Road Blocks in Making Platelets for Transfusion

Summary

The production of lab-generated human platelets is necessary to meet present and future transfusion needs. This manuscript will identify and define the major roadblocks that must be overcome to make human platelet production possible for clinical use, and propose solutions necessary to accelerate development of lab-generated human platelets to market.

Significance

Platelets (PLTs) are a critical first-line treatment for hematological diseases and trauma, for which growing demand is severely limited by blood donor supply, PLT unit outdate, bacterial contamination, and disease transmission[1, 2]. PLTs are derived exclusively from human volunteer donors and are especially vulnerable to depletion during emergencies[3]. Unlike plasma (stored frozen for years) and red blood cells (stored at 4°C for 42 days), PLTs must be stored at room temperature to maintain viability, which limits their shelf-life to only 5 days[4, 5]. In contrast to a sharp decline in red blood cell use between 2008 and 2011, PLT transfusions rose by 7.3%[3]. The market for platelets is expected to grow at 5.3% per annum over the next decade as a result of a growing and aging population[3]. The market for PLTs in the US today is over $1.4B [3] and is expected to grow to over $2B by the end of the decade. Globally, the market for blood products is six times larger[6], suggesting a global PLT market worth over $12B.

While the therapeutic use of PLTs is life-saving, storage and chain-of-custody issues, as well as their short shelf-life complicate supply management and logistics. Because bacterial screening of PLT units takes 2 days, and transport exhausts a 3rd day, blood centers typically do not have more than a 1.5-day PLT inventory available for transfusion(4–6). This results in frequent stock-outs from even minor events such as cold weather or summer vacation, and PLT unit inventory is often rapidly depleted in emergencies. PLTs cannot be refrigerated since cold temperature induces conformational and functional changes and reduces viability[7]. Unused supplies must therefore be discarded, and their cost is not recoverable, adding to their true unit cost. According to the United States Department of Health and Human Services National Blood Collection and Utilization Report of 2011, 15% of the total produced PLT supplies were left unused due to outdate. This figure does not include another 7% of the inventory ultimately characterized as “unaccounted” for[8]. Due to unmet PLT unit demand, hospital orders for PLT units went incomplete an average of 21 days in 2008 and 7.6 days total in 2011[3]. Associated inpatient costs of postponed treatments/surgeries are difficult to assess without a more detailed national study (data lacking) but add to overall healthcare cost.

Additionally, should a radiation disaster occur in the US the availability of PLTs on a large scale will be necessary to effectively treat the large population affected. An analysis of three potential attack scenarios of the United States with an improvised nuclear device conducted by the United States Health and Human Services organization indicated that under current conditions of harvesting, storing and managing blood products, a significant shortfall in PLTs would occur[9].

Because they are derived from human volunteer donors, PLT units are inherently at risk of bacterial contamination at collection (1/1000), resulting in sepsis following transfusion (1/1000–1/100,000), which can cause death (~1/500,000)[4]. The associated healthcare cost for bacterial contamination of PLT units constitutes an additional ~$120M per year in the United States. There is also the pervasive risk of viral contamination since we cannot screen for unknown viruses. This risk is exacerbated for immunocompromised patients who are especially vulnerable to bacterial/viral infection. Patients who receive multiple PLT transfusions, such as those with various types of cancer, also often develop PLT refractoriness due to human leukocyte antigen (HLA) alloreactivity and immune rejection (3/100), requiring additional transfusions with HLA-matched donor PLTs[10]. Finding alternative sources of non-immunogenic PLTs will help alleviate chronic shortages in the supply of PLTs and reduce the risks for refractoriness. Donor-independent lab-generated human PLTs can address all major logistical and safety challenges affecting the current PLT market. While recent advances in stem cell, megakaryocyte (MK) and PLT biology suggest this solution is increasingly likely, major scale, regulatory, and financial roadblocks must still be overcome to make ex vivo human PLT production possible for clinical use. This manuscript aims to identify and define these roadblocks, and propose solutions to accelerate development of lab-generated human PLTs to market.

