Emerging Cellular and Gene Therapies for Congenital Anemias

Abstract

Congenital anemias comprise a group of blood disorders characterized by a reduction in the number of peripherally circulating erythrocytes. Various genetic etiologies have been identified that affect diverse aspects of erythroid physiology and broadly fall into two main categories: impaired production or increased destruction of mature erythrocytes. Current therapies are largely focused on symptomatic treatment and are often based on transfusion of donor-derived erythrocytes and management of complications. Hematopoietic stem cell transplantation represents the only curative option currently available for the majority of congenital anemias. Recent advances in gene therapy and genome editing hold promise for the development of additional curative strategies for these blood disorders. The relative ease of access to the hematopoietic stem cell compartment, as well as the possibility of genetic manipulation ex vivo and subsequent transplantation in an autologous manner, make blood disorders among the most amenable to cellular therapies. Here we review cell-based and gene therapy approaches, and discuss the limitations and prospects of emerging avenues, including genome editing tools and the use of pluripotent stem cells, for the treatment of congenital forms of anemia.

Introduction

Hematopoietic stem cells (HSCs) are rare cells in the bone marrow that give rise to progressively more lineage-restricted progenitor cells that ultimately form all of the cellular components of the blood [Orkin and Zon, 2008]. Erythropoiesis is the coordinated process of differentiation by which these HSCs and progenitor cells (HSPCs) are able to produce erythrocytes responsible for the transport of oxygen [Palis, 2014]. This process is characterized by the massive accumulation of hemoglobin, distinct changes in morphology, and enucleation of the cells prior to entry into the circulation (Fig 1A). Human erythrocytes have a life span of 120 days and under homeostatic conditions healthy adults produce approximately two million erythrocytes every second to balance production and turnover [Palis, 2014]. The major cytokine that regulates erythropoiesis, erythropoietin (EPO), is produced by interstitial fibroblasts in the kidney that sense tissue oxygenation [Bunn, 2013].

Figure 1.

(A) A model for erythropoiesis. Early progenitor cells undergo progressive maturation through successive cellular stages that are accompanied by distinct changes in morphology. Burst forming unit – erythroid (BFU-E); colony forming unit – erythroid (CFU-E); proerythroblast (ProE); basophilic erythroblast (BasoE); polychromatic erythroblast (PolyE); orthochromatic erythroblast (OrthoE); reticulocyte (Retic); red blood cell (RBC). (B) Therapeutic avenues for the treatment of congenital anemias. Gene correction and editing of patients’ autologous cells presents a highly desired strategy to overcome limitations presented by allogeneic HSC transplantation and limited availability of HLA-matched allogeneic donors. Generally, this involves the isolation of the patient’s HSPCs from the bone marrow and subsequent exposure to gene modifying agents using viral or other transient delivery methods. Depending on the genetic entity of the disease, correction may be achieved by the introduction of a functional gene, repair of the mutated gene at its endogenous locus by homology directed repair (HDR) (e.g. the SCD mutation), or by targeting a modifier gene or its regulators known to ameliorate disease (e.g. knockdown or disruption of BCL11A in the hemoglobinopathies). The corrected cells would subsequently be re-infused into the patient to allow repopulation of the bone marrow and generation of healthy cells. Patient-specific induced pluripotent stem cells (iPSCs) derived through reprogramming of somatic cells present another possible approach. They have an unlimited propagation capacity allowing for more faithful genetic modification as they can be more extensively screened for adequate correction compared to manipulated HSPCs. Subsequently, they may be directed to differentiate into hematopoietic cells. Alternatively, a patient’s somatic cells could be directly converted into hematopoietic cells. However, in both instances the derivation of HSPCs with multi-lineage potential and engraftment capability remain key challenges to be overcome.

Anemia is defined as a decrease in the number of circulating erythrocytes. While anemia is often acquired, monogenic disorders leading to anemia are among the most prevalent genetic conditions, particularly among children [Sankaran and Weiss, 2015; Weatherall, 2010]. In these congenital forms of anemia, numerous distinct defects have been identified, affecting diverse aspects of erythroid physiology, including hemoglobin production, membrane stability, metabolism, vesicular trafficking, and ribosome biogenesis [Sankaran and Weiss, 2015]. These defects lead to reduced or ineffective production and maturation of erythroid cells in the bone marrow, or a shortened life span due to increased destruction of mature erythrocytes. Some forms of anemia involve a combination of these defects. Treatment remains largely supportive and symptomatic, consisting of transfusion with healthy donor erythrocytes and management of specific complications. Many patients with such disorders have significant morbidity and mortality, despite receiving the best available therapies [Marsella and Borgna-Pignatti, 2014; Yazdanbakhsh et al., 2012]. Numerous studies have provided important insight into the molecular pathophysiology of these disorders and have led to the development of a variety of compounds, with many under pre-clinical and clinical development [Archer et al., 2015; Ataga and Stocker, 2015]. However, allogeneic transplantation of HSPCs from a healthy donor (ideally a matched sibling) into a conditioned and myeloablated recipient is currently the only curative option available in a clinical setting for these disorders [Lucarelli et al., 2012]. Nevertheless, such transplants can be limited by human leukocyte antigen (HLA)-matched donor availability and transplant-associated complications, such as graft-versus-host disease and graft failure, which can cause considerable morbidity and mortality [Kekre and Antin, 2014; Mattsson et al., 2008; Petersdorf, 2013]. Given these limitations, the development of approaches to genetically manipulate autologous HSPCs holds considerable promise for improved therapies for congenital anemias. Advances in the field of gene therapy and genome editing now raise the prospect of correcting the genetic defect as a curative and more broadly available option. This review will focus on these and other emerging approaches for cellular and gene therapies for congenital forms of anemia.

