The process of HIV-1 particle production is a multi-step process directed by the viral structural protein Gag. As Gag is the only viral protein required to form virus-like particles, it presents a viable target for anti-viral therapeutics of which there are currently none. Although the functions of Gag during the particle assembly process have been well characterized, one of the least known parts of the assembly process is how Gag is targeted to the site of virus assembly. Two main virus assembly sites have been identified in cells that support HIV-1 replication: the plasma membrane or multivesicular bodies (MVBs). However the mechanism by which Gag is targeted to either of these sites remains unknown. The δ subunit of Adaptor Protein Complex 3 has previously been identified as a cellular co-factor for HIV-1 Gag and was reported to mediate Gag trafficking to MVBs, providing a mechanism for Gag targeting to this assembly site. Additionally, AP-3δ was reported to be required for HIV-1 production, suggesting that Gag to MVB targeting is also required for HIV-1 production. The work presented in this thesis further investigates the role of AP-3δ in Gag trafficking to MVBs and its role in HIV-1 production in previously unexplored host environments. Through the use of RNA interference-mediated depletion of AP-3δ, we determined that AP-3δ is dispensible for virus replication in infected HeLa cells, chronically infected HeLa-LAV cells and infected primary human monocyte-derived macrophages. We concomitantly disrupted AP-3 function by disrupting its association with membranes and observed no effect on virus production. Collectively, these results demonstrate that AP-3δ is not required for HIV-1 replication. However, AP-3δ was demonstrated to be required for Gag targeting to MVBs thus presenting a new model for the function of AP-3δ in the context of HIV-1 replication.

The Human Immunodeficiency Virus (HIV) is able to infect CD4+ T cells as well as macrophages. Macrophage-tropism has been linked to determinants in the envelope of HIV. These determinants allow envelopes to exploit low levels of CD4 for infection. Macrophages are an important reservoir of virus, especially during chronic infection, and are likely responsible for the bulk of virus produced after CD4+T cells have declined. Viral factors that may impact the ability to infect macrophages are worth studying because this cell type is so important in infection. It was previously reported that the macrophage-tropic primary isolate AD8 was vpu-independent. The molecular clone YU-2, derived from brain tissue without culture, was also reported to be macrophage-tropic despite having a mutation in the vpu start codon. It was therefore possible that vpu-independent envelopes could evolve in vivo. To examine this possibility, I constructed chimeras containing wild type or defective vpu start codons, and gp160 sequences from AD8, YU-2 or SF162 (a vpu-dependent control). I also used full length AD8 and YU-2 with wild type or defective vpu start codons. I infected macrophages with equal amounts of virus, and measured viral output over two weeks. Viruses with defective vpu start codons were released to lower levels compared to their wild type vpucounterparts. In contrast to previous reports, the AD8 envelope is not vpu-independent for replication in macrophages. The YU-2 envelope is also not vpu-independent. Macrophage-tropic envelopes from late stages of infection can be sensitive to antibodies that bind the CD4 binding site on gp120, implying that macrophage-tropic envelopes have more exposed CD4 binding sites. Neutralizing antibodies may act as modulators of macrophage-tropism over the course of infection. Using chimeras containing gp120 sequences derived from the PBMC of four HIV+patients, I examined the capacity for envelopes to infect macrophages. Three patients (MM1, 4, and 8) had macrophage-tropic envelopes before and after developing autologous neutralizing antibodies. Three patients (MM1, 4, and 23) developed heterologous antibodies against IIIB, an easily neutralized T-cell line adapted strain of HIV-1. This data indicates that macrophage-tropism in these patients is not modulated by the presence of neutralizing antibodies. The macrophage-tropism of envelopes tends to segregate depending on the tissue origin of the virus. Envelopes from two separate tissues from the same patient exhibit very different infectivity characteristics. The B33 envelope, from brain tissue, is very infectious and is macrophage-tropic, while the LN40 envelope, from lymph node tissue, is weakly infectious and is not macrophage-tropic. Replacing the entire gp41 of LN40 with that of B33 restores some infectivity to LN40. The cytoplasmic domain of gp41 contains many motifs important for assembly and infectivity. To examine which motifs are responsible for the weak infectivity of LN40, I made chimeras of gp41, as well as point mutations in gp41. The LN40 chimera containing the entire gp41 of B33 restored the most infectivity. Point mutations in the palmitoylation site, Pr55gagbinding region, and dileucine motif at the C-terminus also restored infectivity when combined. Determinants in the gp41 cytoplasmic domain are responsible for the weak infectivity of LN40; however, it is possible that there are contributing determinants in gp120, such as the ability to use low levels of CD4. Here, I examined how changes in the vpu and env genes of HIV-1 can impact infectivity, especially infectivity of macrophages. Changes that adversely impact the virus’ ability to infect macrophages may also impact the overall course of disease. However, the data here show that retaining the ability to infect, and replicate in, macrophages give HIV an advantage. I speculate that retaining the ability to infect macrophages gives the virus a reservoir for later in disease, when CD4+ T cells have been depleted, as well as way of avoiding neutralizing antibodies. This work further defines the importance of macrophages in HIV-1 infectivity and disease.