Current State of the Art in Platelet Production

PLTs (2–3 µm in diameter) are derived from much larger (~25 µm in diameter) progenitor cells called megakaryocytes. MKs, in turn, are differentiated from CD34⁺ hematopoietic stem cells that mainly reside in the bone marrow, but are also found in the yolk sac, fetal liver, and spleen during early development[11, 12]. During MK differentiation CD34⁺ hematopoietic stem cells undergo a period of endomitosis, whereby MKs become polyploid through multiple cycles of DNA replication without cell division, yielding a polylobulated nucleus with up to 128n copies of DNA[13, 14]. As MKs increase in size they also refine their transcriptional and proteomic profiles, acquire a number of highly specialized granules including α-granules and dense-granules, and develop a highly invaginated membrane system, which are hallmarks of MK maturation and development[15–17]. To produce PLTs, MKs migrate adjacent blood vessels in the bone marrow through which they extend long structures called proPLTs into the circulation[18]. ProPLTs function as the assembly lines for PLT production, and sequentially release multiple anucleate PLTs from their ends (Supplementary Fig.1). For lab-generated PLT production to be commercially viable, lab-generated PLT yields must be at least comparable to donor-derived PLT unit collection. One apheresis-equivalent donor-derived PLT unit comprises ~3x10¹¹ PLTs. At the theoretical yield of ~2000 PLTs per MK (physiological estimate), ~1.5x10⁸ MKs are required to generate one PLT unit. For context, ~5x10⁶ CD34⁺ stem cells are collected from a single human umbilical cord-blood (UCB) unit[19], and can vary depending on stem cell source. If 1 CD34⁺ stem cell can be expanded to yield ≥1000 MKs over multiple passages under standard culture condition (as is the case with UCB stem cells), at least 1x10⁵ CD34⁺ UCB stem cells are required to produce ~1.5x10⁸ MKs. Ex vivo human PLT production therefore hinges on 3 major questions:

1. What is the stem cell source?

2. What is the process of MK differentiation?

3. What is the approach to PLT production?

1. Stem cell source

Today, human MKs can reasonably be derived from 3 major stem cell sources[1, 20]: human CD34⁺ embryonic stem cells (ESCs); human CD34⁺ umbilical cord blood stem cells (UCBs); and human induced pluripotent stem cell (iPSCs). Other stem cell sources that will not be discussed because their application for PLT production is currently scale-prohibitive include bone marrow, fetal liver, and peripheral blood[21].

ESCs are pluripotent stems cells derived from the inner cell mass of an early-stage preimplantation embryo called a blastocyst[22]. Multiple human ESC lines are currently available, are validated for clinical application[23, 24], and have been used to derive MKs and PLTs previously[25–27]. Drawbacks include a limited expansion capability and ethical concerns, as generation of ESCs requires the destruction of an embryo. Because they are primary cells, there is also the concern of viral contamination risk, HLA incompatibility, and congenital diseases; and ESCs should be genetically screened prior to use. This can lead to privacy issues and questions regarding ownership of the donated cells that must be addressed for regulatory approval at collection. In the United States, regulatory requirements around Good Manufacturing Practice (GMP)-compliance also require that a current medical history be available for a stem cell donor (see Supplementary Material: Regulatory Requirements for Clinical Application).

UCBs are pluripotent stem cells derived from blood that remains in the placenta and the attached umbilical cord after childbirth[28]. Human UCBs possess many of the same benefits and drawbacks as human ESCs[19], but are potentially less ethically tenuous since the placenta and umbilical cord blood are a byproduct of childbirth that is otherwise discarded[29]. A major drawback of employing UCBs is that their availability is donation-dependent, which makes it more difficult to scale PLT production, and increases their risks of viral/bacterial contamination and HLA incompatibility. Because human UCBs must be donated/purchased on a regular basis their use will likely lead to higher overall PLT unit cost in the long term.

iPSCs are a type of pluripotent stem cell that can be generated from adult cells by inducing timed expression of particular transcription factors, and engineered to produce MKs and PLTs[30]. Pioneered by Shinya Yamanaka[31], iPSCs can be expanded and potentially maintained in culture indefinitely, supporting scalability and helping further reduce PLT unit cost. Because they are generated from adult cells, iPSCs are also the least ethically tenuous stem cell source of the three.