Gene therapy approaches in congenital anemias

The frequent occurrence of monogenic congenital anemias and the relative ease of isolation of human HSPCs, as well as the possibility of gene modification or correction of HSPCs ex vivo and subsequent reinfusion into an affected patient, make such disorders prime candidates for these emerging therapeutic approaches (Fig 1B) [Sankaran and Weiss, 2015]. Significant advances in gene therapy methods and recent progress in the field of genome editing, such as use of the clustered regularly-interspaced short palindromic repeats (CRISPR)/Cas9 and other approaches, suggest that genetic correction of specific mutations may be feasible in the future [Gupta and Musunuru, 2014; Hsu et al., 2014]. In classical gene therapy, various approaches have been pursued to meet the challenges posed by the diverse set of congenital anemias, which can be caused by a single mutation of a single gene (e.g. sickle cell disease), a variety of different mutations at a single gene locus (e.g. β-thalassemia), or a number of different genes. Potential gene therapy or editing strategies include the introduction of a functional gene, the repair of a gene at its endogenous locus, or targeting a modifier gene known to ameliorate disease severity. While individually tailored approaches are possible in principle, there are practical limitations in applying these advances to disorders caused by diverse mutations that occur in multiple genes.

The addition of an intact gene by means of gene transfer is a strategy long pursued, particularly for the treatment of sickle cell disease (SCD) and the thalassemia syndromes, collectively known as the hemoglobinopathies. They represent the most prevalent forms of congenital anemia in the world and are caused by mutations in the hemoglobin genes [Sankaran et al., 2010]. Thus, the discussion of gene therapy will focus largely on the hemoglobinopathies, with discussion of other conditions at the end of this section.

The human genome has two genes for α-globin (HBA1 and HBA2) on each copy of chromosome 16 and a single gene for β-globin on each copy of chromosome 11 (HBB). The adult form of human hemoglobin (HbA, α₂β₂) consists of two α- and two β-globin chains forming a heterotetramer with each subunit bound to the iron-containing heme prosthetic group responsible for reversible oxygen binding. Thalassemia is caused by mutations that inhibit the production of either the α- or β-globin chains, which results in an imbalance of both chains required for the formation of the functional HbA tetramer. In β-thalassemia, reduced production of β-globin chains leads to aggregation of excessive free α-globin molecules that damage and destroy erythroid precursors, leading to ineffective erythropoiesis. The resulting anemia and compensatory mechanisms lead to organomegaly and erythroid hyperplasia with cortical thinning resulting in bone abnormalities. In α-thalassemia, there is an excess of free β-globin subunits, which are more stable than free α-globin molecules, and the disorder primarily manifests with hemolysis of mature erythrocytes in the circulation. As there are a total of four α-globin genes, the deletion of one or two copies results in no or mild anemia, while the loss of three α-globin genes leads to hemolytic anemia of variable severity, and deletion of all four α-globin genes is incompatible with life [Galanello and Origa, 2010; Harteveld and Higgs, 2010; Marengo-Rowe, 2007; Peters et al., 2012; Piel and Weatherall, 2014]. In SCD, the substitution of glutamate by valine at position 6 in the β-globin chain leads to the formation of a qualitatively defective hemoglobin S (HbS). The replacement of a hydrophilic amino acid with a hydrophobic residue creates rigid linear HbS polymers upon deoxygenation, deforming red blood cells (RBCs) into the characteristic sickle cell morphology. The sickled RBCs adhere to each other and the endothelium of small blood vessels, predisposing the patient to vaso-occlusive events characterized by tissue hypoxia / ischemia and resultant organ damage. Consequently, SCD is a multi-systemic disorder with numerous clinical manifestations that arise from impaired oxygen delivery to tissues [Frenette and Atweh, 2007; Rees et al., 2010; Steinberg, 1999].

Accordingly, the delivery of a functional β-globin gene has long been a primary strategy to treat SCD and β-thalassemia. However, attempts to use integrating γ-retroviral vectors initially failed to achieve sufficient and erythroid-specific transgene expression when gene expression was driven by viral long terminal repeats (LTRs) [Karlsson et al., 1988]. Appropriate expression was later enabled by the discovery of essential endogenous regulatory elements, such as the locus control region (LCR), a critical aggregate of upstream enhancers of the β-globin gene cluster [Talbot et al., 1989]. Incorporation of components of the LCR into gene transfer vectors significantly enhanced the transcription of the delivered globin genes [Plavec et al., 1993; Sadelain et al., 1995]. An additional obstacle in successful gene therapy was the silencing of transgenes randomly integrated into the genome and the altered expression of genes adjacent to the site of integration due to newly inserted regulatory elements. These problems were overcome through the identification and use of chromatin insulator elements to shield repressive and enhancing effects of endogenous and exogenous chromatin elements, respectively (Figure 2) [Evans-Galea et al., 2007; Rivella and Sadelain, 1998]. Deregulated gene expression due to a tendency of γ-retroviral vectors to insert near transcriptional start sites, frequently resulting in insertional oncogenesis due to integration near proto-oncogenes, led to the more frequent use of lentiviral vectors that exhibit a safer integration profile [Biffi et al., 2011; Cattoglio et al., 2007; Kvaratskhelia et al., 2014]. Nonetheless, lentiviruses have a relatively random integration pattern and tend to insert in actively transcribed genes [Schroder et al., 2002]. Furthermore, lentiviral vectors are more efficient at transducing the primary target cells for gene therapy, the quiescent HSCs. A sufficient level of modification in such HSCs presents a significant bottleneck for gene therapy approaches to be successful. To further enhance safety, modern self-inactivating (SIN) vectors carry deletions in their 3’ LTR to further minimize transactivation of nearby genes [Miyoshi et al., 1998]. These findings highlight the principal challenges for all approaches of gene therapy relying on gene integration to enable high-level and lineage-specific expression without compromising safety.

Figure 2.

Simplified vector scheme illustrating a commonly used organization of a lentiviral vector for gene therapy. The vector contains the therapeutic gene of interest and transcriptional regulatory elements driving its expression. These elements are in reverse orientation with respect to viral transcription to avoid interference with viral processing, as is often the case in such vectors. The long terminal repeat region (LTR) is segmented into U3, R and U5 regions. For safety, the U3 repeat region is in a self-inactivating (SIN) configuration (Δ). An insulator element (I) has been introduced into the LTR to shield transcriptional effects of endogenous and exogenous chromatin elements after integration.

