Adam Drake

Densely Interconnected Transcriptional Circuits Control Cell States in Human Hematopoiesis

Though many individual transcription factors are known to regulate hematopoietic differentiation, major aspects of the global architecture of hematopoiesis remain unknown. Here, we profiled gene expression in 38 distinct purified populations of human hematopoietic cells and used probabilistic models of gene expression and analysis of cis-elements in gene promoters to decipher the general organization of their regulatory circuitry. We identified modules of highly coexpressed genes, some of which are restricted to a single lineage but most of which are expressed at variable levels across multiple lineages. We found densely interconnected cis-regulatory circuits and a large number of transcription factors that are differentially expressed across hematopoietic states. These findings suggest a more complex regulatory system for hematopoiesis than previously assumed.

Sensitizing Protective Tumor Microenvironments to Antibody-Mediated Therapy

Therapy-resistant microenvironments represent a major barrier toward effective elimination of disseminated malignancies. Here, we show that select microenvironments can underlie resistance to antibody-based therapy. Using a humanized model of treatment refractory B cell leukemia, we find that infiltration of leukemia cells into the bone marrow rewires the tumor microenvironment to inhibit engulfment of antibody-targeted tumor cells. Resistance to macrophage-mediated killing can be overcome by combination regimens involving therapeutic antibodies and chemotherapy. Specifically, the nitrogen mustard cyclophosphamide induces an acute secretory activating phenotype (ASAP), releasing CCL4, IL8, VEGF, and TNFα from treated tumor cells. These factors induce macrophage infiltration and phagocytic activity in the bone marrow. Thus, the acute induction of stress-related cytokines can effectively target cancer cells for removal by the innate immune system. This synergistic chemoimmunotherapeutic regimen represents a potent strategy for using conventional anticancer agents to alter the tumor microenvironment and promote the efficacy of targeted therapeutics.

Mesenchymal Stem Cells Secreting Angiopoietin-Like-5 Support Efficient Expansion of Human Hematopoietic Stem Cells Without Compromising Their Repopulating Potential

Clinical and preclinical applications of human hematopoietic stem cells (HSCs) are often limited by scarcity of cells. Expanding human HSCs to increase their numbers while maintaining their stem cell properties has therefore become an important area of research. Here, we report a robust HSC coculture system wherein cord blood CD34(+) CD133(+) cells were cocultured with mesenchymal stem cells engineered to express angiopoietin-like-5 in a defined medium. After 11 days of culture, SCID repopulating cells were expanded ~60-fold by limiting dilution assay in NOD-scid Il2rg(-/-) (NSG) mice. The cultured CD34(+) CD133(+) cells had similar engraftment potential to uncultured CD34(+) CD133(+) cells in competitive repopulation assays and were capable of efficient secondary reconstitution. Further, the expanded cells supported a robust multilineage reconstitution of human blood cells in NSG recipient mice, including a more efficient T-cell reconstitution. These results demonstrate that the expanded CD34(+) CD133(+) cells maintain both short-term and long-term HSC activities. To our knowledge, this ~60-fold expansion of SCID repopulating cells is the best expansion of human HSCs reported to date. Further development of this coculture method for expanding human HSCs for clinical and preclinical applications is therefore warranted.

Human CD34+ CD133+ Hematopoietic Stem Cells Cultured with Growth Factors Including Angptl5 Efficiently Engraft Adult NOD-SCID Il2rγ−/− (NSG) Mice

