Drosophila Dicer-2 generates small interfering RNAs (siRNAs) from long double-stranded RNA (dsRNA), whereas Dicer-1 produces microRNAs from premicroRNA. My thesis focuses on the functional characteristics of two Drosophila Dicers that makes them specific for their biological substrates. We found that RNA binding protein partners of Dicers and two small molecules, ATP and phosphate are key in regulating Drosophila Dicers’ specificity. Without any additional factor, recombinant Dicer-2 cleaves pre-miRNA, but its product is shorter than the authentic miRNA. However, the protein R2D2 and inorganic phosphate block pre-miRNA processing by Dicer-2. In contrast, Dicer-1 is inherently capable of processing the substrates of Dicer, long dsRNAs. Yet, partner protein of Dicer-1, Loqs-PB and ATP increase the efficiency of miRNA production from pre-miRNAs by Dicer-1, therefore enhance substrate specificity of Dicer-1. Our data highlight the role of ATP and regulatory dsRNA-binding partner proteins to achieve substrate specificity in Drosophila RNA silencing. Our study also sheds light onto the function of the helicase domain in Drosophila Dicers. Although Dicer-1 doesn’t hydrolyze ATP, ATP enhances miRNA production by increasing Dicer-1’s substrate specificity through lowering its KM. On the other hand, Dicer-2 is a dsRNA-stimulated ATPase that hydrolyzes ATP to ADP, and ATP hydrolysis is required for Dicer-2 to process long dsRNA. Wild-type Dicer-2, but not a mutant defective in ATP hydrolysis, is processive; generating siRNAs faster than it can dissociate from a long dsRNA substrate. We propose that the Dicer-2 helicase domain uses ATP to generate many siRNAs from a single molecule of dsRNA before dissociating from its substrate. Piwi-dependent small RNAs, namely piRNAs, are a third class of small RNAs that are distinct from miRNAs and siRNAs. Their primary function is to repress transposons in the animal germline. piRNAs are Dicer-independent, and require Piwi family proteins for their biogenesis and function. Recently in addition to their presence in animal germlines, the presence and function of piRNA-like RNAs in the somatic tissues have been suggested (Yan et al. 2011; Morazzani et al. 2012; Rajasethupathy et al. 2012). We have investigated whether the piRNA-like reads in our many Drosophila head libraries come from the germline as a contaminant or are soma-specific. Most of the piRNA reads in our published head libraries show high similarity to germline piRNAs. However, piRNA-like reads from manually dissected heads are distinct from germline piRNAs, proving the presence of somatic piRNA-like small RNAs. We are currently asking the question whether these distinct piRNA-like reads in the heads are dependent on the Piwi family proteins, like the germline piRNAs.

In most metazoans, early embryogenesis is controlled by the translational regulation of maternally supplied mRNA. Sequence-specific RNA-binding proteins play an important role in regulating early embryogenesis, yet their specificities and regulatory targets are largely unknown. To understand how these RNA-binding proteins select their targets, my research focused on the C. elegans CCCH-type tandem zinc finger protein POS-1. Embryos lacking maternally supplied POS-1 die prior to gastrulation, and exhibit defects in the specification of pharyngeal, intestinal, and germline precursor cells. To identify the regulatory targets that contribute to the POS-1 mutant phenotype, we set out to determine the sequence specificity of POS-1 in vitro, and then use this information to identify regulatory targets in vivo. Using a candidate-based search, we identified a twelve-nucleotide fragment of the mex-3 3' untranslated region (3' UTR) to which POS-1 binds with high affinity. Using quantitative fluorescent electrophoretic mobility shift assays, I determined the affinity of the RNA-binding domain of POS-1 for a panel of single nucleotide mutations of this sequence, and then defined a consensus binding element based on this dataset. POS-1 recognizes the degenerate element UAU 2-3 RDN 1-3 G, where R is any purine (adenosine or guanine), and D is any base except cytosine. A bioinformatics analysis revealed the presence of this element in approximately 40% of C. elegans 3' UTRs, suggesting that POS-1 is capable of binding to and perhaps regulating many transcripts in vivo. POS-1 binding sites alone are not sufficient to pattern the expression of a reporter, suggesting that other factors may contribute to POS-1 specificity. To address the mechanism of POS-1-mediated translational regulation, I investigated the translational regulation of the C. elegans Notch homolog glp-1. Previous work demonstrated that glp-1 translation is repressed in the early embryo in a POS-1-dependent fashion, though it was not clear if this regulation was direct. The glp-1 3' UTR contains two POS-1 binding sites within five nucleotides of each other, and these sites are within a thirty nucleotide region of the 3' UTR required for proper spatiotemporal translation of glp-1. The POS-1 sites overlap with a negative regulatory element that is recognized by GLD-1, and a positive regulatory element recognized by an unknown factor. Both POS-1 and GLD-1 bind to an RNA containing these sites in vitro, and POS-1 competes with GLD-1 for binding. Both proteins are required for translational repression of a glp-1 3' UTR reporter in embryos. Furthermore, only one of the two POS-1 binding sites is required for repression, and the required site is wholly contained within a previously characterized positive regulatory element. Based on this, we propose that POS-1 does not regulate its targets by recruiting regulatory machinery, but instead by competing with factors that do. Thus, sites of POS-1 regulation are highly context dependent, which may contribute to POS-1 specificity.