If developing a pluripotent stem cell line, the stem cell source should be derived from female Type 0 Rh- (ABO blood type). This is necessary to reduce the risk of immune-mediated complications following infusion. In addition, PLTs express 3 HLAs corresponding to major histocompatibility complex (MHC) class I (A, B, and C). PLTs do not express HLA class II. Because multiple alleles exist for each polymorphic HLA gene, there is no ‘universal donor.’ Potential approaches to minimize rejection of mismatched infusions have been described previously[32, 33]. Algorithms to identify maximal homozygous HLA -A, -B, -DR matching combinations in targeted populations have yielded maximal zero HLA mismatch rates of ~17%[33]. To address HLA incompatibility issues resulting in PLT refractoriness, stem cells should be genetically engineered to remove HLA-A, B and C class I antigens[34]. Riolobos et al. recently demonstrated proof of concept by knocking out HLA antigens in ESCs via two distinct genetic approaches; 1) gene targeting and mitotic recombination to derive HLA-null ESCs; and 2) targeted disruption of both alleles of the β2 microglobulin (B2M) gene in ESCs[32]. Because the absence of B2M can lead to lysis by Natural Killer cells (“missing self” response), B2M-/- cells should be engineered to express recombinant class I molecules, such as a non-polymorphic HLA-E molecule. When selecting a gene editing approach one should avoid technologies that introduce double-stranded breaks, generate of off-target alterations to the genome, produce unwanted mutations at the target site, or introduce nuclease genes into the cell which may have unintended effects and can result in tumorigenicity. It is important that following genetic engineering, the stem cells are screened for possible off-target integrations or effects and that only the parent pluripotent stem cell containing the correct integration be expanded. This will need to be validated again by screening the engineered stem cell line during release testing, as is required for GMP compliance. Because PLTs are anucleate, there is no risk that wild-type DNA viruses prevalent in the human population could lead to replication and packaging of transgenes during superinfection by a helper virus, leading to viral mobilization. PLT refractoriness affects only ~5% of patients receiving leukoreduced PLTs, and the engineering of HLA-null PLTs should be considered as a follow-on approach after ‘general’ lab-generated PLTs have been advanced through the regulatory process and into commercial use[35, 36]. This is desired to minimize risks associated with genetic engineering, avoid perceived risk that could delay clinical trials or market adoption, and accelerate the time to market.

In every case, CD34⁺ hematopoietic stem cells must by expanded and maintained in cell culture using defined growth medium (oftentimes with the addition of serum) that usually contains a combination of cytokines such as stem cell factor (SCF, also known as Kit ligand), thrombopoietin (TPO), FMS-like tyrosine kinase 3 (Flt3L), angiopoietin-like 5 (ANGPTL5, insulin-like growth factor binding protein 2 (IGFBP2), insulin-like growth factor 2 (IGF-2), interleukins 3 and 6 (IL-3, -6), granulocyte colony stimulating factor (G-CSF), and notch ligands 45 and 46, among others[1, 37]. Stem cell cultures also typically rely on embryonic fibroblast feeder cells that can potentially be contaminated with xenogeneic pathogens and increases the risk for an immunogenic reaction in humans. Alternatives to animal serum (often horse or bovine in origin) and feeder cells (often mouse) are desired to avoid the introduction of animal products into human tissue culture, to reduce or to avoid harvesting of cells/serum from animal fetuses, and to ensure safe and animal product-free conditions for GMP[38]. Because iPSCs offer greater genetic control of the parent stem cell and can potentially be maintained in culture indefinitely, we believe that human iPSCs will eventually provide the least risky and most scalable/economical source of PLTs in the future, and therefore represent the most attractive stem cell source from which to advance human PLT production today.