A number of β-globin gene-containing lentiviral vectors were subsequently shown to correct SCD and β-thalassemia in mouse models [Levasseur et al., 2003; May et al., 2000; Pawliuk et al., 2001] and in human HSPCs transplanted into immunodeficient mice [Breda et al., 2012; Puthenveetil et al., 2004]. A recent study reported the lentiviral transduction of human bone marrow (BM) derived HSPCs from SCD donors to achieve exogenous expression of the human β-globin gene, which prevented sickling of differentiated erythroid cells [Romero et al., 2013]. The anti-sickling β-globin gene activity was retained for at least three months following transplantation into immunodeficient mice, demonstrating a sustained therapeutic effect. Several groups have also evaluated the utility of synthetic β-globins or the use of γ-globin for therapeutic purposes. β-globin variants carrying substitutions at E22A and T87Q have potent anti-sickling properties by affecting axial and lateral contacts essential for the formation of linear HbS polymers [McCune et al., 1994; Negre et al., 2016]. The application of γ-globin vectors stems from the observation of potent anti-sickling effects of fetal hemoglobin (HbF, α₂γ₂), which ameliorates disease morbidity and mortality in SCD and β-thalassemia syndromes [Musallam et al., 2012; Perumbeti et al., 2009; Pestina et al., 2009; Platt et al., 1994; Wilber et al., 2011]. In fact, patients with either SCD or β-thalassemia and hereditary persistence of fetal hemoglobin exhibit reduced disease severity and in some cases are entirely asymptomatic [Musallam et al., 2012; Sankaran, 2011; Sankaran and Orkin, 2013]. In addition to a direct benefit of replacing defective or dysfunctional β-globin chains in the hemoglobinopathies, HbF also allows for increased RBC survival, so such replacement strategies result in additional alleviation of pathogenesis.

The expression of fetal hemoglobin is primarily confined to the gestational period, after which it is replaced by adult hemoglobin during the first several months of life [Sankaran and Nathan, 2010]. Consequently, identifying the genetic factors contributing to persistent fetal hemoglobin levels and the mechanisms related to the developmental switch of the hemoglobin genes have been areas of intense investigation over many decades. A major discovery came from follow up of genome-wide association studies (GWAS) that identified common polymorphisms associated with fetal hemoglobin levels in the BCL11A gene [Lettre et al., 2008; Menzel et al., 2007; Uda et al., 2008]. Although expressed in the brain and in lymphocytes, there was no known role for BCL11A in erythroid development. Seminal work showed that BCL11A was a major transcriptional regulator of the switch from fetal γ- to adult β-globin expression, and knockdown of BCL11A significantly increased HbF expression without altering erythroid development [Sankaran et al., 2008; Sankaran et al., 2009]. As expected, deletion of Bcl11a in a transgenic mouse model of SCD corrected the associated hematologic and pathologic defects through induction of HbF [Xu et al., 2011]. Importantly, three patients with heterozygous microdeletions of chromosome 2p15p16 encompassing BCL11A presented with developmental delay and autism spectrum disorder, but they had no signs of compromised hematologic and immunologic function and exhibited persistence of HbF expression with levels ranging from 16–30% [Basak et al., 2015]. Subsequent follow-up studies of additional patients with BCL11A deletions or mutations have supported these findings [Dias et al., 2016; Funnell et al., 2015]. These findings support the rationale for partial suppression and hematopoietic-specific therapeutic targeting of BCL11A to treat the hemoglobinopathies, as complete deletion of Bcl11a has been shown to severely affect murine lymphoid development [Yu et al., 2012]. To avoid potential side effects in non-erythroid cells by short hairpin RNA (shRNA) mediated knockdown of BCL11A, erythroid-specific expression cassettes have been developed for use in potential gene therapy vectors [Guda et al., 2015]. The delivery of such shRNAs using SIN lentiviral vectors resulted in HbF induction to levels more than 50% of total hemoglobin in human erythroid cultures, although further studies of such vectors are needed. Furthermore, RNA polymerase II-driven shRNAs appeared to be less cytotoxic than RNA polymerase III-based vectors – perhaps due to lower overall expression levels of the shRNAs – implying possible preferences for using polymerase II-based strategies in clinical settings more generally [Brendel et al., 2016; Guda et al., 2015]. Further strategies may involve the targeting of erythroid-specific regulatory elements of BCL11A instead of coding sequences, an approach discussed in more detail below. Currently, a number of clinical trials utilizing various vector systems aiming to introduce globin variants, as well as approaches to induce HbF for the treatment of SCD and β-thalassemia by targeting BCL11A, are open or in preparation and have recently been reviewed in detail [Archer et al., 2015; Hoban et al., 2016b].

Recently, several additional potential targets for HbF induction have emerged. The leukemia/lymphoma-related factor (LRF), a transcription factor that is encoded by the ZBTB7A gene, has emerged as another potential target. Disruption of ZBTB7A induces HbF in primary human erythroblasts [Masuda et al., 2016]. LRF and BCL11A suppress γ-globin via potentially distinct mechanisms, and concomitant knockout of both genes results in higher fetal/adult globin ratios in erythroid cell lines than knockout of each gene alone [Masuda et al., 2016]. However, LRF is necessary for normal erythropoiesis as well as lymphocyte production, so it is unlikely to be an ideal therapeutic target [Maeda et al., 2007; Sakurai et al., 2011]. An additional potential target is the hematopoietic transcription factor MYB, whose suppression in erythroid progenitors is associated with elevated HbF [Galarneau et al., 2010; Lettre et al., 2008; Sankaran et al., 2011; Stadhouders et al., 2014]. Furthermore, common variants in the hematopoietic enhancers of this gene significantly affect not only on HbF levels, but also the clinical course of SCD and β-thalassemia, suggesting potential beneficial roles of these variants, which warrant more detailed study [Sankaran et al., 2013; Sankaran and Orkin, 2013; Stadhouders et al., 2014]. Rare mutations in the erythroid transcription factor KLF1 have also been identified in individuals with elevated HbF. However, KLF1 mutations in humans can cause variable phenotypes, including severe forms of anemia, and are not always associated with elevated HbF levels [Arnaud et al., 2010; Borg et al., 2010; Giardine et al., 2011; Sankaran and Orkin, 2013; Satta et al., 2011]. Another strategy for HbF induction has been the forced chromatin looping by tethering the self-association domain of Ldb1 to an artificial zinc finger protein targeting the γ-globin promoter [Deng et al., 2014]. In primary adult human erythroblasts, this resulted in increased production of γ-globin mRNA and HbF with a concomitant reduction in adult β-globin expression. It remains to be investigated whether this strategy and its therapeutic effect may be applicable in the long-term in vivo.