The p53 transcription factor responds to various cellular stressors by regulating the expression of numerous target genes involved in cellular processes such as cell cycle arrest, apoptosis, and senescence. As these downstream pathways are harmful to the growth and development of normal cells when prolonged or deregulated, p53 activity needs to be under tight regulatory control. The Mdm2 oncoprotein is the chief negative regulator of p53, and many mouse models have demonstrated that absence of Mdm2 expression leads to constitutive p53 activation in a variety of cell types. While unregulated p53 can be deleterious to cells, functional p53 is essential for tumor suppression, as many human cancers harbor p53 mutations and p53 knockout mice rapidly develop spontaneous tumors. Therefore, the mechanisms that control p53 regulation by Mdm2 are critical to ensure p53 activity in the appropriate cellular context. Many genetically engineered mouse models have been created to analyze p53 and Mdm2 functions and these studies have yielded valuable insights into their physiological roles. This dissertation will describe the generation and characterization of novel mutant Mdm2 mouse models and their use to interrogate the roles of p53-Mdm2 signaling in tissue homeostasis and cell stress responses. Deletion of Mdm2 in epidermal progenitor cells of the skin and hair follicles resulted in progressive hair loss and decreased skin integrity, phenotypes that are characteristic of premature aging. Furthermore, p53 protein levels, p53 target gene expression, and cellular senescence were all upregulated in the skins of these mice, and epidermal stem cell numbers and function were diminished. These results indicate that Mdm2 is necessary to limit p53 activity in adult tissues to ensure normal stem cell function. Additional mouse models used to determine the role of Mdm2 phosphorylation will also be presented. DNA damage triggers an extensive cellular response, including activation of the ATM kinase. ATM activity is necessary for p53 protein stabilization and, therefore, p53 activation, but in vivo evidence suggests that phosphorylation of p53 itself had little effect on p53 stability. ATM was previously shown to phosphorylate MDM2 at serine residue 395 (394 in mice), and we generated knock-in mutant mouse models to study the role of this posttranslational modification in vivo. Absence of this phosphorylation site led to greatly diminished p53 stability and function in response to γ-irradiation and increased spontaneous tumorigenesis in mice. Conversely, a phosphomimic model demonstrated prolonged p53 activation in cells treated with γ-irradiation, which revealed that phosphorylation of this Mdm2 residue controls the duration of the DNA damage response. Therefore, these mouse models have uncovered new roles for the p53-Mdm2 regulatory axis in vivo and will be useful reagents in future studies of posttranslational modifications in oncogene and DNA damage-induced tumorigenesis.

