxyurea and TPO levels work in synergy on platelet production, for example, when initiating hydroxyurea treatment or increasing its dose when TPO level is low. This may relate to the observation that frequent dose change, often times increasing the dose when platelet count was high and decreasing when it was low, caused bigger swings in platelet count.⁷⁹

Second, hydroxyurea altered the pathogenic landscape enabling CTP phenotype. Compared with the pretreatment stages, steady hydroxyurea treatment resulted in substantial platelet oscillations around a reduced platelet count baseline and discontinuation of the treatment stopped the oscillation.^7,8,52 The cell cycle-specific inhibition property of hydroxyurea is known to synchronize cell proliferation, and higher synchronization could increase the chance to resonate with the rhythm of TPO-mediated oscillation and amplify the oscillation amplitude. Referring to Morley’s prediction with CyN,¹⁴ the baseline reduction from the JAK2 V617F-caused overproduction may increase system “gain” and susceptibility to substantial oscillations. And this baseline shift may mirror the CTP cases induced in patients with ITP background by cyclosporine treatment or during their remission processes^63,65

Therapeutic approaches for CTP

Some patients with CTP exhibit a mild phenotype and can be managed with observation and supportive care. Additionally, spontaneous remission can occur. For example, 2 of the 9 CTP patients in Kyrle’s group had spontaneous remission after years of platelet oscillations.⁴

For CTP associated with thrombocytosis and hydroxyurea therapy, discontinuation of hydroxyurea usually stops the oscillation, and the oscillation does not occur after switching to other treatments (Figure 2).⁸ For thrombocytopenia-based CTP, due to the multiple pathologic backgrounds involved, identifying the most effective treatment modality remains challenging (Figure 3).

Figure 3. Strategies for management of thrombocytopenia-based CTP.

Identifying the underlying pathogenic conditions and finding effective treatment for individual patient remains challenging. Currently, cyclosporin A that inhibits cytotoxic clonal T cells has more successes than other treatment regimens such as oral contraceptives that exert hormonal regulations. Some patients can be managed with close observation and supportive care. Some may have spontaneous recovery. Moreover, new targeted or systematic strategies, such as STAT3 inhibitors or hypomethylating agents, may warrant further investigation.

Steinbrecher et al. reported treatment information of 9 CTP patients, where 4 had a durable remission after treatment, 3 with cyclosporine A, 1 with rituximab.⁴ Other studies on individual cases also documented clinical responses to cyclosporine A,^51,67 danazol,^{2,70,88,107,108} azathioprine,⁶² low dose hormonal contraception,⁸⁷ and thyroxine and corticosteroid replacement.⁸⁵ However, other patients have been treatment refractory to these agents and thus far selection is largely trial and error.^6,54,65,109

As TPO levels are 180 degrees out of phase with platelet count in CTP, modulation of cycling using rhTPO or TPO receptor agonists to attenuate cycling has been attempted. These TPO-mimetic agents caused exaggerated platelet oscillations, often results in extreme thrombocytosis peaks and thrombocytopenic nadirs.^{2,4,54,55,109} In one occasion, it caused a prolonged severe thrombocytopenic period after the first stimulated platelet peak.⁵⁵ Nevertheless, in two cases, reduction of the dose and frequency of TPO-mimetic agents over a period of time mitigated the oscillation amplitude with increased platelet nadir and decreased platelet peaks, provided clinical benefits.^54,109 A similar response pattern has been seen with G-CSF treatment in CyN, where G-CSF also resulted in bigger swing in neutrophil count with an increased baseline level. However, in addition to stimulating neutrophil production, G-CSF also reduces the bone marrow transit time of maturing neutrophil, thus reduces the time delay of the oscillatory circus.^110,111 As a result, G-CSF treatment shortens the neutrophil cycling period from 21d to 14d, thereby reduces the time that patients were under neutropenic condition, providing favorable clinical outcome in the cases of CyN (Table 1).^25,26 In contrast, TPO-mimetic agents do not alter the platelet cycling period. Based on the intrinsic oscillatory features of CTP, one potential direction to improve clinical efficacy of TPO-mimetic agents might be to tailor its administration schedule to compensate patients’ intrinsic TPO nadirs, which might help mitigate platelet nadirs through bone marrow thrombopoiesis.

Better identification of underlying pathogenic factors in CTP patients may help developing targeted therapies that are less toxic and more beneficial for the patients. For example, for patients possessing STAT3 gain-of-function mutations, targeting mutation-bearing cell population by STAT3 inhibitors, such as OPB-31121, may be a potential therapeutic avenue.¹¹²

Alternative treatment approaches to CTP might be drawn from examples of other disruptions of nonlinear dynamical systems perturbed into cyclical patterns, the classic example being cardiac sinus rhythm transforming to ventricular tachycardia, where sinus rhythm is restored by electrical shock. In this sense, exerting a chemical equivalent of cardioversion to the system may provide a reset to the system with resumption of a physiologic platelet cycle. One particularly interesting pharmacological candidate comes in the form of DNA hypomethylating agents such as decitabine.^113–115 In one study low dose decitabine was shown to be efficacious in patients with treatment refractory ITP.¹¹⁴ A follow up study by Ni et al. proposed that this effect is in part due to decitabine promoting the expansion of myeloid-derived suppressor cells, creating an immunosuppressive environment.¹¹³ Whether this approach can directly impact CTP patients, such as those with pathogenic somatic DNMT3A or other clonal hematopoiesis remains to be determined, but given the beneficial effects of hypomethylating agents in refractory ITP, their effects on CTP merits further clinical investigation. Better understanding of CTP will not only allow clinicians to readily identify this under-diagnosed disease, but also ideally develop minimally toxic therapeutic strategies that result in prolonged and clinically significant responses. In the ideal world, we could achieve the equivalent of “cardioversion’ and shock the counts back to a stable “normal’ with minimal amplitude platelet cycling.