Regulatory mechanisms driving the random monoallelic expression of the natural killer cell receptor genes

AbstractRegulatory mechanisms driving the random monoallelic expression of the natural killer cell receptor genesbyDjem U. KissiovDoctor of Philosophy in Molecular and Cell BiologyUniversity of California, Berkeley,Professor David H. Raulet, Chair Natural killer (NK) cells constitute the first line of defense against many foreign pathogens and cancerous cells. Unlike T and B lymphocytes, NK cells do not rearrange their receptor genes and instead generate diversity for MHC I ligands by drawing on a pool of germline-encoded receptors in a stochastic fashion. These receptors are encoded by C-type lectin domain-containing genes in a tandem array on mouse chromosome 6, and are expressed in a random, monoallelic and mitotically stable pattern (RME). Genes are generally transcribed from both alleles, but in recent years RME has arisen as a notable exception and may describe up to ~10% of genes. While progress has been made toward the understanding of the prevalence and phenomenology of RME, the driving mechanisms are not understood. Research has been hampered by the lack of an established in vivo genetic model to dissect the role of specific regulatory elements in RME expression patterns. NK cell receptors provide the opportunity to generate such a model. NK cell receptors are regulated proximally, greatly simplifying the search for the relevant regulatory elements as they should occur near the gene locus itself. Furthermore, our lab has previously developed allele-specific antibodies, allowing the assessment of allelic expression on single cells rapidly and with a high degree of confidence by flow cytometry, circumventing the technical challenges, costs and time associated with experiments such as single cell RNA-seq. Finally, primary NK cells are readily cultured in vitro in medium containing IL-2 such that questions about mitotic stability of expression states are easily addressed. Deletion of enhancers in vivo has been greatly simplified with the advent of CRISPR/Cas9-based genome editing. Germline enhancer deletion in mice can now be achieved on timescales of months rather than years, and much more reliably than by traditional gene targeting methods. Additionally, analysis of the chromatin states of silent and active alleles has been historically limited by the requirement of large numbers of cells. Recent advances in chromatin profiling technologies (ATAC-seq and CUT&RUN) allow experiments to be performed with tens of thousands of cells rather than tens of millions, allowing the profiling of subsets of NK cells sorted with respect to allelic expression status using allele-specific antibodies. Using the power and flexibility provided by these new approaches, this thesis addresses the following questions. First, what is the role of enhancer elements in regulating the expression frequencies of the variegated NK receptor genes? Chapter 3 addresses this question through a series of enhancer deletions and F1 hybrid genetics in vivo. Furthermore, in Chapter 3 I leverage the power of allele-specific antibodies and flow cytometry to search for RME expression patterns in genes previously thought to be ubiquitously expressed by a given cell type. Strikingly, RME-like expression patterns are identified in all assayed receptors: NKG2D by NK cells, CD45 by T cells and B cells, CD8 by cytotoxic T cells, and Thy1 by both cytotoxic and helper T cells. This supports a model where RME is the consequence of generalized stochastic properties of gene expression and can be detected in many and perhaps all genes. Next, this thesis addresses the chromatin features of both silent and active NK receptor gene alleles in vivo and deduces clues as to the mechanism of mitotic stability in RME. Chapter 4 discusses the results of chromatin analyses in sorted primary cells from F1 hybrid mice. Additionally, this chapter addresses the role of enhancer elements in the maintenance of active RME alleles. The data presented in this thesis results in a unified model of RME and enhancer function derived from the broadly probabilistic properties of gene expression. Enhancers display constitutive activation but only probabilistic effects on target gene expression, suggesting enhancer activation is generally decoupled from target gene activation. Deletion of individual enhancers from a set of enhancers regulating a target gene results in a reduction of the proportion of cells expressing the gene at both the Ly49g and Nkg2d loci. A particularly notable result is the transformation of the (apparently) ubiquitously expressed Nkg2d gene to a stable RME gene via enhancer deletion, displaying all the fundamental properties of the natural RME NK receptor genes. These deletions had no large effect on the expression level of these genes per cell. These results strongly support the binary on/off model of enhancer action. Rather than being specific to a set of genes, stochastic allele activation appears to be a general property of gene expression and is not restricted to a biologically meaningful set of genes. Surprisingly, silent NK receptor alleles lack a repressive chromatin state, and more closely resemble the chromatin of lineage non-specific genes. Mitotic stability in RME is likely a result of allelic classification as lineage appropriate or lineage non-specific by stochastic enhancer action. We propose that previously documented examples of RME are extreme manifestations of a general property, rather than a result of a dedicated mechanism. RME, therefore, does not seem to be an exception to the rules and instead describes gene expression broadly. This model of gene expression conceptually resembles a bistable multivibrator, in which an initial signal determines one of two possible states which are then maintained in the absence of the initial signal. In the context of developmental gene regulation, this signal is presumably provided during cellular differentiation, and may be provided to inducible genes in fully differentiated cells. Importantly, gene induction (whether developmental or through some stimulatory signal) is read out by the varying strength of enhancer activity as a rheostat, raising or lowering the probability of stable allelic activation