The innate immune system is the first line of defense against infectious agents. It is essential for protection against pathogens and stimulation of long-term adaptive immune responses. Therefore, deciphering the mechanisms of the innate immune system is crucial for understanding the integrated systems of host defense against microbial infections, which is conserved from insects to humans. Despite lacking a conventional adaptive immune system, insects can mount a robust immune response against a wide array of microbial pathogens. These innate immune mechanisms have been widely studied in Drosophila melanogaster, because of the model system’s powerful genetic, genomic and molecular tools. The Drosophila immunity relies on cellular and humoral innate immune responses to fight pathogens. The hallmark of the Drosophilahumoral immune response is the rapid induction of antimicrobial peptide genes in the fat body, the homolog of the mammalian liver. Expression of these antimicrobial peptide genes is controlled by two distinct immune signaling pathways, the Toll pathway and the IMD (immune deficiency) pathway. The Toll pathway is activated by fungal and Gram-positive bacterial infections, whereas the IMD pathway responds to Gram-negative bacteria. Both pathways culminate in activation of the Rel/NF-кB transcription factors DIF (Dorsal-related immunity factor), Dorsal and Relish, which in turn translocate to the nucleus to induce the antimicrobial peptide genes. DIF and Dorsal are activated by the Toll pathway and control induction of antimicrobial peptide genes such as Drosomycin. The NF-кB precursor Relish, which is composed of an N-terminal Rel homology domain and a C-terminal IкB-like domain, is activated by the IMD pathway and initiates transcription of antimicrobial peptide genes such as Diptericin. Although many components of the Drosophila immune signaling pathways have been identified, the detailed mechanisms of signal trans

Neurons exhibit highly polarized structures, including two morphologically and functionally distinct domains, axons and dendrites. Dendrites and axons receive versus send information, and proper execution of each requires different sets of molecules. Differential distribution of membrane proteins in distinct neuronal compartments plays essential roles in neuronal functions. The major goal of my doctoral thesis was to study the molecular mechanisms that govern the differential distribution of membrane proteins in neurons, using the Drosophilalarval mushroom body (MB) as a model system. My work was initiated by an observation of differential distribution of distinct Dscam isoforms in neurons. Dscam stands for Down Syndrome Cell Adhesion Molecule, which is a Drosophila homolog of human DSCAM. According to genomic analysis, DrosophilaDscam gene can generate more than 38,000 isoforms through alternative splicing in its exons 4, 6, 9 and 17. All Dscam isoforms share similar domain structures, with 10 immunoglobulin domains and 6 fibronectin type III repeats in the ectodomain, a single transmembrane domain and a cytoplasmic endodomain. There are two alternative exons in exon 17 (17.1 and 17.2), which encodes Dscam’s transmembrane domain. Interestingly, in ectopic expression, Dscam isoforms carrying exon 17.1 (Dscam[TM1]) can be preferentially localized to dendrites and cell bodies, while Dscam isoforms carrying exon 17.2 (Dscam[TM2]) are distributed throughout the entire neuron including axons and dendrites. To unravel the mechanisms involved in the differential distribution of Dscam[TM1] versus Dscam[TM2], I conducted a mosaic genetic screening to identify the possible factors affecting dendritic distribution of Dscam[TM1], established an in vivoTARGET system to better distinguish the differential distribution of Dscam, identified the axonal and dendritic targeting motifs of Dscam molecules and further showed that Dscam’s differential roles in dendrites versus axons are correlated with its localization. Several mutants affecting dendritic distribution of Dscam[TM1] have been identified using a MARCM genetic screen. Three of these mutants (Dlis1, Dmn and p24) are components of the dynein/dynactin complex. Silencing of other dynein/dynactin subunits and blocking dynein function with a dominant-negative Glued mutant also resulted in mislocalization of Dscam[TM1] from dendrites to axons. However, microtubule polarity in the mutant axons was maintained. Taken together, this was the first demonstration that the dynein/dynactin complex is involved in the polarized distribution of membrane proteins in neurons. To further examine how dynein/dynactin is involved in the dendritic distribution of Dscam[TM1], I compromised dynenin/dynactin function with dominant-negative Glued and transiently induced Dscam[TM1] expression. The results suggested that dynein/dynactin may not be directly involved in the targeting of newly synthesized Dscam[TM1] to dendrites. Instead, it plays a role in maintaining dendritic restriction of Dscam[TM1]. Notably, dynein/dynactin dysfunction did not alter distribution of another dendritic transmembrane protein Rdl (Resistant to Dieldrin), supporting involvement of diverse mechanisms in distributing distinct molecules to the