2. Process of MK differentiation

There are currently multiple ongoing efforts to optimize MK differentiation and maturation ex vivo, which have mostly comprised defined growth mediums containing different combinations of hematopoietic cytokines such as SCF, TPO, Flt3L, IL-1, IL-3, IL-6, IL-9, IL-11, stromal-derived factor 1 (SDF-1), and platelet-derived growth factor BB (PDGF-BB) among others. A number of excellent papers have recently reviewed current approaches to MK differentiation and scale-up[1, 2, 39, 40], which will therefore not be discussed here. Validation of MK differentiation should include the following exhaustive list of early quality assessment metrics. As this field advances, efforts should be made to narrow these criteria to a smaller targeted set of metrics that predict (or otherwise correlate with) PLT functionality and safety.

1. Morphology

2. Ultrastructure

3. Cytoskeletal organization

4. Granule content

   • •Immunofluorescence microscopy of CD31, LDL uptake, VWF, β1-tubulin, Phalloidin, Serotonin, Thrombospondin-4, Platelet Factor 4

5. Biomarker expression

   • •Flow cytometry analysis of CD31, CD34, CD43, VE-Cad, CD144, CD105, CXCR4, KDR, CD14, CD41a, CD61, CD42a, CD42b, CD235a, and GP6 expression

6. Ploidy

7. Gene Expression

   • •Relative mRNA expression of MYC, GATA1, NFE2, MEIS1, PBX1, RUNX1, ZFPM1

8. ProPLT Formation and PLT Release

Primary human MKs should serve as a physiological control[41]. To bypass current inefficiencies in MK differentiation Nakamura et al. recently developed stable, immortalized MK progenitor cell lines (imMKCLs) from pluripotent stem cell-derived hematopoietic progenitors by over-expressing both BMI1 and BCL-XL[20]. The resulting immortalized MKs could be expanded in culture over extended periods of time (4–5 months), and support cryopreservation, which is desirable to support rapid “surge” capacity when large numbers of PLTs are needed. However, imMKCLs are dependent on virus-mediated genetic reprogramming, and serum and feeder cell culture, which is a potential source of viral and other-cell contamination, increased costs, and is a major rate limiting step during scale-up.

In collaboration with Ocata Therapeutics, we have shown that it is feasible to generate functional MKs from both human ESC[25] and iPSC[34] cultures. This method also permits the cryopreservation of MK progenitors, which can be thawed, differentiated, and triggered to produce PLTs within a few days. Most importantly, this approach employs a serum- and feeder cell layer-free protocol which decreases the risk of an immunogenic reaction in humans, improves scalability, increases time efficiency from MK progenitor to PLT and decreases the overall cost of PLT unit generation. Ultrastructural/morphological analyses showed no major differences between iPSC PLTs and blood PLTs. iPSC PLTs formed aggregates, lamellipodia, and filopodia after activation and circulated in macrophage-depleted animals and incorporated into developing mouse thrombi in a manner identical to human PLTs.

Because differentiated MKs today are mostly derived from embryonic/fetal tissues or modeled after gene transcription profiles from these tissues, resulting MKs are likely more ‘neonatal-like’ than ‘adult-like’. This may be desired as Liu et al recently showed that mouse neonatal PLTs survive longer than their adult counterparts in vivo, which may extend lab-generated PLT unit storage[42]. Nevertheless, efforts should be made to model adult MK differentiation in cell culture to determine if this might represent a more suitable PLT source.

