Autophagy in the pathogenesis of myelodysplastic syndrome and acute myeloid leukemia

Abstract

Autophagy is a conserved cellular pathway responsible for the sequestration of spent organelles and protein aggregates from the cytoplasm and their delivery into lysosomes for degradation. Autophagy plays an important role in adaptation to starvation, in cell survival, immunity, development and cancer. Recent evidence in mice suggests that autophagic defects in hematopoietic stem cells (HSCs) may be implicated in leukemia. Indeed, mice lacking Atg7 in HSCs develop an atypical myeloproliferation resembling human myelodysplastic syndrome (MDS) progressing to acute myeloid leukemia (AML). Studies suggest that accumulation of damaged mitochondria and reactive oxygen species result in cell death of the majority of progenitor cells and, possibly, concomitant transformation of some surviving ones. Interestingly, bone marrow cells from MDS patients are characterized by mitochondrial abnormalities and increased cell death. A role for autophagy in the transformation to cancer has been proposed in other cancer types. This review focuses on autophagy in human MDS development and progression to AML within the context of the role of mitochondria, apoptosis and reactive oxygen species (ROS) in its pathogenesis.

Autophagy

Autophagy, a lysosomal/vacuolar-mediated catabolic pathway for degradation of cytoplasmic constituents, has approximately 30 identified direct molecular players.1 Autophagy takes place constitutively at basal levels which can be induced to higher levels in response to multiple physiological conditions, including starvation, hormonal imbalance and oxidative stress.

Autophagy is initiated by the Atg1–Atg13 protein complex, which, in mammals, is activated by the absence of signaling from the nutrient-sensing kinase mammalian target of rapamycin (mTOR). Next, the class III phosphoinositide-3-kinases (PI3K) (hVps34 and hVps35)—Beclin 1 (mammalian homolog of Atg6) complexes are key to the nucleation step, which results in the formation of an isolation membrane. Elongation of the isolation membrane is then mediated by two ubiquitin-like conjugation systems; one where Atg7 (E1-like) and Atg10 (an E2-like protein) act to conjugate Atg5 to Atg12, and another in which the Atg5–Atg12 conjugate (which functions as an E3-like protein) acts in concert with Atg7 and Atg3 (an E2-like protein) to conjugate Atg8 to phosphatidylethanolamine in the membrane of the growing autophagosome. Finally, autophagy culminates in the fusion of the autophagosome with lysosomal compartments and the degradation of autophagosomal contents by lysosomal hydrolases.2

Autophagic Processes in Cancer

Autophagy has been shown to play several important roles in cancer.3 Indeed, multiple autophagy genes, including Beclin 1 and UVRAG, have now been characterized as tumor suppressor genes.4 Autophagy is implicated in cancer due to its extensive crosstalk with apoptosis, a programmed cell death pathway that is often impaired during tumorigenesis. Autophagy also has a protective role through controlling oxidative stress as well as the build-up of potentially DNA damaging waste.3,5,6

Myelodysplastic Syndrome and Acute Myeloid Leukemia

MDS comprises a group of anemic disorders of uncertain etiology characterized by abnormal cell morphology in the bone marrow (BM) and peripheral blood cytopenias. Increased caspase-dependent apoptosis, reactive oxygen species (ROS) and mitochondrial damage are observed in BM cells from MDS patients.7 The classification of MDS subtypes and their underlying mechanisms has remained contentious;8 several systems for classifying cases of this disorder have been devised based on morphology and/or prognostic parameters. The World Health Organization (WHO) recognizes seven main MDS subtypes, primarly based on refractory anemias/cytopenias: (1) refractory cytopenias with unilineage dysplasia (RCUD), (2) refractory anemia with ring sideroblasts (RARS), (3) refractory cytopenia with multilineage dysplasia (RCMD), (4 and 5) refractory anemia with excess blasts (RAEB-1 and RAEB-2), (6) MDS-unclassified (MDS-U) and (7) MDS associated with isolated del (5q).9 Although classification issues make prevalence hard to determine8 recent studies have estimated MDS to be more common than previously thought, with a prevalence of 2/1,000 above the age of 65 in the US and Canada.10 MDS has an overall transformation rate of 30% to acute myeloid leukemia (AML).