dendritic membrane. To identify the targeting motifs of Dscam, I incorporated the TARGET (Temporal and regional gene expression targeting) system into mushroom body (MB) neurons, and this allowed the demonstration of the differential distribution of Dscam[TM1] and Dscam[TM2] with more clarity than conventional overexpression techniques. Using the TARGET system, I identified an axonal targeting motif located in the cytoplasmic juxtamemebrane domain of Dscam[TM2]. This axonal targeting motif is dominant over the dendritic targeting motif located in Dscam’s ectodomain. Scanning alanine mutagenesis demonstrated that two amino acids in the axonal targeting motif were essential for Dscam’s axonal distribution. Interestingly, swapping the cytoplasmic juxtamembrane portions between TM1 and TM2 not only reversed TM1’s and TM2’s differential distribution patterns but also their functional properties in dendrites versus axons. My thesis research also involved studying endodomain diversity of Dscam isoforms. Besides the diversity originally found in the ectodomain and transmembrane domain of Dscam, my colleagues and I further demonstrated the existence of four additional endodomain variants. These four variants are generated by skipping or retaining exon 19 or exon 23 through independent alternative splicing. Interestingly, different Dscam endodomain isoforms are expressed at different developmental stages and in different areas of the nervous system. Through isoform-specific RNA interference, we showed the differential involvement of distinct Dscam endodomains in specific neuronal morphogenetic processes. Analysis of the primary sequence of the Dscam endodomain indicated that endodomain variants may confer activation of different signaling pathways and functional roles in neuronal morphogenesis. In Summary, my thesis work identified and characterized several previously unknown mechanisms related to the differential distribution of membrane proteins in neurons. I showed that there may be a dynein/dynactin-independent mechanism for selective transport of dendritic membrane proteins to dendrites. Second, dynein/dynactin plays a maintenance role in dendritic restriction of Dscam[TM1]. Third, different membrane proteins may require distinct combinations of mechanisms to be properly targeted and maintained in certain neuronal compartments. Further analysis of the mutants indentified from my genetic screen will definitely help to resolve the missing pieces of the puzzle. These findings provide novel mechanistic insight into the differential distribution of membrane proteins in polarized neurons.

This thesis focuses on several aspects of olfactory processing in Drosophila. In chapter I and II, I will discuss how odorants are encoded in the brain. In both insects and mammals, olfactory receptor neurons (ORNs) expressing the same odorant receptor gene converge onto the same glomerulus. This topographical organization segregates incoming odor information into combinatorial maps. One prominent theory suggests that insects and mammals discriminate odors based on these distinct combinatorial spatial codes. I tested the combinatorial coding hypothesis by engineering flies that have only one class of functional ORNs and therefore cannot support combinatorial maps. These files can be taught to discriminate between two odorants that activate the single functional class of ORN and identify an odorant across a range of concentrations, demonstrating that a combinatorial code is not required to support learned odor discrimination. In addition, these data suggest that odorant identity can be encoded as temporal patterns of ORN activity. Behaviors are influenced by motivational states of the animal. Chapter III of this thesis focuses on understanding how motivational states control behavior. Appetitive memory in Drosophilaprovides an excellent system for such studies because the motivational state of hunger promotes reliance on learned appetitive cues whereas satiety suppresses it. We found that activation of neuropeptide F (dNPF) neurons in fed flies releases appetitive memory performance from satiety-mediated suppression. Through a GAL4 screen, we identified six dopaminergic neurons that are a substrate for dNPF regulation. In satiated flies, these neurons inhibit mushroom body output, thereby suppressing appetitive memory performance. Hunger promotes dNPF release, which blocks the inhibitory dopaminergic neurons. The motivational drive of hunger thus affects behavior through a hierarchical inhibitory control mechanism: satiety inhibits memory performance through a subset of dopaminergic neurons, and hunger promotes appetitive memory retrieval via dNPF-mediated disinhibition of these neurons. The aforementioned studies utilize sophisticated genetic tools for Drosophila. In chapter IV, I will talk about two new genetic tools. We developed a new technique to restrict gene expression to different subsets of mushroom body neurons with unprecedented precision. We also adapted the light-activated adenylyl cyclase (PAC) from Euglena gracilis as a light-inducable cAMP system for Drosophila. This system can be used to induce cAMP synthesis in targeted neurons in live, behaving preparations.