Modulation of the expression of the endogenous δ-globin gene may present an alternative strategy to increase the level of hemoglobin A2 consisting of two α- and two δ-globin chains (HbA-2, α₂δ₂). HbA2 is physiologically expressed at low levels in adults and could be a potential substitute for HbA in the hemoglobinopathies. In a proof-of-principle experiment, the activation of δ-globin in transgenic mice led to a robust improvement of hematologic parameters in β-thalassemic mice [Manchinu et al., 2014]. Therefore, it is possible that this may be an alternative strategy to replace either defective or dysfunctional adult β-globin production.

An alternative approach for the β-thalassemia syndromes may be the targeted reduction of α-globin expression to restore the quantitative balance in globin chain production, thereby reducing pathologic effects of excess α-globin aggregates [Mettananda et al., 2015]. This rationale is supported by the observation that the number of functional α-globin genes has direct clinical consequences as exemplified by the beneficial effects of co-inherited α-thalassemia mutations in β-thalassemia patients [Kan and Nathan, 1970; Thein, 2008] and a more severe clinical phenotype upon increased copy number of the α-globin gene cluster [Premawardhena et al., 2005; Sollaino et al., 2009]. As anticipated, in proof-of-principle experiments, RNA interference strategies that reduce the amount of α-globin mRNAs have resulted in phenotypic improvement in murine models of β-thalassemia [Voon et al., 2008; Xie et al., 2011; Xie et al., 2007].

The efficacy of gene therapy approaches has also been evaluated in murine models for other congenital anemias, including Diamond-Blackfan anemia (DBA). DBA is characterized by an absence of erythroid precursors in the bone marrow and various developmental anomalies [Vlachos et al., 2008]. In 50–60% of cases, DBA is caused by heterozygous loss of function mutations in one of sixteen ribosomal proteins, with ribosomal protein S19 (RPS19) being mutated in ~25% of affected individuals [Danilova and Gazda, 2015]. Using lentiviral vectors, Jaako and colleagues restored ribosomal protein S19 (Rps19) expression, rescuing the anemia in an Rps19 RNA interference knockdown mouse model of DBA [Jaako et al., 2014]. However, some ribosomal protein gene lesions have only been identified in a fraction of patients, and while the development of appropriate gene therapy vectors is possible, their testing of efficacy and safety is laborious and may not be suitable for a very limited number of patients. Recently, the altered translation of GATA1 mRNA has been identified as a common defect downstream of the knockdown of various ribosomal protein genes mutated in DBA [Ludwig et al., 2014]. Although it has also been suggested that the erythroid defect may occur as a result of excessive p53 activation [Ball, 2011; Boultwood et al., 2012; Ellis, 2014], the reduced production of the erythroid-essential transcription factor GATA1 appears to play a fundamental role with respect to the specificity of the observed defect, and rarely, GATA1 mutations are sufficient to cause DBA [Sankaran et al., 2012; Weiss et al., 2012]. Accordingly, lentiviral mediated expression of GATA1 in patient cells with ribosomal protein haploinsufficiency partially restored erythroid differentiation [Ludwig et al., 2014]. Thus, gene therapy approaches aimed at increasing the expression of GATA1 may provide a promising alternative and would be amenable to a large proportion of patients with DBA, independent of their genetic lesion. Erythroid-specific expression of GATA1 is particularly important as ubiquitous expression can promote the differentiation of HSCs with concomitant loss of self-renewal activity [Gentner et al., 2010; Iwasaki et al., 2003]. Gene expression within the HSC compartment may be prevented by the incorporation of microRNA-126 target sequences into GATA1 gene expression vectors. MicroRNAs present a class of short noncoding RNAs that bind to mRNAs and control their gene expression at the post-transcriptional level. MicroRNA-126 is highly expressed in HSCs but quickly downregulated during the early stages of hematopoietic differentiation and this strategy has successfully suppressed ectopic gene expression in HSCs, while maintaining expression in differentiated cells [Gentner et al., 2010]. Furthermore, erythroid-specific regulatory elements for GATA1 expression have already been identified in mice and provide a framework for lineage-restricted expression of GATA1 in hematopoietic cells [Shimizu et al., 2013; Takai et al., 2013]. In the future, unbiased screening approaches, such as massively parallel reporter assays, may have value for identifying regulatory elements with desired expression properties to allow for regulated delivery of such transgenes [Ulirsch et al., 2016].

Fanconi anemia is a rare bone marrow failure disorder (affecting all blood lineages) with patients developing varying degrees of hematologic disease, including anemia, other cytopenias, myelodysplastic syndrome, and acute myeloid leukemia [Kee and D'Andrea, 2012]. Mutations in twenty-one different genes have been identified, with FANCA being the most commonly mutated (in up to 60% of patients) [Bluteau et al., 2016; Rio et al., 2014]. All gene products appear to function in a common pathway, the Fanconi anemia pathway, which is primarily involved in double stranded DNA repair by homologous recombination. Consequently, patients have an increasing predisposition to cancer with age. Lentiviral-mediated gene therapy in Fanca deficient mice efficiently corrected the phenotype without any sign of toxicity [Molina-Estevez et al., 2015] and human gene therapy trials are currently underway [EuroFancolen, 2015; Tolar et al., 2012].