AML is thought to occur due to the transformation of an immature hematopoietic stem or progenitor cell (HSPC), leading to myeloid leukemia. The dysregulated production and proliferation of myeloid cells in the BM interferes with normal hematopoiesis resulting in a reduction in red blood cell (RBC), platelet and leukocyte numbers. The WHO classifies AML into: (1) AML with characteristic genetic abnormalities (translocations and inversions), (2) AML with multilineage dysplasia (progressed from MDS), (3) therapy-related AML and (4) AML not otherwise categorized. AML is the most common myeloid leukemia, with a prevalence of 3–8 cases per 100,000 which rises to 17–19 per 100,000 in those 65 and older. It is often associated with common mutations and chromosomal abnormalities; as with many cancers a multiple genetic-hit model for its initiation has been proposed.11

Here, we will review the accumulating evidence for an important role of autophagy in MDS and AML. Recent evidence arising from mouse models deficient for autophagy in the hematopoietic system will be reviewed, followed by an assessment of the cellular hallmarks of the human disease with possible links to autophagy. Finally, we will present our in-silico analysis of existing studies on genomic alterations in autophagy genes found in AML cells.

Autophagy, Mitophagy and Stem Cell Maintenance

When autophagy is induced by nutrient deprivation, autophagic substrate selection is thought to be mostly random; on the other hand, in nutrient-rich conditions, cargo-selective autophagy takes place to remove superfluous, damaged organelles and protein aggregates. The selective removal of mitochondria by autophagy (mitophagy) has been particularly well studied.12,13

Some of the pathways leading to mitophagy have now been elucidated.12 In yeast, several of the main autophagy players, such as ATG1, ATG5, ATG6, ATG8, ATG9, ATG12 and ATG13, have been found important for mitophagy.14 Based on knockout models of several core autophagy genes, there is good evidence that mitophagy also utilizes some of the classical autophagic proteins in mammalian cells.

Within three years of the term “autophagy” being introduced by Christian de Duve in 1963,15 autophagosomes were described engulfing mitochondria in human RBCs.16 Yet, it was only recently that, thanks to studies of autophagy gene knockout models, mitophagy was confirmed to be essential for erythroid development.14,17–19 In mammals, erythrocytes mature from HSC-derived early progenitors, primarily in the BM, starting as erythroblasts before losing their nuclei and becoming reticulocytes. Newly formed reticulocytes develop further into erythrocytes in peripheral circulation, where they complete protein synthesis, acquire a biconcave shape and get rid of all their remaining organelles, primarily mitochondria.20 Mitophagy is essential for this developmentally regulated mitochondrial degradation. It has also been found to play an important role in T lymphocytes, in which the regulation of mitochondrial load is essential for their development and survival.21–23

Mice deficient for Atg7 in the hematopoietic system develop severe anemia, lymphopenia and atypical myeloproliferation.13,23 Importantly, these symptoms are all hallmarks of MDS and/or AML. These mice also displayed an impaired self-renewal capacity of their HSCs, as assessed by serial replating of colony-forming unit assays and in vivo reconstitution assays, leading to the conclusion that Atg7 and autophagy are vital to HSC maintenance. Moreover, mice lacking Atg7 in the hematopoietic system, despite the near complete loss of true HSCs, were found to have an expanded Lin⁻Sca⁺c-kit⁺ (LSK) compartment, normally containing hematopoietic stem and progenitor cells. The expanded LSK compartment in the absence of Atg7 showed increased mitochondrial content, higher levels of mitochondrial ROS, DNA damage, an enhanced proliferation rate and higher apoptosis levels. In other words, autophagy-deficient hematopoietic progenitor cells show evidence of mitochondrial damage and ROS build-up along with increased rates of proliferation and apoptosis, which are all frequently observed in human MDS patient samples. Mice lacking autophagy in the hematopoietic system died within 12 weeks and post-mortem analysis of their tissues showed myeloid blast infiltrates in multiple organs including the thymus, liver, spleen, intestines, skin, heart, pancreas and kidney. Histopathological observations revealed the presence of over 20% myeloid blasts in the BM. This, combined with significantly increased numbers of myeloid CD11b⁺Gr1⁺ cells, strongly suggests that these mice developed a myeloproliferative/myelodysplastic overlap disease resembling human AML of the acute myelomonocytic subtype. Moreover, transplantation of the entire marrow or the selected expanded Atg7^−/− LSK compartment into immunocompromised mice replicated this atypical myeloproliferation.13