Cells need to complete DNA replication in a timely and error-free manner. To ensure that replication is completed efficiently and in a finite amount of time, cells regulate origin firing. To prevent any errors from being transmitted to the next generation, cells have the checkpoint mechanism. The S-phase DNA damage slows replication to allow the cell to repair the damage. The mechanism of replication slowing by the checkpoint was not clear in fission yeast, Schizosaccharomyces pombe, at the start of my thesis. The downstream targets of the DNA damage checkpoint in fission yeast were also unclear. I worked on identifying the downstream targets for the checkpoint by studying if Cdc25, a phosphatase, is a target of the checkpoint. Work from our lab has shown that origin firing is stochastic in fission yeast. Origins are also known to be inefficient. Inefficient origins firing stochastically would lead to large stretches of chromosome where no origins may fire randomly leading to long replication times, an issue called the random gap problem. However, cells do not take a long time to complete replication and the process of replication itself is efficient. I focused on understanding the mechanism by which cells complete replication and avoid the random gap problem by attempting to measure the firing efficiency of late origins. Genome-wide origin studies in fission yeast have identified several hundred origins. However, the resolution of these studies can be improved upon. I began a genome-wide origin mapping study using deep sequencing to identify origins at a greater resolution compared to the previous studies. We have extended our origin search to two other Schizosaccharomyces species- S. octosporus and S. japonicus.There have been no origin mapping studies on these fission yeasts and identifying origins in these species will advance the field of replication. My thesis research shows that Cdc25 is not a target of the S-phase DNA damage checkpoint. I showed that DNA damage checkpoint does not target Cdc2-Y15 to slow replication. Based on my preliminary observation, origin firing might be inhibited by the DNA damage checkpoint as a way to slow replication. My efforts to measure the firing efficiency of a late replicating sequence were hindered by potentially unidentified inefficient origins firing at a low rate and replicating the region being studied. Studying the origin efficiency was maybe further complicated by neighboring origins being able to passively replicate the region. To identify origins in recently sequenced Schizosaccharomyces species, we initiated the genome-wide origin mapping. The mapping was also done on S. pombe to identify inefficient origins not mapped by other mapping studies. My work shows that deep sequencing can be used to map origins in other species and provides a powerful tool for origin studies.