Numerous metabolic defects in erythroid cells have been described and gene replacement strategies have proven to be an effective means of permanent correction in models of these disorders. Glucose-6-phosphate dehydrogenase (G6PD) is a housekeeping gene expressed in all cells and deficiency of its activity is a highly prevalent condition in humans [Luzzatto et al., 2016]. G6PD is part of the pentose phosphate pathway and plays a major role in NADPH production and resistance to oxidative stress in erythrocytes, which lack mitochondria to regenerate antioxidants. Most mutations in G6PD are mild and only become clinically apparent upon challenge of the patient with oxidant drugs or foods. However, some mutations cause severe instability or reduction in gene expression, which therefore lead to lifelong chronic hemolytic anemia [Mason et al., 1995]. Early proof-of-principle experiments have shown stable retroviral transduction of murine and human HSCs with human G6PD in a mouse transplant model and phenotypic correction in G6PD deficient embryonic stem cells [Rovira et al., 2000]. Pyruvate kinase (PK) deficiency is a chronic hemolytic anemia due to mutations in the PKLR gene. PKLR catalyzes the last ATP-generating reaction in the glycolytic pathway [Zanella et al., 2005]. Mature erythrocytes only express the R-type-specific isoform, so any loss of PKLR activity results in reduced ATP formation and increased metabolic and mechanical vulnerability. The resulting hemolytic anemia may range from mild to life threatening in severe cases. Successful gene replacement therapy has recently been demonstrated in a mouse model of PK deficiency [Garcia-Gomez et al., 2016]. Using a lentiviral vector expressing the human PKLR cDNA, Garcia-Gomez and colleagues transduced murine PK-deficient HSCs and showed functional correction of the glycolytic pathway and normalization of the hematologic phenotype.

Congenital erythroid porphyria results from the deficiency in uroporphyrinogen III synthase (UROS), the fourth enzyme of the heme biosynthetic pathway [Di Pierro et al., 2016]. The enzyme defect results in the accumulation of toxic porphyrin metabolites, which cause cutaneous photosensitivity, photo-mutilation, and chronic hemolytic anemia. Mice carrying homozygous mutations in Uros closely mimic the symptoms observed in patients, and lentiviral-mediated transfer of UROS cDNA into mouse HSCs resulted in a complete and long-term enzymatic, metabolic, and phenotypic correction of the disease, favored by a survival advantage of corrected red blood cells [Robert-Richard et al., 2008]. Together, these reports underscore the utility and feasibility of gene therapy approaches in the treatment of various congenital forms of anemia.

Genome editing tools and applications in congenital anemias

The correction of a gene at its endogenous locus presents an appealing solution for the treatment of blood disorders and there have been considerable advances in the field of genome editing over the past years. However, precise editing while completely avoiding the creation of unintended mutations, such as insertions or deletions (indels), remains a major challenge. Current genome editing approaches rely on the expression of a nuclease to induce specific double-stranded DNA breaks (DSBs) that trigger homologous recombination and/or non-homologous end-joining (NHEJ) that leads to the formation of indels [Kim and Kim, 2014] (Fig 3). The process of homologous recombination may be directed by the introduction of a donor template to insert a desired sequence leading to homology-directed repair (HDR) at the endogenous locus, or at a so-called safe harbor locus in a specific region of the genome [Cox et al., 2015; Genovese et al., 2014; Hsu et al., 2014; Kim and Kim, 2014; Sadelain et al., 2012; Sankaran and Weiss, 2015]. While NHEJ occurs throughout the cell cycle, the activity of the HDR machinery is primarily confined to the S and G2 phases [Mao et al., 2008; Orthwein et al., 2015; Saleh-Gohari and Helleday, 2004]. Considering that the most primitive repopulating HSCs with multi-lineage and long-term engraftment potential are largely quiescent, this presents a major hurdle for achieving therapeutically beneficial HDR-mediated editing efficiently, as these cells appear to preferentially repair DSBs by the error-prone NHEJ pathway [Genovese et al., 2014; Mohrin et al., 2010]. Some marginal improvements in homology-directed genome editing can be achieved through the use of rationally designed asymmetric donor DNA templates [Richardson et al., 2016], and modulation of the balance between the activation of HDR vs. NHEJ may also slightly increase precise editing efficiency [Chu et al., 2015; Maruyama et al., 2015; Pinder et al., 2015; Yu et al., 2015], but it remains unclear whether such manipulation of HSPCs may affect engraftment or repopulating capability. Nevertheless, NHEJ can be utilized to induce deletions or other disruptions in coding or regulatory regions, although nuclease off-targets effects must still be considered. Delivery of the nuclease and donor template to HSPCs can be enabled by the use of integrase-defective non-integrating lentiviral vectors (IDLV) or these products can be introduced via transient delivery of plasmids or nucleoprotein complexes [Cai et al., 2016; Giani et al., 2016; Mandal et al., 2014; Nightingale et al., 2006; Schumann et al., 2015; Shaw and Cornetta, 2014]. Consequently, the effect of altered gene expression and insertional oncogenesis due to the integration of foreign DNA would be minimized, as this remains a major concern for classical gene therapy viral-based approaches even with the use of safer SIN vectors [Aiuti et al., 2013; Hacein-Bey-Abina et al., 2008].

Figure 3.

Nuclease-induced genome-editing. (A) Induction of a double strand break (DSB) by a nuclease either leads to non-homologous end joining (NHEJ) with indel formation (small random deletions or small random insertions), or results in homology-directed repair (HDR) after introduction of an either single-stranded oligodeoxynucleotide (ssODN) or a double-stranded donor DNA repair template containing the corrected sequence and flanking sequences that are homologous to the target locus. Alternatively, HDR may be utilized for targeted insertion of a gene, either at the endogenous locus or at a so-called safe harbor locus. (B) Induction of two DSBs by two nucleases on the same chromosome can result in inversion or large deletion of the flanking region. (C) Induction of two DSBs by two nucleases on different chromosomes may result in translocations. Figure has been modified from Kim and Kim [Kim and Kim, 2014].

Several classes of targeted nucleases are available in the genome editing toolkit and have recently been reviewed in detail [Cox et al., 2015; Gaj et al., 2013; Gupta and Musunuru, 2014; Kim and Kim, 2014]. These include meganucleases, zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and most recently the CRISPR/Cas systems (Table 1). Meganucleases are sequence-specific endonucleases that have relatively long target sites, 14–40 base pairs (bp), limiting their application due to the rare presence of target sequences at the majority of human genes [Cox et al., 2015; Hafez and Hausner, 2012; Paques and Duchateau, 2007]. Nonetheless, despite challenges, it is possible to engineer these meganucleases to target diverse sequences within the genome [Silva et al., 2011; Smith et al., 2006]. There are only limited examples of the use of meganucleases for gene correction related to congenital anemias or other blood disorders. For example, one study reported the correction of DCLRE1C deficiency in murine HSCs, albeit in a rather inefficient manner [Riviere et al., 2014].