FIP200, part of the Ulk1 (mammalian Atg1 homolog) autophagy initiation complex, was also recently studied in hematopoiesis. It was conditionally deleted by Liu et al. (2010),24 in early hematopoietic cells using a Tie2-Cre system, which lead to perinatal lethality. Similar to the conditional Atg7 knockouts in the hematopoietic system, hematopoietic FIP200 knockout animals displayed severe anemia and depletion of the HSC compartment. Moreover, FIP200-deficient fetal HSCs, like adult Atg7^−/− HSCs, lost the ability to repopulate lethally irradiated hosts and displayed abnormally increased proliferation along with increased mitochondrial mass and ROS levels. Interestingly, the parallels between the two conditional hematopoietic autophagy gene knockouts did not end there. Despite their early lethality, FIP200 null mice also displayed aberrant myeloproliferation, with a 17-fold increase in the frequency of myeloid cells in the blood of knockout embryos. Although the FIP200 phenotype is associated with fetal hematopoiesis rather than adult hematopoiesis as was the case in the Vav-iCre driven Atg7 knockouts, these data support the idea that autophagy/mitophagy is responsible for the two phenotypes described, rather than an autophagy-independent function of either the FIP200 or Atg7 genes, and that removal of autophagy/mitophagy genes results in anemia along with abnormal myeloproliferation.

Mitochondrial Abnormalities in MDS/AML: An Indication of Defective Mitophagy?

The phenotypes in mice with impaired autophagy in the hematopoietic system raise the question of whether autophagy/mitophagy may also contribute to hallmarks of human MDS and its development and progression to AML. If so, one might expect abnormalities in mitochondria and mitochondrial-linked processes to be associated with those diseases. Multiple studies have observed structurally abnormal mitochondria associated with MDS and enlarged mitochondria25 have been described in pro-erythroblasts, basophilic erythroblasts and, less commonly, in more mature erythroblasts26 of MDS patients. As mitochondria are the main regulators of the caspase-dependent apoptosis pathway, it is not surprising that, in addition to structurally abnormal mitochondria, increased apoptosis is commonly observed in MDS.25 The anemia that is part of the MDS diagnosis is often described as the result of an imbalance between proliferation and apoptosis of HSCs. Consistent with this, murine Atg7^−/− erythrocytes and hematopoietic progenitors showed increased susceptibility to caspase 3-dependent apoptosis.23 In a study by Lin et al. (2002), the higher than normal proliferation observed in MDS was accompanied by significantly increased apoptosis as measured by caspase-3 activity, especially in the HSC-containing CD34⁺ BM population. This was in contrast to the AML samples tested, where the CD34⁺ population showed significantly lower apoptosis as compared to normal controls. Similarly, Encomopoulou et al. (2010), found caspase-3 activity to be positively correlated with the S-phase of the cell cycle, an indication of proliferation, as well as being correlated with cyclin-dependent kinase (CDK) expression and apoptosis. Thus, MDS is characterized by mitochondrial abnormalities, implying possible defects in mitochondrial homeostasis as a consequence of defective mitophagy.

Another consequence of the accumulation of dysfunctional mitochondria would be increased levels of ROS. Multiple sources in the cell generate ROS, including the plasma membrane, peroxisomes and the cytosol. However, most cellular ROS (around 90%) are produced by mitochondria.29 The primary generators of ROS within mitochondria are the electron transport chain cytochromes that create the largest potential energy flux: complexes I and to a lesser extent III.29,30 ROS have long been known to cause DNA mutations;31 thus, as ROS levels increase with the mitochondrial load in the cell, they could begin to damage genetic material in both the mitochondrial and nuclear genomes. If severe enough, this could trigger apoptosis.7 Indeed, multiple mitochondrial genes involved in the electron transport chain have been characterized as tumor suppressors which follow a two-hit model of deletion32 highlighting the tumor-promoting potential of damaged mitochondria. While some studies have found increased incidence of mutations in electron transport chain proteins, such as cytochrome c oxidase I and II,33 and other functionally relevant proteins in MDS patients34 other studies have found no such increased mitochondrial genome instability.35 Therefore, the role that mitochondrial genome mutations play in MDS remains controversial. That being said, recent evidence of clonal expansion of cells with mtDNA mutations in MDS patients supports the idea of ROS/age-related damage to an erythroid precursor.36

Mitochondrial iron accumulation, possibly due to ineffective hematopoiesis, is common in MDS cases,36,37 particularly in the refractory anemia with ring sideroblast (RARS) form. Importantly, excess cellular iron is hypothesized to catalyze ROS forming reactions.38 This would mean that the already potentially pathological accumulation of ROS could be further exacerbated by the presence of excess iron in MDS patients.