Innate immunity is the first line of defense against invading pathogens. It functions to eliminate pathogens and also to control infections. The innate immune response is also important for the development of pathogen-specific adaptive immune responses. As a result, the study of innate immune signaling pathways is crucial for understanding the interactions between host and pathogen. Unlike mammals, insects lack a classical adaptive immune response and rely mostly on innate immune responses. Innate immune mechanisms have been widely studied in the fruit fly, Drosophila melanogaster. The genetic and molecular tools available in the Drosophila system make it an excellent model system for studying immunity. Furthermore, the innate immune signaling pathways used by Drosophila show strong homology to those of vertebrates making them ideal for studying these pathways. 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 production of these antimicrobial peptides is regulated by two immune signaling pathways-Toll and Immune Deficency (IMD) pathways. The Toll pathway responds to many Gram-positive bacterial and fungal infections , while the IMD pathway is potently activated by DAP-type peptidoglycan (PGN) from Gram-negative bacteria and certain Gram-positive bacteria. Two receptors, PGRP-LC and PGRP-LE, are able to recognize DAP-type PGN at the cell surface or in the cytosol, respectively, and trigger the IMD pathway. Upon binding DAP-type PGN, both PGRP-LC and PGRP-LE dimerize/ multimerize and signal to the downstream components of IMD pathway. It is unclear how the receptor activates its downstream components. My work has focused on understanding the molecular events that take place at the receptors following there activation. In these studies I have identified a common motif in the N-terminal domains of both the receptors, known as the RHIM-like domain. The RHIM-like domain is critical for signaling by either receptor, but the mechanism(s) involved remain unclear. IMD, a downstream component of the pathway, associates with both PGRP-LC and -LE but the interaction of PGRP-LC with IMD is not mediated through its RHIM-like domain. Also, mutations affecting the PGRP-LC RHIM-like motif are defective in all known downstream signaling events. However, the RHIM-like mutant receptors are capable of serving as a platform for the assembly of all known components of a receptor proximal signaling complex. These results suggest that another, unidentified component of the IMD signaling pathway may function to mediate interaction with the RHIM-like motif. I performed a yeast two-hybrid screen to identify proteins that might interact with the receptor PGRP-LC through its RHIM- like domain. With this approach, two new components of the IMD pathway were identified. The first component I characterized is called Rudra and it is a critical feedback inhibitor of peptidoglycan receptor signaling. The other factor is known as RYBP, it includes a highly conserved ubiquitin binding motif (NZF), and RNAi studies suggest it is a critical component of the IMD pathway. The identification and characterization of these two new components of the IMD pathway has provided a new insight into the molecular events that take place proximal to the receptor.

Human papillomavirus infection is the cause of more than 99% of cervical cancer cases. The current vaccine is ineffective therapeutically; highlighting the need for continued papillomavirus research. One avenue that could be explored in this regard is the function of the papillomavirus E2 regulatory proteins. HPV E2 represses expression of the viral E6 and E7 oncoproteins. Reintroduction of E2 into cervical carcinoma cells results in growth arrest and cellular senescence. Understanding the mechanism of how E2 regulates the early promoter may be key to developing new therapeutic and prophylactic vaccines. Here, we describe regulation of E2 through acetylation and possibly through direct interaction with a novel cellular interacting protein, RINT1. Histone acetyltransferase (HAT) proteins have been demonstrated to interact with Bovine Papillomavirus (BPV) and Human Papillomavirus (HPV) E2 proteins as well as enhance E2 dependant transcription luciferase reporter plasmid containing E2 binding sites. We demonstrate that HATs p300, CBP, and pCAF are limiting for E2 dependant transcriptional activation and that each protein functions independently. We have also identified that BPV-1 E2 is a substrate for acetylation by p300. Mutants of E2 that cannot be acetylated on lysines 111 or 112, display abnormal transcriptional phenotypes. Cells deficient in p300 display similar transcriptional defects that are intensified by CBP depletion. We propose that acetylation of BPV-1 E2 is necessary for transcriptional activation. Acetylation generates a binding site through which a co-factor may interact via a bromodomain. Regulation of E2 dependent transcriptional activation through a post-transcriptional modification represents a novel method through which BPV-1 controls gene expression. We also present evidence for a direct interaction between BPV-1 E2 and the cellular factor RINT1. This interaction does not appear to be critical for transcriptional regulation; however, several other functional pathways are indicated by the cellular complexes in which RINT1 functions. Some of these, such as ER/Golgi vesicular transport and hTERT independent telomere maintenance, are pathways in which E2 has no known role. Further investigation into regulation and consequences of E2 acetylation and the biological significance of the interaction between E2 and RINT1 could prove important in understanding the complex role of E2 in papillomavirus infection.