Table 1.

Key features of the major classes of programmable nucleases

	Meganucleases	ZFNs	TALENs	CRISPR/Cas
Nuclease	Homing endonucleases	FokI	FokI	Cas9
Type of recognition	Protein-DNA	Protein-DNA	Protein-DNA	RNA-DNA
Recognition site length	14–40 bp	18–36 bp	30–40 bp	~20 bp
Engineering complexity	high	high	moderate	low
Off-target effects	low	high	low	variable
Size	~1 kb	~1 kb/ monomer	~3 kb/ monomer	~3.3 kb – ~4.2 kb

ZFNs are site-specific endonucleases composed of a DNA-binding domain consisting of three to six tandem arrays of Cys₂His₂ zinc fingers, each able to specifically bind 3 bp of DNA, and a catalytic DNA-cleavage domain, typically derived from the endonuclease FokI [Scott, 2005]. Multiple zinc fingers can be incorporated to recognize a DNA target sequence up to 36 bp in length to increase specificity. DNA cleavage is then induced by hetero-dimerization of the nuclease domains of two independent ZFNs, each binding to adjacent sequences [Gaj et al., 2013; Perez-Pinera et al., 2012]. To avoid homo-dimerization of individual ZFNs resulting in potential non-specific cleavage, FokI variants have been designed that efficiently cleave DNA only when paired as a heterodimer with a second ZFN leading to improved specificity [Doyon et al., 2011; Miller et al., 2007; Szczepek et al., 2007]. Design, construction, and validation present a challenging process that prohibits the wider application of ZFNs [Ramirez et al., 2008]. Nevertheless, ZFNs have been successfully applied to target the CCR5 locus in HSPCs [Holt et al., 2010; Wang et al., 2015] and to correct genetic lesions in patient-specific induced pluripotent stem cells (iPSCs) with SCD and Fanconi anemia mutations [Rio et al., 2014; Sebastiano et al., 2011]. Correction in primary HSPCs from patients with SCD has also recently been reported [Hoban et al., 2015], although this process is rare and inefficient. Therefore, the low rate of successful gene editing presents a significant bottleneck for therapeutic purposes. For example, in SCD samples, the ZFN-mediated correction of HbS to HbA did not exceed 10%. This is consistent with a study that demonstrated site-specific gene correction in human HSPCs derived from patients with severe combined immune deficiency and showed particularly low efficiency of precise editing in quiescent HSCs [Genovese et al., 2014]. This presents a major obstacle for clinical applications, as HSPCs are the primary target for editing strategies.

TALENs comprise a customizable DNA-binding domain that allows for targeting any sequence of interest and a restriction endonuclease, FokI, to introduce DSBs [Cox et al., 2015; Joung and Sander, 2013; Kim and Kim, 2014]. The DNA-binding domain consists of highly conserved TALE repeats arranged in a tandem array. Each repeat is composed of 33–35 amino acids binding to a single base of DNA that is recognized by two hypervariable amino acid residues, usually at positions 12 and 13 of the repeat [Joung and Sander, 2013; Perez-Pinera et al., 2012]. Multiple TALE repeats can be designed in any order of interest and their targeted DNA sequences are typically 15–20 bp in length. Similar to ZFNs, the target DNA is cleaved after dimerization of the nuclease domains of two individual TALENs in the 12–21 bp spacer region between their respective recognition sites [Gaj et al., 2013; Joung and Sander, 2013]. Hybrid nucleases known as megaTAL combining a TALE DNA binding domain with an engineered, sequence-specific meganuclease may promote even more efficient cleavage with increased sequence specificity and improved delivery [Boissel et al., 2014]. Several groups have utilized TALENs in erythroid cell lines and in SCD and β-thalassemia PSC lines to mediate site-specific correction or targeted integration of β-globin at its endogenous locus [Sun and Zhao, 2014; Voit et al., 2014; Xu et al., 2015]. TALEN-mediated homologous recombination has also been used for therapeutic editing of pyruvate kinase deficiency in PSCs with subsequent generation of a high number of erythroid cells from these edited cells [Garate et al., 2015].

TALENs and ZFNs have been utilized successfully to target erythroid-specific regulatory elements of BCL11A to induce fetal globin expression. Follow-up of GWAS signals led to the identification of an erythroid-specific enhancer of BCL11A [Bauer et al., 2013]. TALEN directed targeting of this enhancer resulted in disruption of BCL11A expression in a mouse erythroleukemia cell line, with robust induction of embryonic globin expression. A subsequent study reported the deletion of the +55 kb and +58 kb DNase I–hypersensitive sites of the erythroid-specific BCL11A enhancer in primary human HSPCs using TALENs and ZFNs resulting in elevation of γ-globin mRNA and HbF levels [Vierstra et al., 2015]. Similarly, CRISPR/Cas9 mediated ablation of BCL11A expression has been achieved by targeting its erythroid-specific enhancer in human hematopoietic cells, including primary HSPCs, which robustly induced HbF expression [Canver et al., 2015].

The CRISPR/Cas endonuclease systems rely on the guidance of a small RNA to their target DNA site, which distinguishes it from meganucleases, ZFNs, and TALENs, all of which rely on protein-DNA interactions (Table 1) [Cox et al., 2015; Doudna and Charpentier, 2014]. In the CRISPR/Cas systems a 20 bp single guide RNA (sgRNA) complementary to the target DNA sequence and the presence of a downstream protospacer adjacent motif (PAM) sequence, NGG, are sufficient to allow cleavage by the Cas9 nuclease. This feature of RNA-guided genome targeting allows CRISPR/Cas systems to be incredibly versatile in targeting numerous regions of the genome without the need to engineer DNA binding proteins [Cox et al., 2015; Doudna and Charpentier, 2014; Hsu et al., 2014; Sander and Joung, 2014]. Modification of this system by using Cas9 nickase mutants to introduce single strand DNA breaks (nicks), in combination with using paired sgRNAs to introduce multiple targeted breaks in the desired locus, results in particularly high specificity and significant reduction of off-target effects [Ran et al., 2013b]. Over the past few years, multiple CRISPR/Cas systems with various characteristics have been described and will inevitably further advance genome editing applications [Cong et al., 2013; Ran et al., 2015; Zetsche et al., 2015].