3. Approach to PLT Production

Currently, the most significant obstacle to ex vivo PLT production has been triggering MKs to produce PLTs at yields necessary to make production of a PLT transfusion unit both clinically and financially practical. While functional human PLTs were first grown in vitro in 1995[21], to-date only ~10% of human MKs initiate PLT production in culture. Evidence suggests that key physiological characteristics of the bone marrow environment including extracellular matrix composition[43, 44], bone marrow stiffness[45], endothelial cell contacts[46, 47], and vascular shear stresses[48, 49] support proPLT production and may help trigger PLT release. Differences in fetal versus adult PLT production support these observations and suggest that differences in MK progenitors can affect the functionality and lifespan of the resulting PLTs[42, 50–52]. Approaches to PLT production for clinical use can generally be classified into two major categories: in vivo and ex vivo. A number of excellent papers have recently reviewed current approaches to PLT production [1, 2, 39, 40], and will therefore not be described in detail here.

In vivo

One approach has been to infuse mature MKs intravenously with the goal using the lung capillary bed as a means to trigger PLT production[53–58]. When infused by retro-orbital or tail-vein injection, mouse fetal liver cell culture-[59] and bone marrow-derived[53] MKs primarily trap in the lungs and release PLTs over a period of 24 hours. MKs are very rarely observed in the lungs of healthy animals, and previous studies have shown that the resulting PLTs are rapidly cleared, limiting their potential effectiveness and raising the concern of unwanted activation/aggregation and immunogenicity in the pulmonary microcirculation. Recent studies by Fuentes et al. suggest that a second wave of PLTs may be generated by infused MKs in the lung that are longer lived[53]. While the physiological significance of the lung as a site of primary PLT production remains unclear, the lung might nevertheless provide the ideal microenvironment to trigger PLT production and warrants further study.

Additionally, Xi et al. recently performed a Phase I clinical trial to determine the safety and feasibility of directly infusing a median of 5.45x10⁶ cord blood-derived CD41⁺ MK progenitor cells/kg of body weight in 24 patients with hematological malignancy (12–60 years of age; post-chemotherapy; 20–60x10⁹ PLTs/L; no serious damage to liver and kidney function)[60]. MK progenitor cells were not matched for ABO or HLA, and patients were monitored weekly for 4 weeks and after 1 year post-infusion. No severe adverse effects were observed in patients, although two patients experienced fever (<38°C), four patients exhibited increased transaminases (present before MK progenitor cell infusion), one patient experienced increased jaundice (also present prior to MK progenitor cell infusion), and one patient experienced vomiting – all of which were resolved with standard clinical treatment. Graft-versus-host disease was not observed at the 1 year time point. Of the 19 patients who did not receive a further PLT transfusion, PLT recovery (>100x10⁹ PLTs/L) occurred in only 12 patients within an average of 8 days following MK progenitor cell infusion. This is not a substantial improvement over TPO mimetics that take ~5 days to increase PLT counts and ~12 days to reach maximal effect[61]. Although there have been some advancements with the direct MK infusion method, the potential pitfalls will likely outweigh the benefits. The major issue will be regulating time to PLT production when comparing infusion of MKs versus direct infusion of lab-generated PLTs. Also, to be clinically approved, it will be necessary to determine how many MKs/progenitor cells are necessary to achieve 3x10¹¹ PLTs (human transfusion unit) in vivo, particularly if MKs/progenitor cells are becoming trapped in the microcirculation of the lung and can pose complications from blood vessel blockages and resulting inflammation or immunogenicity (eg. Transfusion-related acute lung injury, TRALI).

Unlike PLTs, MKs/progenitor cells are nucleated and also pose an oncogenic or teratogenic risk if derived from a genetically modified stem cell source[62], although there is evidence to suggest that mature MKs (but not MK progenitors) may be amenable to irradiation[63–65]. As with lab-generated PLTs, quantitation and quality assessment of PLTs resulting from infusion of MKs/progenitor cells are necessary and will be complicated by the need to distinguish/isolate them from other circulating PLTs in the blood/lungs or other tissues.