Around thirty years ago HIV-1 was identified, and from that point the known epidemic has grown to over 30 million infected individuals. Early on in the course of HIV-1 research, viruses were classified as either syncytia inducing, CXCR4-using, T-cell tropic or non-syncytia inducing, CCR5-using, macrophage tropic. Since that time, several groups have shown that this is an oversimplification. There is a great deal of diversity amongst CCR5-using HIV-1 variants. There remains a great deal to be discovered regarding HIV-1 CCR5-tropism and how this affects other aspects of HIV-1 infection. The CD4 binding site (CD4bs) on the HIV-1 envelope plays a major role in determining the capacity of R5 viruses to infect primary macrophages. Thus, envelope determinants within or proximal to the CD4bs have been shown to control the use of low CD4 levels on macrophages for infection. These residues affect the affinity for CD4 either directly or indirectly by altering the exposure of CD4 contact residues. In this thesis, a single amino acid determinant is described in the V1 loop that also modulates macrophage tropism. I identified an E153G substitution that conferred high levels of macrophage infectivity for several heterologous R5 envelopes, while the reciprocal G153E substitution abrogated infection. Shifts in macrophage tropism were associated with dramatic shifts in sensitivity to the V3 loop monoclonal antibody (MAb), 447-52D and soluble CD4, as well as more modest changes in sensitivity to the CD4bs MAb, b12. These observations are consistent with an altered conformation or exposure of the V3 loop that enables the envelope to use low CD4 levels for infection. The modest shifts in b12 sensitivity suggest that residue 153 impacts on the exposure of the CD4bs. However, the more intense shifts in sCD4 sensitivity suggest additional mechanisms that likely include an increased ability of the envelope to undergo conformational changes following binding to suboptimal levels of cell surface CD4. In summary, a conserved determinant in the V1 loop modulates the V3 loop to prime low CD4 use and macrophage infection. In addition to determinants, this thesis seeks to evaluate the roles of macrophage tropic and non-macrophage tropic envelopes during the course of infection. Non-macrophage tropic virus predominates in immune tissue throughout infection, even in individuals suffering from HIV-associated dementia (HAD) who are known to carry many macrophage tropic viruses. There must be some advantage for these non-macrophage tropic viruses allowing them to persist in immune tissue throughout the disease. This thesis demonstrates that there is no advantage for these viruses to directly infect CD4+ T-cells, nor is there an advantage for them to be preferentially transmitted by dendritic cells to CD4+ T-cells. Given that transmitted/founder (T/F) viruses may preferentially interact with α4β7, and T/F viruses are non-macrophage tropic, I tested whether non-mac viruses could utilize α4β7 to their advantage. These experiments show that macrophage tropism does not play a role in gp120 interactions with α4β7. I evaluated whether there was a distinct disadvantage to macrophage tropic Envs, given their ability to infect dendritic cells and possibly stimulate the innate immune response. Using infected monocyte-derived dendritic cells (MDDCs), it was shown that mac-tropic Envs do not generate a significant immune response. These experiments demonstrate that there does not appear to be any advantage to non-macrophage tropic Envs, and that macrophage tropic Envs are able to infect CD4+ T-cells more efficiently, as well as DCs.

Replication of HIV-1 requires the assembly and release of mature and infectious viral particles. In order to accomplish this goal, HIV-1 has evolved multiple methods to interact with the host cell. HIV-1 recruits the host cell ESCRT machinery to facilitate the release of nascent viral particles from the host cell membrane. Recruitment of these cellular factors is dependent on the presence of short motifs in Gag referred to as Late-domains. Deletion or mutation of these domains results in substantial decrease in the release of infectious virions. However, previously published work has indicated that over-expression of the E3 ubiquitin ligase, NEDD4.2s is able to robustly rescue release of otherwise budding-defective HIV-1 particles. This rescue is specific to the NEDD4.2s isoform as related E3 ubiquitin ligases display no ability to rescue particle release. In addition, rescue of particle release is dependent on the presence of the partial C2 domain and a catalytically active HECT domain of NEDD4.2s. Here I provide evidence supporting the hypothesis that a partial C2 domain of NEDD4.2s constitutes a Gag interacting module capable of targeting the HECT domains of other E3 ubiquitin ligases to HIV-1 Gag. Also, by generating chimeras between HECT domains shown to form poly-ubiquitin chains linked through either K48 or K63 of ubiquitin, I demonstrate that the ability of NEDD4.2s to catalyze the formation of K63-polyubiquitin chains is required for its stimulation of HIV-1 L-domain mutant particle release. In addition, I present findings from on-going research into the role of the HIV-1 accessory protein Nef during viral replication using the culture T-cell line, MOLT3. My current findings indicate that downregulation of CD4 from the host cell membrane does not solely account for the dramatic dependence of HIV-1 replication on Nef expression in this system. In addition, I present evidence indicating that Nef proteins from diverse HIV-1 Groups and strains are capable of enhancing HIV-1 replication in this system. Analysis of a range of mutations in Nef known to impact interaction with cellular proteins suggest that the observed replication enhancement requires Nef targeting to the host cell membrane and may also require the ability to interact with select Src-kinases. Lastly, we find that the ability of Nef to enhance replication in this system is separate from any increase in viral particle infectivity, in agreement with current literature.