Increasing demand for human hematopoietic stem cells (HSCs) in clinical and research applications necessitates expansion of HSCs in vitro. Before these cells can be used they must be carefully evaluated to assess their stem cell activity. Here, we expanded cord blood CD34+ CD133+ cells in a defined medium containing angiopoietin like 5 and insulin-like growth factor binding protein 2 and evaluated the cells for stem cell activity in NOD-SCID Il2rg−/− (NSG) mice by multi-lineage engraftment, long term reconstitution, limiting dilution and serial reconstitution. The phenotype of expanded cells was characterized by flow cytometry during the course of expansion and following engraftment in mice. We show that the SCID repopulating activity resides in the CD34+ CD133+ fraction of expanded cells and that CD34+ CD133+ cell number correlates with SCID repopulating activity before and after culture. The expanded cells mediate long-term hematopoiesis and serial reconstitution in NSG mice. Furthermore, they efficiently reconstitute not only neonate but also adult NSG recipients, generating human blood cell populations similar to those reported in mice reconstituted with uncultured human HSCs. These findings suggest an expansion of long term HSCs in our culture and show that expression of CD34 and CD133 serves as a marker for HSC activity in human cord blood cell cultures. The ability to expand human HSCs in vitro should facilitate clinical use of HSCs and large-scale construction of humanized mice from the same donor for research applications.

Engineering humanized mice for improved hematopoietic reconstitution

Humanized mice are immunodeficient animals engrafted with human hematopoietic stem cells that give rise to various lineages of human blood cells throughout the life of the mouse. This article reviews recent advances in the generation of humanized mice, focusing on practical considerations. We discuss features of different immunodeficient recipient mouse strains, sources of human hematopoietic stem cells, advances in expansion and genetic modification of hematopoietic stem cells, and techniques to modulate the cytokine environment of recipient mice, in order to enhance reconstitution of specific human blood lineage cells. We highlight the opportunities created by new technologies and discuss practical considerations on how to make best use of the widening array of basic models for specific research applications.

Rapid generation of human B-cell lymphomas via combined expression of Myc and Bcl2 and their use as a preclinical model for biological therapies

Although numerous mouse models of B-cell malignancy have been developed via the enforced expression of defined oncogenic lesions, the feasibility of generating lineage-defined human B-cell malignancies using mice reconstituted with modified human hematopoietic stem cells (HSCs) remains unclear. In fact, whether human cells can be transformed as readily as murine cells by simple oncogene combinations is a subject of considerable debate. Here, we describe the development of humanized mouse model of MYC/BCL2-driven ‘double-hit’ lymphoma. By engrafting human HSCs transduced with the oncogene combination into immunodeficient mice, we generate a fatal B malignancy with complete penetrance. This humanized-MYC/BCL2-model (hMB) accurately recapitulates the histopathological and clinical aspects of steroid-, chemotherapy- and rituximab-resistant human ‘double-hit’ lymphomas that involve the MYC and BCL2 loci. Notably, this model can serve as a platform for the evaluation of antibody-based therapeutics. As a proof of principle, we used this model to show that the anti-CD52 antibody alemtuzumab effectively eliminates lymphoma cells from the spleen, liver and peripheral blood, but not from the brain. The hMB humanized mouse model underscores the synergy of MYC and BCL2 in ‘double-hit’ lymphomas in human patients. Additionally, our findings highlight the utility of humanized mouse models in interrogating therapeutic approaches, particularly human-specific monoclonal antibodies.