An initial effort to assess the editing efficiency of the CRISPR/Cas9 system in HSPCs reported highly efficient CCR5 ablation with the use of a single guide RNA, which could be improved further with a dual sgRNA approach, achieving biallelic gene deletion rates of ~27% with minimal detectable off-target mutations [Mandal et al., 2014]. The edited HSPCs retained multi-lineage potential in vitro and in vivo upon transplantation in immunodeficient mice. Similarly, CRISPR/Cas9 has also been successfully applied for gene correction by HDR in congenital anemias, including β-thalassemia and SCD patient-derived primary hematopoietic cells, iPSCs and in fibroblasts from patient with Fanconi anemia [Hoban et al., 2016a; Huang et al., 2015; Niu et al., 2016; Osborn et al., 2015; Xie et al., 2014; Xu et al., 2015]. Recently, CRISPR/Cas9-mediated genome editing of a promoter region in the γ-globin gene recapitulated a naturally occurring mutation in people with hereditary persistence of fetal hemoglobin and lead to the production of RBCs with modest elevation of HbF [Traxler et al., 2016]. However, the authors of this work did not comprehensively ascertain the range of indels that may form, particularly if larger deletions occurred, and additionally how effective this strategy is in engraftable HSCs was not tested.

These advances emphasize the promise of genome editing approaches utilizing autologous cells to overcome limitations in allogeneic transplantation for the treatment of congenital anemias. Currently, the efficiency of precise genome editing remains the major bottleneck for site-specific gene correction in HSPC. Furthermore, DNA breaks themselves, such as those created by the vast majority of genome editing approaches, may compromise HSC function and have prompted the development of new nucleotide editing tools that may allow more precise genome modifications without the creation of DNA breaks [Komor et al., 2016]. It is important to note that in many instances, complete correction of all HSPCs will not be required. Even at a smaller frequency, gene-corrected cells would exhibit a selective survival advantage to fill the previously defective erythropoietic niche and increase RBC lifespan to allow for a clinically relevant benefit [Sankaran and Weiss, 2015]. However, at this time HSPCs cannot be expanded indefinitely in vitro while maintaining long-term repopulation and multi-lineage capabilities [Boitano et al., 2010; Fares et al., 2014]. Analogous to classical gene therapy approaches, this significantly hampers the identification and enrichment of site-specific corrected cells and prohibits reliable verification of editing and screening of adverse off-target effects. These limitations of HSPCs prompted the exploration of other potential target cell types for genetic manipulation and correction.

The use of pluripotent stem cells for cell therapies

The ability of pluripotent stem cells (PSCs) to differentiate into cell types derived from any of the three germ layers has made them an attractive cellular target for the purposes of regenerative medicine. Recent advances in the derivation of PSCs, including human embryonic stem cells and induced PSCs (iPSCs), have opened up many new possibilities for the study of human disease and potential cellular therapies [Chen et al., 2014; Takahashi et al., 2007]. Their capacity to expand and maintain pluripotency in vitro offers a potentially unlimited supply for screening and selecting gene-edited clones and has successfully been demonstrated for various congenital anemias, including SCD, β-thalassemia, and PK deficiency [Garate et al., 2015; Sebastiano et al., 2011; Xu et al., 2015; Zou et al., 2011]. Furthermore, mature erythroid cells have been successfully derived from PSCs and may provide a means to overcome deficits in the blood supply for transfusion purposes [Giani et al., 2016]. Despite the potential of PSCs, their application for curative cellular therapies remains restricted by the insufficient generation of fully functional HSCs from PSCs (Fig 1B) [Daniel et al., 2016; Kim and Sankaran, 2016; Vo and Daley, 2015]. The generation of HSCs by mimicking the natural differentiation pathway has long been considered the optimal route. However, although hematopoietic differentiation of PSCs has been found to mirror aspects of in vivo embryonic development, current efforts have been unable to produce the third hematopoietic wave of ontogeny that gives rise to HSCs [Nandakumar et al., 2016; Yoder, 2014]. The first hematopoietic wave, commonly known as the primitive wave, arises in the yolk sac at embryonic day 7.5 (E7.5) of murine development and transiently produces primitive erythroblasts and primitive macrophages and megakaryocytes. A second wave of yolk sac hematopoiesis is characterized by the emergence of erythro-myeloid progenitors (EMPs) at approximately E8.25 that also colonize the fetal liver. EMPs arise from vascular endothelial cells that are present in blood island capillaries of the yolk sac and give rise to erythro-myeloid and lymphocyte progenitor cells. The second wave is considered “definitive” because the produced blood cells display adult-type features and functions, but this wave does not produce HSCs or other progenitors capable of long-term engraftment. The third wave begins at E10.5 and is characterized by the emergence of HSCs from a unique population of endothelial cells, termed the hemogenic endothelium (HE) that can be found in dorsal aorta of the aorta-gonad-mesonephros region and within the arteries found in the umbilical, vitelline, cranial, yolk sac, and placental regions [Yoder, 2014]. These HSCs migrate to the fetal liver before ultimately seeding the bone marrow and giving rise to all blood lineages; they are capable of long-term engraftment after transplantation. To date, protocols for PSC differentiation have only been able to produce the first two waves of hematopoiesis successfully and do not fully recapitulate certain blood disorders, such as the hemoglobinopathies, since the adult β-globin gene is generally poorly expressed [Yang et al., 2014].