The family of runt related transcription factors (RUNX/Cbfa/AML/PEBP2) are essential for cellular differentiation and fetal development. RUNX factors are distributed throughout the nucleus in punctate foci that are associated with the nuclear matrix/scaffold and generally correspond with sites of active transcription. Truncations of RUNX proteins that eliminate the C-terminus including a 31-amino acid segment designated the nuclear matrix targeting signal (NMTS) lose nuclear matrix association and result in lethal hematopoietic (RUNX1) and skeletal (RUNX2) phenotypes in mice. These findings suggest that the targeting of RUNX factors to subnuclear foci may mediate the formation of multimeric regulatory complexes and contribute to transcriptional control. In this study, we hypothesized that RUNX transcription factors may dynamically move through the nucleus and associate with subnuclear domains in a C-terminal dependent mechanism to regulate transcription. Therefore, we investigated the subnuclear distribution and mobility of RUNX transcription factors in living cells using enhanced green fluorescent protein (EGFP) fused to RUNX proteins. The RUNX C-terminus was demonstrated to be necessary for the dynamic association of RUNX with stable subnuclear domains. Time-lapse fluorescence microscopy showed that RUNX1 and RUNX2 localize to punctate foci that remain stationary in the nuclear space in living cells. By measuring fluorescence recovery after photobleaching, both RUNX1 and RUNX2 were found to dynamically and rapidly associate with these subnuclear foci with a half-time of recovery in the ten-second time scale. A large immobile fraction of RUNX1 and RUNX2 proteins was observed in the photobleaching experiments, which suggests that this fraction of RUNX1 and RUNX2 proteins are immobilized through the C-terminal domain by interacting with the nuclear architecture. Truncation of the C-terminus of RUNX2, which removes the NMTS as well as several co-regulatory protein interaction domains, increases the mobility of RUNX2 by at least an order of magnitude, resulting in a half-time of recovery equivalent to that of EGFP alone. Contributions of the NMTS sequence to the subnuclear distribution and mobility of RUNX2 were further assessed by creating point mutations in the NMTS of RUNX2 fused to EGFP. The results show that these point mutations decrease, but do not abolish, association with the nuclear matrix compared to wild-type EGFP-RUNX2. Three patterns of subnuclear distribution were similarly observed in living cells for both NMTS mutants and wild-type RUNX2. Furthermore, the NMTS mutations showed no measurable effect on the mobility of RUNX2. However, the mobility of RUNX proteins in each of the different subnuclear distributions observed in living cells were significantly different from each other. The punctate distribution appears to correlate with higher fluorescence intensity, suggesting that the protein concentration in the cell may have an effect on the formation or size of the foci. These findings suggest that the entire NMTS and/or the co-regulatory protein interaction domains may be necessary to immobilize RUNX2 proteins. Because RUNX factors contain a conserved intranuclear targeting signal, we examined whether RUNX1 and RUNX2 are targeted to common subnuclear domains. The results show that RUNX1 and RUNX2 colocalized in common subnuclear foci. Furthermore, RUNX subnuclear foci contain the co-regulatory protein CBFβ, which heterodimerizes with RUNX factors, and nascent transcripts as shown by BrUTP incorporation. These results suggest that RUNX subnuclear foci may represent sites of transcription containing multi-subunit transcription factor complexes. RUNX2 transcription factors induce expression of the osteocalcin promoter during osteoblast differentiation and to study both RUNX2 and osteocalcin function, it would be helpful to have transgenic mice in which OC expression could be easily evaluated. Therefore, to assess the in vivo regulation of osteocalcin by RUNX protein, we generated transgenic mice expressing EGFP controlled by the osteocalcin promoter. Our results show that EGFP is expressed from the OC promoter in a cultured osteosarcoma cell line, but not in a kidney cell line, and is induced by vitamin D3. Furthermore, the OC-EGFP transgenic mice specifically express EGFP in osteoblasts and osteocytes in bone tissues. Moreover, EGFP is expressed in mineralized bone nodules of differentiated bone marrow derived from transgenic mice. Thus, these mice produce a good model for studying the in vivo effects of RUNX-mediated osteocalcin regulation and for developing potential drug therapies for bone diseases. Taken together, our results in living cells support the conclusion that RUNX transcription factors dynamically associate with stationary subnuclear foci in a C-terminal dependent mechanism to regulate gene expression. Moreover, RUNX subnuclear foci represent transcription sites containing nascent transcripts and co-regulatory interacting proteins. These conclusions provide a mechanism for how RUNX transcription factors may associate with subnuclear foci to regulate gene expression. Furthermore, the OC-EGFP transgenic mice now provide a useful tool for studying the in vivo function and regulation of osteocalcin by RUNX proteins during osteoblast differentiation and possibly for developing therapeutic drugs for treatment of bone diseases in the future.