Of mice and men: what rodent models don't tell us

Work recently published in PNAS by The Inflammation and Host Response to Injury, Large Scale Collaborative Research Program1 compared the transcriptional responses in peripheral blood to inflammatory injuries (burns, blunt force trauma) and to endotoxin in human patients and in corresponding mouse models. The results were startling: while the human patients showed similar transcriptional responses to burns, trauma and endotoxemia, the mouse models showed little correlation either to each other or the human response. These differences extended beyond acute gene response and were also apparent in markedly slower recovery from injury, with gene expression not returning after weeks or months in patients as opposed to days in mouse models. The results of this study shed light on an ongoing problem in therapeutic development. Approximately 90% of the drugs reported to have clinical efficacy in high profile journals based on animal models fail in clinical trials.2 Drugs to treat sepsis and acute inflammation have a much worse track record in clinical trials, with all ∼150 drugs trialed to date failing.1 This report highlights a startling example of how the in vivo models used in preclinical drug development can bear little relationship to the human diseases at the transcriptional level. Given the similarity between the human responses to pro-inflammatory stimulation, and the divergent responses in mice it would be easy to conclude that in the ∼75 million years3 since humans and mice shared a common ancestor there has been tremendous divergence in the way in which inflammatory responses are generated and resolved. Indeed, there is strong evidence to support this view in the data mining reported by the authors, who show that systematic regulation of entire pathways associated with inflammatory responses occurs across human inflammation. This coordinated response does not occur across mouse models. The authors then examined other published datasets further strengthening the broad and general nature of their findings. One of the most striking differences this group observed was in the regulation of the Toll-like receptors. These sensors of pathogen-associated molecular patterns exhibit broad transcriptional upregulation in all human inflammation examined, while the changes in expression levels in mice were variable, notably with very little response to endotoxin exposure. This difference could account for relatively greater sensitivity to endotoxic shock and severe pathology seen in human patients, which is absent from mouse models. The authors report upregulation of Toll-like receptor pathway genes following endotoxin exposure, which should result in increased sensitivity to subsequent pathogen-associated molecular patterns encounter, leading to further sensitization. This could result in both a cytokine storm and long-term alterations in the immune system—two of the major challenges of sepsis.4 As well as helping to explain differences in mouse and human responses, the data presented make a compelling case for caution when evaluating mouse models in the development of anti-inflammatory drugs. Medicines act at the molecular level and difference in the transcriptional response between human and mouse could, at the very least, reduce treatment effectiveness in humans. At worst, these differences could lead to unforeseen in vivo side effects. However, there are significant caveats that need to be considered in the context of these results. Firstly, the patients studied were all receiving medical care. This variable combination of changes in activity, diet, medication and environment is not recapitulated in mouse studies and may account for some of the correlation between different human injuries. Secondly, only a single mouse strain, C57BL/6, was assessed and other strains may prove to be better models of human responses at the molecular level. Thirdly, total leukocytes were assessed. The composition of mouse and human blood leukocytes vary significantly—notably with dramatic differences in the circulating levels of neutrophils. These cells are also representative of only a single organ, and changes in gene expression in other tissues such as the liver, kidneys and vascular endothelium are also important for clinical outcome. Finally, this study assesses acute inflammatory conditions and does not speak to the predictive power of mouse models of chronic inflammation or models where inflammation is not a key feature of pathology. Despite these limitations, the results of this study are extremely important, as commonly used models were assessed and found to be profoundly different to the corresponding human condition. This study is the first to systematically address at the molecular level why mouse models that faithfully anatomically mimic human trauma have lead to no clinically effective therapeutics. While mortality from acute sepsis in hospital has fallen from ∼75% to ∼25% this is largely due to improvements in supportive care rather than effective medicines.5 The stark differences in gene expression between patients and mouse models helps to explain why there has been a persistent failure to address this major cause of morbidity. Going forward this study will provide a general template for future evaluation of preclinical data and a starting point for experimental strategies to further understand inflammation. These points should lead to the development of better models, yielding more relevant preclinical data, leading to new therapeutics. Each of these points is worthy of consideration in greater detail. By highlighting the differences in gene expression with respect to induction/repression, magnitude and duration of response between patients and animal models, the authors have provided a new standard for testing models and interpreting drug efficacy results. While future studies will build on this initial work by refining the fractionation of cells and broadening the tissues examined the basic template has now been defined: regulatory authorities should begin asking for gene expression data validating the relevance of in vivo preclinical studies and research funding bodies should consider devoting resources to developing on the datasets presented here to further clarify what particular model systems offer. The combination of the caveats regarding interpretation and the clear differences in gene expression point to several important directions where new model systems are needed or where clarification is vital to understanding the relevance of preclinical data. Firstly, experiments are needed to assess the effects of supportive care and identify which pathways need to be targeted to compliment existing hospital treatment. Secondly, all rodent strains and other animal models need to be assessed in the same way as the common C57BL/6 mouse model was in this report. Directing drug development to use the best existing models or utilizing less widely used strains could yield substantially better outcomes without extensive development and testing. Thirdly, a more nuanced analysis is needed where fractionated cell subtypes are assessed and the differences in expression phenotype are attributed to specific populations rather than the stoichiometry of the mixture of leukocytes isolated. Finally, the methods laid out in this study need to be replicated for models of other conditions, especially where failure rates in clinical trials are high. For diseases such as cancer, these sorts of efforts are already underway. Based on the results of Seok et al., it is also clear that there is an opportunity to develop new models for inflammation. While work continues to develop and refine in vitro models none envisaged to date can recapture the complexity of a living organism. Since human and mouse leukocytes respond so differently, in vivo models where the immune cells more closely resemble those in patients are particularly attractive. For example, humanized mice 6,7 offer many of the advantages of small animal models with the majority of leukocytes being of human origin. However, these cells having developed in a murine environment have their own set of limitations imposed by incompatibilities between graft and host. Primate models offer excellent similarity due to their close evolutionary relationship with humans; this, however, comes at the expense of using outbred animals with far slower generation times and greater housing and care requirements. Whatever models are proposed will have to exceed the existing rodent systems in a rigorous cost/benefit analysis before being adopted. The comparative study of Seok et al. provides a framework for molecular assessment to complement existing pathological and anatomical criteria. In conclusion, this paper highlights the fact that there are major differences in gene expression between patients and existing murine inflammation models. This raises a number of important questions about why the models were so poorly predictive and provides a baseline to assess whether new models come closer to clinical reality. Perhaps the strongest lesson is an old one that is all to often forgotten in the rush of progress: all models are flawed, imperfect representations of reality. They are still useful if their limitations are understood and respected. Seok et al. have highlighted unappreciated weaknesses of a broad class of models and begun a process of adjustment that will hopefully lead to new treatments.