Compounding this effect, evidence suggests that high levels of ROS can inhibit the activity of certain phosphatases, including the tumor suppressor PTEN,39,40 thus it is possible that ROS can have transformative influence independent of genetic damage. Finally, a recently proposed model highlights the role of oxidative stress in surrounding stromal support tissue as a mechanism of genetic instability and autophagic induction in tumors.41 This also marks the autophagy-triggering role of ROS, as a possible response aimed at removing a major potential source: aging/damaged mitochondria.

Mitophagy in the Progression of MDS to AML

Recent work has shown that erythroid precursors from low-risk MDS patients (those less likely to progress to AML, as compared to high-risk patients) display increased numbers of mitochondria-laden autophagosomes.42 The authors hypothesize that this increase could indicate intrinsic differentiation defects or could be a response to the higher levels of dysfunctional, iron-saturated mitochondria. However it may also be an indication of a positive feedback cycle consisting of mitophagy defects in hematopoietic stem or progenitor cells leading to improper degradation of targeted mitochondria, promoting increased ROS levels that cause DNA damage. This, combined with increased iron levels (which catalyze ROS formation), contributes to further metabolic defects as relevant genes are altered. This would continue until the abnormal mitochondria release enough cytochrome c for the cell to undergo apoptosis, unless the ROS-induced DNA damage occurs in such a way as to induce progression to AML (Fig. 1). Consistent with this, a study by Boultwood et al. (1996), found increased amounts of mtDNA in AML cells, which we hypothesize may be due to mitophagy defects leading to increased numbers of mitochondria in those cells. So, the high levels of mitochondria-laden autophagosomes observed in low-risk MDS patients may indicate that high levels of functional mitophagy protect these patients from the ROS build-up due to iron-damaged mitochondria until the cells undergo apoptosis from other causes, thus rendering the MDS low risk for transformation. In other words, functional autophagy may be required to prevent AML progression from MDS. The high prevalence of MDS in the aging population may be linked to the fact that autophagy levels decrease with age.44

Similarly, there is indirect evidence that autophagy is involved in human AML. The PI3K/AKT/mTOR signaling pathway is constitutively active in AML patients,45 a phenomenon shown to confer leukemogenic potential to mouse hematopoietic cells.46 Indeed, mTOR inhibitors, such as Rapamycin, have been shown to reduce AML cell clonogenic properties.47 From a functional perspective, a recently published article reported that autophagy is induced by all-trans retinoic acid treatment of acute promyelocytic leukemia (APL) and that autophagy plays an important role in the basal turnover, as well as therapy-induced proteolysis of the PML/RARA oncoprotein.48

Genetic Alterations of Autophagy Genes in Human MDS/AML

Overall, characteristics found in cells from MDS and AML patients suggest a link between these diseases and autophagy/mitophagy; therefore, some abnormalities in autophagic gene expression could be expected to be revealed. However, as yet, studies of expression data from CD34⁺⁴⁹ and CD15⁺⁵⁰ cell populations, representing stem cell populations in MDS and AML myeloid progenitors respectively, have not highlighted autophagy genes as possible targets.

Indeed, a detailed review of expression data from two of the larger recent studies of MDS expression data revealed no significant drop in autophagy gene expression in BM mononuclear cells51 or in HSCs.52 However, as noted by Lee et al. 2006,50 each MDS karyotype shows different expression profiles, and it is likely that any given mechanism for MDS development applies to a specific subset of cases, which would be diluted in a general sampling of all MDS patients. Secondly, it is not yet clear whether autophagy is regulated at a transcriptional level and therefore whether a transcriptional analysis would reflect alterations in autophagic activity. Also, gene expression data would not necessarily reveal functionally deleterious point mutations in the coding regions of mitophagy/autophagy genes.

Similarly, multiple studies have examined genetic changes in myeloid neoplasm progression, including research into the role of mechanisms like acquired uniparental disomy53 and methylation.54 Such research has revealed alterations in a collection of tumor suppressor genes, such as Casitas B-cell lymphoma gene53 and tet oncogene family member 2,55 as prevalent in certain types of myeloproliferative neoplasms. However, these studies have, as yet, not found autophagy genes to be altered in these diseases.