Initial experiments successfully generated hematopoietic progenitors by directed differentiation from embryoid bodies, three-dimensional aggregates of PSCs, but their multi-lineage potential and engraftment capacities were limited as these cells primarily exhibited characteristics of the first two hematopoietic waves, as we note above [Kaufman et al., 2001; Keller et al., 1993; Vodyanik et al., 2005]. Further research led to significant progress in the identification of guiding factors and redefined PSC differentiation protocols through HE to derive hematopoietic cells with multi-lineage-potential, but yet have failed to generate genuine HSCs [Boiers et al., 2013; Ditadi et al., 2015; Dzierzak and Speck, 2008; Gori et al., 2015; Kennedy et al., 2012; Ramos-Mejia et al., 2010; Sturgeon et al., 2014].

Recently, the generation of HSCs from human PSCs has been reported through the formation of teratomas in immunodeficient mice [Amabile et al., 2013; Suzuki et al., 2013]. Teratoma-derived CD34⁺ cells were capable of reconstitution and multi-lineage engraftment upon xenotransplantation. These studies provide a proof-of-principle that PSC-derived human hematopoietic cells with repopulating and reconstituting potential can possibly be obtained, although the described tumor-forming approaches are not suitable for potential clinical applications and certainly cannot be efficiently scaled to generate HSCs reliably from PSCs. The identification of intrinsic and extrinsic factors necessary for the production of HSCs presents a significant challenge in the use of PSCs for cellular therapeutic applications in hematopoietic disorders.

Direct conversion of cellular identities by transcription factor mediated reprogramming without going trough a pluripotent stage has been demonstrated for various somatic cell types and presents an alternative source for the generation of gene-corrected cells for therapeutic purposes [Vo and Daley, 2015] (Fig 1B). Several studies have shown the instructive roles of lineage-specific transcription factors like GATA1, CEBPA, CEBPB, and SPI1 in the direct conversion between hematopoietic cell types [Iwasaki et al., 2003; Laiosa et al., 2006; Xie et al., 2004] and from non-hematopoietic to hematopoietic cell types [Capellera-Garcia et al., 2016]. These approaches may be improved further to increase cellular yield when combined with additional methods of manipulating hematopoietic cytokine signaling [Giani et al., 2016]. The generation of hematopoietic progenitor cells from human and murine fibroblasts and endothelial cells has been demonstrated, but to date show limited self-renewal capacity, engraftment potential, or multi-lineage differentiation potential [Batta et al., 2014; Sandler et al., 2014; Szabo et al., 2010]. However, a similar reprogramming approach aided in the identification of a murine hemogenic precursor cell in vivo capable of maturation into potential bona fide HSCs [Pereira et al., 2016; Pereira et al., 2013]. The first study to successfully generate murine HSCs screened the conversion capability of 36 hematopoietic factors in mature hematopoietic cells to generate engraftable hematopoietic cells [Riddell et al., 2014]. The authors identified a cocktail of 8 transcription factors (Run1t1, Hlf, Lmo2, Prdm5, Pbx1, Zfp37, Mycn and Meis1) to endow lymphoid and myeloid cells with the capacity for multi-lineage engraftment and reconstitute hematopoiesis in serial transplantation. However, it remains unclear to what extent successful reprogramming and maintenance of induced HSCs relied on in vivo cues after the transplantation of transduced cells, currently limiting this approach for potential clinical translation. Whether the same factors can similarly reprogram human hematopoietic cells also remains to be investigated.

Direct conversions from PSCs have also been attempted using a variety of transcription factors, and these manipulations have generated HE and hematopoietic progenitor cell populations, but not yet functional HSCs [Doulatov et al., 2013; Elcheva et al., 2014; Elcheva et al., 2015; Ran et al., 2013a]. While significant progress has been made in the use of PSCs, further studies will be required to optimize transcription factor combinations and to identify additional (extrinsic) factors to refine the process of directed differentiation and direct conversion from pluripotent and readily obtainable somatic cells to generate autologous HSCs. The comprehensive assessment of their functionality, measured by their ability to fully repopulate a recipient upon transplantation and direct molecular comparison of their transcriptional and epigenomic profiles to native HSCs, present fundamental hurdles before their clinical use can be considered. However, if such cells could be obtained, and given that PSCs can readily be manipulated with genome editing tools to enable precise modifications and clonally selected [Giani et al., 2016], this would create an incredibly valuable source of autologous cell therapies to treat congenital forms of anemia.

Concluding remarks

Congenital forms of anemia are a genetically heterogeneous group of disorders that result in significant morbidity and mortality worldwide. At the current time, allogeneic transplantation of HSPCs from a healthy donor is the only cure available. Numerous advances in the fields of gene therapy and genome editing have paved the way to genetically correct patient cells and enable an autologous form of cell therapy with numerous pre-clinical and clinical studies in progress. Further refinement in genome editing tools with improved specificity paired with more elaborate cell culture protocols for HSPCs and PSCs may ultimately enable new curative options for patients with a variety of blood disorders. Despite these promising developments, many congenital forms of anemia are particularly common in resource-poor countries, and most cell and gene therapy approaches may therefore benefit only a minority of patients worldwide. Unless technical advances allow these therapies to be more readily applied, there remains a substantial need to pursue other strategies, such as the development of small molecules. Nonetheless, the insight gained from clinical studies of gene and cell therapy may provide insight into the development of other more readily distributable therapeutics. Thus, major challenges remain and will require the combined efforts of basic scientists and clinical investigators to safely advance these strategies to the clinic.

Biographies

Leif S. Ludwig completed his medical studies at the Charité Universitätsmedizin Berlin, Germany and graduated with his Ph.D. in Biochemistry at the Freie Universität Berlin, Germany. He is currently working as a postdoctoral researcher studying hematology and genetics at Boston Children’s Hospital and the Broad Institute.

Rajiv K. Khajuria is an M.D. Ph.D. student at the Charité Universitätsmedizin Berlin, Germany. He is currently conducting his thesis research studying congenital blood disorders at Boston Children’s Hospital and the Broad Institute.

Vijay G. Sankaran is an Assistant Professor of Pediatrics at Boston Children’s Hospital and Harvard Medical School, an Attending Physician in Hematology/ Oncology, and an Associate Member of the Broad Institute. His laboratory utilizes human genetics to understand normal blood production and how this process is perturbed in disease states.