Ex vivo

A complementary approach has been to develop scalable, MK/progenitor cell-compatible PLT bioreactors that employ biologically-inspired engineering to integrate the major chemical and physical components of the bone marrow stroma within micro- and macro-fluidic systems[66]. PLT bioreactors simplify the regulatory pathway (see Supplementary Material: Performance Requirements and Quality Assessment for lab-generated PLTs) and because they are independent of MK source, are also easily adapted to future advancements in stem cell culture. Attempts to model vascular flow have included perfusing MKs/progenitors over extracellular matrix-coated glass slides[48, 67] and centripetally agitating MKs in an incubator shaker[49, 68], both of which have yielded shear-mediated increases in PLT production. Coating bioreactor surfaces with extracellular matrix proteins (eg. fibrinogen, fibronectin, laminin, collagen type IV), and regulating media composition (eg. SDF-1, RANTES, S1P) have also been shown to improve PLT quality and increase PLT yield[46, 48, 49, 68–73], although their use can further complicate the regulatory pathway. Nevertheless, extracellular matrixcoated silk-based tubes that reproduce bone marrow sinusoids[74], and modular 3D knitted polyester scaffolds perfused under continuous fluid flow[75] have suggested that physiological PLT production can be replicated ex vivo and clinically useful numbers of lab-generated human PLTs are attainable.

We recently developed a scalable microfluidic PLT bioreactor that recapitulates many of the major characteristics of human bone marrow including stiffness, extracellular matrix composition, micro-channel size, hemodynamic vascular shear stress, and endothelial cell contacts[66]. By exposing MKs to physiological shear stresses of ~600 mPa we showed an improved rate and extent of PLT production to yield 30–70 PLTs per MK (1x10⁶ PLTs per 300 µL)[66]. In our bioreactor, time to initiation of PLT production was reduced from 6 hrs. to immediately, the percent of PLT-producing progenitors was increased from 10% to more than 90%, the time to completion of PLT production was reduced from 18 hours to two hours, and PLT morphology, ultrastructure and function of our lab-generated PLTs was consistent with blood PLTs[66]. To assess lab-generated PLT quality/safety in vitro and perform function studies in vivo, bioreactors will need to achieve a total yield of at least 1x10⁸ PLTs per 300 µL.

To achieve 3x10¹¹ PLTs per 300 mL (human transfusion unit), bioreactor designs will need to be further adapted to support continuous media perfusion and parallelization, maximize MK/progenitor cell zonal distribution and trapping, and equalize shear stress exposure, while limiting hemodynamic-activation of PLTs[76]. This is best approached by generating 3D computer models of the bioreactors that include membrane/channel porosity, tortuosity and void fraction, and computational modeling of their fluid dynamics based on Finite Volume numerical schemes. Fluid properties should be adjusted to match culture media density and viscosity, and computer-generated flow simulations should be compared to physical fluid modeling and corrected to reflect experimental data. While yields of ~100 PLTs per MK are sufficient to support pre-clinical/clinical studies, yields of ≥1000 PLTs per MK (physiological PLT yield) will be needed so that lab-generated PLTs can be priced competitively relative to donor-derived PLTs. Required PLT yields per MK and resulting cost structures have been reviewed prior[1, 7], and are discussed in greater detail in Supplementary Material: Financial Requirements for Commercialization.

Conclusion

PLTs are presently derived entirely from human volunteer donors, and are of limited supply, and have safety concerns. While it is currently possible to isolate and culture human MKs from ESCs, UCB, and iPSCs, the major bottleneck remains triggering human MKs to produce PLTs at clinically and commercially viable scale. By identifying and reproducing key physiological triggers of PLT production in scalable micro/macro-fluidic bioreactors, we believe it will be possible to achieve less risky and functional human lab-generated PLTs at yields necessary for clinical use, and costs amenable to market adoption. Recent advances in stem cell, MK and PLT biology suggest this solution is finally within our grasp. By overcoming the remaining scale, regulatory, and financial roadblocks, it is increasingly likely that cost-effective lab-generated human PLTs will be realized in our lifetime.