Persistence of tumor-infiltrating CD8 T cells is tumor-dependent but antigen-independent

How tumor-infiltrating lymphocytes (TILs) that are tumor-specific but functionally tolerant persist in the antigen-expressing tumor tissue is largely unknown. We have previously developed a modified TRansgenic Adenocarcinoma of the Mouse Prostate (TRAMP) model where prostate cancer cells express the T-cell epitope SIYRYYGL (SIY) recognized by CD8 T cells expressing the 2C T-cell receptor (TCR) (referred to as TRP-SIY mice). In TRP-SIY mice, activated 2C T cells rapidly become tolerant following infiltration into the prostate tumor. In this study, we show that tolerant 2C T cells persist in the prostate tumor of TRP-SIY mice by proliferating slowly in a tumor-dependent, but antigen-, interleukin (IL)-7- and IL-15-independent manner. We also show that disappearance of 2C T cells from the lymphoid organs of TRP-SIY mice are due to antigen-induced T-cell contraction rather than altered trafficking or generalized T-cell depletion in the mice. Finally, we show that clonal T cells unreactive to SIY are equally capable of persisting in the prostate tumor. These findings suggest that while functional tolerance of TILs is induced by antigen, persistence of tolerant TILs in the tumor tissue is mediated by a novel mechanism: slow proliferation independent of antigen and homeostatic cytokines. These results also allow CD8 T-cell survival in the tumor environment to be compared with T-cell survival in chronic infection.

Interleukins 7 and 15 Maintain Human T Cell Proliferative Capacity through STAT5 Signaling.

T lymphocytes require signals from self-peptides and cytokines, most notably interleukins 7 and 15 (IL-7, IL-15), for survival. While mouse T cells die rapidly if IL-7 or IL-15 is withdrawn, human T cells can survive prolonged withdrawal of IL-7 and IL-15. Here we show that IL-7 and IL-15 are required to maintain human T cell proliferative capacity through the STAT5 signaling pathway. T cells from humanized mice proliferate better if stimulated in the presence of human IL-7 or IL-15 or if T cells are exposed to human IL-7 or IL-15 in mice. Freshly isolated T cells from human peripheral blood lose proliferative capacity if cultured for 24 hours in the absence of IL-7 or IL-15. We further show that phosphorylation of STAT5 correlates with proliferation and inhibition of STAT5 reduces proliferation. These results reveal a novel role of IL-7 and IL-15 in maintaining human T cell function, provide an explanation for T cell dysfunction in humanized mice, and have significant implications for in vitro studies with human T cells.