The prevalence of diabetes is approaching epidemic proportions worldwide. There is currently no cure for type 1 diabetes, and successful treatment requires constant monitoring of blood sugars and use of exogenous insulin to prevent hyperglycemia. Diabetes will be curable when pancreatic β-islet cells can be transplanted into diabetes patients without requiring long-term immunosuppression. This will require learning more about the induction of functional tolerance, a state that maintains the competence of the immune system to most antigens but protects graft-specific antigens from immune rejection, permitting transplantation. One known mechanism of peripheral tolerance is T cell anergy, a phenotype of hypo-reponsiveness in CD4+ T cells. The focus of this thesis is a description of factors shown to be specific to the induction and maintenance of T cell anergy, whose loss reverses the anergic phenotype, restoring the ability of the cells to proliferate in response to antigen. The first of these is Egr-2, a zinc-finger transcription factor, whose presence is required for the induction of anergy induced in T cell clones by TCR stimulation in the absence of costimulation. Egr-2 is shown to be important to anergy induction but not anergy maintenance. In contrast, a negative costimulation receptor, PD-1, is shown to be necessary for the maintenance of anergy. It is possible that learning more about the genetic factors that orchestrate T cell anergy will prove useful in the development of tolerance-based protocols for organ and tissue transplantation without the use of long-term immunosuppression.

mRNA degradation is a fundamental process that controls both the level and the fidelity of gene expression. Using a combination of bioinformatic, genomic, genetic, and molecular biology approaches, we have shown that Edc3p, a yeast mRNA decay factor, controls the stability of the intron-containing YRA1 pre-mRNA. We found that Edc3p-mediated degradation of YRA1 pre-mRNA: 1) is a component of a negative feedback loop involved in the autoregulation of YRA1, 2) takes place in the cytoplasm, 3) is independent of translation, 4) occurs through a deadenylation-independent decapping and 5΄ to 3΄ exonucleotic decay mechanism, and 5) is controlled by specific cis-acting elements and trans-regulatory factors. Cis-regulation of YRA1 pre-mRNA degradation is complicated and precise. Sequences in exon1 inhibit YRA1 pre-mRNA splicing and/or promote pre-mRNA export in a size-dependent but sequence-independent manner. Sequences in the intron dictate the substrate specificity for Edc3p-mediated decay. Five structurally different but functionally interdependent modules were identified in the YRA1 intron. Two modules, designated Edc3p-responsive elements (EREs), are required for triggering an Edc3p-response. Three other modules, designated translational repression elements (TREs), are required for repressing translation of YRA1 pre-mRNA. TREs enhance the efficiency of the response of the EREs to Edc3p by inhibiting translation-dependent nonsense-mediated mRNA decay (NMD). Trans-regulation of YRA1 pre-mRNA is governed by Yra1p, which inhibits YRA1 pre-mRNA splicing and commits the pre-mRNA to nuclear export, and the RNP export factors, Mex67p and Crm1p, which jointly promote YRA1 pre-mRNA export. Mex67p also appears to interact with sequences in the YRA1 intron to promote translational repression and to enhance the Edc3p response of YRA1 pre-mRNA. These results illustrate how common steps in the nuclear processing, export, and degradation of a transcript can be uniquely combined to control the expression of a specific gene and suggest that Edc3p-mediated decay may have additional regulatory functions in eukaryotic cells.