In contrast, our review of recent studies of genetic abnormalities in AML patients has identified several key autophagy network genes, as recently defined by Behrends et al. (2010),56 within the commonly deleted regions (CDRs) on 5q, 7q, 16q and 17p.57–59 Rucker et al. (2006) examined 60 AML patients with complex karyotypes and found losses of genetic material in these regions occurring at frequencies of 77% for 5q, 55% for 17p, 45% for 7q and 32% for 16q. This is significant because, as mentioned above, the chromosomal regions these CDRs encompass can include autophagy genes such as: ATG10 and ATG12 in the 5q CDR, ATG9B, PRKAG2 and KLHDC10 in the 7q CDR, GABARAPL2 and MAP1LC3B (both human ATG8 homologs) in the 16q CDR and GABARAP (another ATG8 homolog) in the 17p CDR.57 Moreover, not only are these autophagy genes within CDRs, but Walter et al. (2009), found significantly reduced expression of both ATG12 and KLHDC10 mRNA in patients with the 5q deletion. Similarly, the autophagy gene GABARAPL1 is located in a small region of genetic loss identified by Akagi et al. (2010). In addition, a recent paper has identified ATG16L2 as hypermethylated in samples from both leukemias and lymphomas.60 Taken together, these data suggest a possible role for autophagy in AML consistent with previous data on frameshift mutations in the autophagy genes ATG2, ATG5, ATG9B and ATG12 in gastric and colorectal cancers.61 As mentioned above, a number of studies have shown functional autophagy to aid in cancer resistance. In particular, Beclin 1,62 (ATG6), which is monoallelically deleted in almost half of human breast, ovarian and prostate cancers,4 and UVRAG63 have both been characterized as tumor suppressors. While deletions of at least one copy of autophagy genes is in keeping with the Knudson hypothesis for tumor suppressor genes, there is evidence to suggest that autophagy may not only overall prevent cancer but may also play a role in cancer cell survival,64 such as mediating cell stress during BCR-Abl-driven leukemogenesis.65 It is possible that autophagy and mitophagy are beneficial to primary oncogene-mediated leukemic transformation and for the survival of established tumors in hostile microenvironments, yet also have a protective role in genome stability in healthy cells, especially longer lived stem cells. Furthermore, heterozygous autophagy gene deletions followed by clonal selection first for deleterious mutations in the remaining allele then against them could explain this dual role,6 but more research is required to assess this hypothesis. We suggest specifically this may be the case in MDS/AML.

In this review, we have built the case for a possible connection between autophagy defects, the death of erythrocyte precursors and the transformation of HSCs into leukemia initiating cells. However, it is important to note that the genetic alterations affecting autophagy genes in human MDS/AML may be bystander rather than driver mutations (i.e., may be a result of the genetic instability of transformed cells rather than a causative agent behind the tumorigenesis), or may at best contribute to the transformation process driven by other mutations. To date, it is the phenotype resulting from the ablation of autophagy in murine HSCs that has provided the strongest evidence for a link between autophagy and myeloid malignancy. There is an obvious need to investigate this connection in the human disease in order to establish a clear link between autophagy and leukemia development. This could shed light on the disease etiology and on how autophagy-modulating drugs may be beneficial in treating these diseases.

Abbreviations

MDS	=	myelodysplastic syndrome
AML	=	acute myeloid leukemia
HSC	=	hematopoietic stem cells
ROS	=	reactive oxygen species

Figures and Tables

Figure 1 A simplified model where an HSC or early hematopoietic precursor develops a fault in autophagy/mitophagy (1) which leads to the build up of damaged mitochondria. These in turn cause an increase in metabolic by-products, including ROS (2) which result in cellular damage including genetic instability and more mitochondrial damage. A loss of quiescence results, along with both increased apoptosis (3) and proliferation (4) in an MDS phenotype. Oncogenic mutations may result in transformation and AML (5).

Acknowledgments

We would like to thank the National Sciences and Engineering Council of Canada and The Nuffield Department of Medicine for providing the funding of A.W.'s PhD. We also gratefully acknowledge the Det Frie Forskningsråd Sundhed og Sygdom (FSS) for a generous postdoctoral fellowship for M.M. The Wellcome Trust, the BBSRC and the NIHR have supported this research.