Vascular Endothelial Growth Factor Receptor-3 (VEGFR3/Flt4) and its ligand Vegfc are necessary for development of both blood and lymphatic vasculature in vertebrates. In zebrafish, Vegfc/Flt4 signaling is essential for formation of arteries, veins, and lymphatic vessels. Interestingly, Flt4 appears to utilize distinct signaling pathways during the development of each of these vessels. To identify components of this pathway, we performed a transgenic haploid genetic screen in zebrafish that express EGFP under the control of a blood vessel specific promoter. As a result, we indentified a mutant allele of vascular endothelial growth factor c (vegfc), vegfcum18. vegfcum18 mutants display defects in vein and lymphatic vessel development but normal segmental artery (SeA) formation. Characterization of this allele led to the finding that the primary defect in vegfcum18 mutants was a general failure in vein and lymphatic vessel sprouting. Further genetic and biochemical analysis of this mutant revealed profound paracrine defects, which likely result in the observed loss of lymphatic and venous structures. Furthermore, double mutant analysis demonstrated that defects during SeA formation in vegfcum18 mutants were masked by inputs from the Vegfa signaling pathway. Endothelial cell autonomous expression of vegfcum18 induced angiogenic effects on blood vessels while endothelial cells lacking vegfc displayed defects in tip cell occupancy, suggesting a cell autonomous-autocrine role for Vegfc during developmental angiogenesis. Finally, we present genetic evidence that links processing of Vegfc by Furin during the formation of lymphatics in zebrafish. Together the data presented here suggest two discrete modes of signaling during blood and lymphatic vessel development, thus implying that regulation of Vegfc secretion and processing may play a pivotal role in the formation of these distinct vessel types in zebrafish.

Small interfering RNAs (siRNAs), microRNAs (miRNAs), and piRNAs (piRNA) are conserved classes of small single-stranded ~21-30 nucleotide (nt) RNA guides that repress eukaryotic gene expression using distinct RNA Induced Silencing Complexes (RISCs). At its core, RISC is composed of a single-stranded small RNA guide bound to a member of the Argonaute protein family, which together bind and repress complementary target RNA. miRNAs target protein coding mRNAs—a function essential for normal development and broadly involved in pathways of human disease; small interfering RNAs (siRNA) defend against viruses, but can also be engineered to direct experimental or therapeutic gene silencing; piwi associated RNAs (piRNAs) protect germline genomes from expansion of parasitic nucleic acids such as transposons. Using the fruit fly, Drosophila melanogaster, as a model organism we seek to understand how small silencing RNAs are made and how they function. In Drosophila, miRNAs and siRNAs are proposed to have parallel, but separate biogenesis and effector machinery. miRNA duplexes are excised from imperfectly paired hairpin precursors by Dicer1 and loaded into Ago1; siRNA duplexes are hewn from perfectly paired long dsRNA by Dicer2 and loaded into Ago2. Contrary to this model we found one miRNA, miR-277, is made by Dicer1, but partitions between Ago1 and Ago2 RISCs. These two RISCs are functionally distinct—Ago2 could silence a perfectly paired target, but not a centrally bulged target; Ago1 could silence a bulged target, but not a perfect target. This was surprising since both Ago1 and Ago2 have endonucleolytic cleavage activity necessary for perfect target cleavage in vitro. Our detailed kinetic studies suggested why—Ago2 is a robust multiple turnover enzyme, but Ago1 is not. Along with a complementary in vitro study our data supports a duplex sorting mechanism in which Diced duplexes are released, and rebind to Ago1 or Ago2 loading machinery, regardless of which Dicer produced them. This allows structural information embedded in small RNA duplexes to direct small RNA loading into Ago1 and/or Ago2, resulting in distinct regulatory outputs. Small RNA sorting also has chemical consequences for the small RNA guide. Although siRNAs were presumed to have the signature 2′, 3′ hydroxyl ends left by Dicer, we found that small RNAs loaded into Ago2 or Piwi proteins, but not Ago1, are modified at their 3´ ends by the RNA 2´-O-methyltransferase DmHen1. In plants Hen1 modifies the 3´ ends all small RNAs duplexs, protecting and stabilizing them. Implying a similar function in flies, piRNAs are smaller, less abundant, and their function is perturbed in hen1 mutants. But unlike plants, small RNAs are modified as single-strands in RISC rather than as duplexes. This nicely explains why the dsRNA binding domain in plant Hen1 was discarded in animals, and why both dsRNA derived siRNAs and ssRNA derived piRNAs are modified. The recent discovery that both piRNAs and siRNAs target transposons links terminal modification and transposon silencing, suggesting that it is specialized for this purpose.