Abstract B132: A model of de novo acute myeloid leukemia with an autologous human immune system as a preclinical tool for testing immunotherapies

Based on our understanding of the importance of the immune system in controlling tumor progression and the recent successes of immunotherapies for cancer treatment, there is a need to develop small animal models that faithfully recapitulate the human tumor-immune system interaction. These models would serve as preclinical platforms to test immunotherapies and to better understand immune editing of tumors in vivo. Acute Myeloid Leukemia (AML) is a cancer that primarily occurs in older adults who unfortunately, do not tolerate standard chemotherapy and often die of the disease. While new immunotherapies are being developed to treat AML, the lack of suitable preclinical model makes it difficult to accurately test the efficacy and toxicity of these therapies in vivo where the leukemia exists alongside a normal human immune system which includes T cells, Natural Killer cells and macrophages – immune cells that are required to elicit the cytotoxic effects of immunotherapies. Here, we demonstrate the development of a humanized mouse model of de novo Acute Myeloid Leukemia (AML) with an autologous human immune system. By lentiviral mediated expression of a patient-derived mutation of Nucleophosmin (NPM1) in human hematopoietic stem cells (HSCs), followed by engraftment of a mixed pool of transduced and untransduced HSCs in immune-compromised NOD-scid IL2rg-/- (NSG) mice, we have modeled AML which develops alongside an autologous immune system. NPM1 expression is monitored by GFP expression as both proteins are encoded for by the lentiviral construct and expressed in equal stoichiometry. We chose to model the disease with mutant NPM1 as this genetic lesion occurs in ~30% of adult AMLs and 60% of cases with a normal karyotype. The latency of disease is 14-26 weeks with complete penetrance. Tumorigenicity of the transduced leukemic cells was confirmed with secondary transplantation. In this model, the leukemia resembles the broad M2-M5 category of the human disease as per the French-American-British (FAB) classification system, mirroring the pattern demonstrated by patient AMLs with an NPM1 mutation. As is seen in AML patients and in AML xenograft mouse models, the disease in this model is characterized by the presence of blasts in the peripheral blood and bone marrow, anemia, weight loss, splenomegaly and hypocellularity of the bone marrow. Leukemic blasts are primarily CD13 and CD33 positive with low expression of CD14. In the bone marrow, which is the primary site of disease, we detect the presence of leukemic stem cells (LSC), which are hypothesized to seed the disease and are responsible for disease relapse upon conventional chemotherapy. Transcriptome analysis of the CD123+ LSCs demonstrates their stem cell-like expression profile akin to that seen with patient AML LSCs. Additionally, since NPM1 has been shown to be an early mutation in AML and the absence of other lesions results in a favorable prognosis, we sought to find co-operating hits that were driving disease. In summary, we have developed a model of human AML with an autologous human immune system, which recapitulates important features of the human disease. The co-existence of leukemic cells and normal immune cells in this model makes it a useful pre-clinical tool for testing immunotherapies that require functional normal immune cells. Citation Format: Mandeep Kaur, Amanda Hanson, Adam Drake, Ryan Phennicie, Jianzhu Chen. A model of de novo acute myeloid leukemia with an autologous human immune system as a preclinical tool for testing immunotherapies. [abstract]. In: Proceedings of the CRI-CIMT-EATI-AACR Inaugural International Cancer Immunotherapy Conference: Translating Science into Survival; September 16-19, 2015; New York, NY. Philadelphia (PA): AACR; Cancer Immunol Res 2016;4(1 Suppl):Abstract nr B132.