In the preceding chapters, I have discussed my doctoral research on studying the siRNA loading pathway in Drosophila using both biochemical and genetic approaches. We established a gel shift system to identify the intermediate complexes formed during siRNA loading. We detected at least three complexes, named complex B, RISC loading complex (RLC) and RISC. Using kinetic modeling, we determined that the siRNA enters complex B and RLC early during assembly when it remains double-stranded, and then matures in RISC to generate Argonaute bearing only the single-stranded guide. We further characterized the three complexes. We showed that complex B comprises Dcr-1 and Loqs, while both RLC and RISC contain Dcr-2 and R2D2. Our study suggests that the Dcr-2/R2D2 heterodimer plays a central role in RISC assembly. We observed that Dcr-1/Loqs, which function together to process pre-miRNA into mature miRNA, were also involved in siRNA loading. This was surprising, because it has been proposed that the RNAi pathway and miRNA pathway are separate and parallel, with each using a unique set of proteins to produce small RNAs, to assemble functional RNA-guided enzyme complexes, and to regulate target mRNAs. We further examined the molecular function of Dcr-1/Loqs in RNAi pathway. Our data suggest that, in vivo and in vitro, the Dcr-1/Loqs complex binds to siRNA. In vitro, the binding of the Dcr-1/Loqs complex to siRNA is the earliest detectable step in siRNA-triggered Ago2-RISC assembly. Futhermore, the binding of Dcr-1/Loqs to siRNA appears to facilitate dsRNA dicing by Dcr-2/R2D2, because the dicing activity is much lower in loqslysate than in wild type. Long inverted repeat (IR) triggered white silencing in fly eyes is an example of endogenous RNAi. Consistent with our finding that Dcr-1/Loqs function to load siRNA, less white siRNA accumulates in loqs mutant eyes compared to wild type. As a result, loqs mutants are partially defective in IR trigged whitesilencing. Our data suggest considerable functional and genetic overlap between the miRNA and siRNA pathways, with the two sharing key components previously thought to be confined to just one of the two pathways. Based on our study on siRNA loading pathway, we also elucidated the molecular function of Armitage (Armi) protein in RNAi. We showed that armi is required for RNAi. Lysates from armi mutant ovaries are defective for RNAi in vitro. Native gel analysis of protein-siRNA complexes suggests that armi mutants support early steps in the RNAi pathway, i.e., the formation of complex B and RLC, but are defective in the production of the RISC.

Erythropoiesis maintains stable tissue oxygenation in the basal state, while accelerating red cell production in anemia, blood loss or high altitude. The principal regulator of erythropoiesis is the hormone erythropoietin (Epo). In response to hypoxic stress, Epo can increase a 1000-fold, driving erythropoietic rate by up to 10-fold. It’s been suggested that survival pathways activated by the Epo receptor (EpoR) underlie its regulation of erythropoietic rate. A number of apparently redundant EpoR survival pathways were identified in vitro, raising the possibility of their functional specialization in vivo. Here I assessed the roles of three survival pathways activated by EpoR in erythroblasts in-vivo: the suppression of cell-surface Fas and FasL, the suppression of the pro-apoptotic regulator Bim, and the induction of the anti-apoptotic regulator Bcl-xL. I used the novel CD71/Ter119 flow-cytometric method of identifying erythroblast maturation stages in vivo to measure these apoptotic pathways in fetal liver and adult erythropoietic tissues. I found that these pathways differ markedly in their regulation of erythropoietic rate. Using mouse genetic models, I found that apoptosis mediated by interaction between erythroblasts that co-express cell-surface Fas and FasL plays a key autoregulatory role in stabilizing the size of the erythroblast pool in the basal state. Further, mice mutant for Fas or FasL showed a delayed erythropoietic response to hypoxia or high Epo. This suggests that Fas and FasL accelerate the stress response by providing an apoptotic ‘cell reserve’ that can be rescued by Epo in stress. I also examined the in-vivo behavior of two cell-intrinsic apoptotic regulators, Bcl-xL and Bim, previously unexamined in stress. The induction of Bcl-xL was rapid but transient, whilst the suppression of Bim was slower but persistent. My data suggest that Bcl-xL is a key mediator of EpoR’s anti-apoptotic signal very early in the stress response, before Bim and Fas are suppressed. Bcl-xL adaptation to high Epo occurs through inhibition of Stat5 activation, and resets it for the next acute stress. My findings suggest that in vivo, Epo regulates erythropoietic rate through erythroblast apoptosis, and that various apoptotic regulators play distinct and unique roles in this process. My work provides new molecular insights into erythropoiesis that are relevant to cytokine biology and to clinical approaches of disease treatment.