I. Metabolic State of Mouse Embryonic Stem Cells:

By surveying the abundance of common metabolites in mouse embryonic (ES) cells as a function of differentiation, it has been observed that metabolites associated with one carbon metabolism change radically when undifferentiated ES cells are cued to differentiate into embryoid bodies.  These observations led to the discovery that ES cells are uniquely dependent upon threonine catabolism.  Jian Wang and Peter Alexander have found that undifferentiated ES cells express the threonine dehydrogenase (TDH) enzyme at a 1,000-fold higher level than any other cell line or mouse tissue tested.  TDH is a mitochondrial enzyme that catabolizes threonine into glycine and acetyl-CoA.  The former metabolite is further catabolized by the mitochondrial glycine cleavage system in order to charge tetrahydrofolate (THF) into methyl-THF for use in purine biosynthesis and other forms of one carbon metabolism. The latter metabolite, acetyl-CoA, can be used to feed the tricarboxylic acid (TCA) cycle for the generation of ATP.  The first chapter of this story was published in the summer of 2009 (Wang, Alexander, etc., 2009, Science).    Evolution of this project will now turn towards the possibility that the exceptionally high levels of acetyl-CoA in undifferentiated stem cells might facilitate acetylation of certain target proteins instrumental for the unusually rapid cell division cycle of ES cells. These cells divide once every 4-5 hours, which is a more rapid rate of proliferation than even the fastest growing of all cancer cells.  Efforts have been made to tailor strains of mice wherein the TDH gene can be conditionally eliminated via Cre-mediated recombination, such that it might be possible to rigorously test the necessity of TDH enzyme requirement in the inner cell mass of the blastocyst embryo.  Finally, a high throughput screen of the UTSWMC compound file of 250,000 drug-like chemicals have led to the discovery of potent and specific inhibitors of the mouse TDH enzyme.  These inhibitors are selectively toxic to mouse ES cells, presumably due to their ability to block the activity of the TDH enzyme which is critical for maintaining the unusual metabolic state of mouse ES cells (Alexander et al., 2011, PNAS).

II. Metabolic Cycles:

During the 2004-2008 timeframe the McKnight lab, in efforts primarily led by Benjamin Tu and Jake Chen, studied the growth of a prototrophic yeast strain under nutrient-limiting conditions in a chemostat.  Quite remarkably, the cells enter into a self-synchronized metabolic cycle 4-5 hours in length.  During this metabolic cycle yeast cells fluctuate rhythmically back and forth between respirative and glycolytic metabolism.  DNA microarray studies have shown that over half of the genes in the yeast genome are selectively expressed during specific windows of this yeast metabolic cycle (YMC).  Studies of the grouping of genes that become activated at specific times of the YMC have offered a simple logic to explain how and why the cells are able to perform specific functions at specific times (Tu, Kudlicki, Rowicka and McKnight, 2005, Science).  Moving beyond gene expression studies, Benjamin Tu – in collaboration with Ted Young and his colleagues at the University of Washington – studied fluctuation in the abundance of hundreds of metabolites as a function of the YMC by use of various forms of liquid and gas chromatography and mass spectrometry.  These observed patterns of metabolite fluctuation largely conformed to predictions coming from patterns of gene expression (Tu, Mohler, Liu, Dombek, Young, Synovec and McKnight, 2006, PNAS).

One of the observations made by Tu and Chen was that yeast cells are only able to divide during one of the three defined phases of the YMC.  This phase, designated the reductive building (RB) phase, is when cells cease mitochondrial respiration and enhance the rate of glycolytic metabolism.  It was hypothesized that DNA synthesis is restricted to the RB phase as a means of protecting genome integrity.  This idea was prosecuted in studies of yeast strains bearing mutations that impede progress through the cell division cycle.  Slow-growing strains were found to display correlative reductions in the length of the YMC.  Wild-type yeast cells, when grown in rich culture medium, double in growth about once per hour and display a YMC lasting roughly 4 hours.  Strains missing genes encoding proteins important for cell cycle progression, such as swi6 or sic1, double much more slowly and display YMC’s of roughly two and one hours respectively.  It was found that the latter strains are forced to undergo DNA replication in all phases of the YMC, including the oxidative phase when mitochondrial respiration is taking place.  Analysis of the swi6 and sic1 mutants led to the discovery that they suffer considerably higher rates of spontaneous mutation when grown in the chemostat, but not when grown in glucose-rich culture medium (Chen, Odstrcil, Tu and McKnight, 2007, Science ).  Such observations are consistent with the hypothesis that wild type yeast know how to restrict DNA synthesis and cell division to the reductive phase of the YMC as a means of protecting genome integrity.     

Benjamin Tu has inherited the YMC project as an independent Assistant Professor of Biochemistry here at UT Southwestern Medical Center.  The McKnight lab – in efforts spearheaded by Jake Chen and Shanhai Xie – have turned to mammalian systems in search of phenomena related to the YMC.  Ongoing experiments have given evidence of a metabolic cycle of circadian dimensions in the brain tissue of adult mice.  Using liquid chromatography/mass spectrometry, Drs. Chen and Xie quantitated the abundance of dozens of metabolites surveyed at four hour intervals for mice exposed to running wheels and kept on a 12:12hr light:dark cycle.  A number of metabolites involved in sulfur flux, including S-adenosylmethinonie (SAM), S-adneosylhomocysteine (SAH) and cystothione fluctuated in abundance as a function of the day:night cycle.  Intriguingly, the metabolites that fluctuate in abundance most robustly in the yeast metabolic cycle correspond closely with those that fluctuate most robustly in the brain tissue of adult mice.  By subjecting animals to sleep deprivation, it has been observed that metabolite fluctuation is driven by sleep, not by circadian rhythm.  We do not yet understand the significance of these observations, but are encouraged to consider the idea that the metabolic state of neurons may be dynamic.  If so, it is possible that neurons are able to perform certain tasks more optimally at certain times of the day:night and/or sleep:wake cycle than others.

III. Discovery of Pro-Neurogenic Chemicals:

No substantive therapeutics are available for the treatment of almost any form of disease entailing nerve cell death. Patients suffering from any of a wide spectrum of neurodegenerative diseases, including Parkinson’s disease, Alzheimer’s disease, amyotrophic lateral sclerosis, fronto-temporal dementia, and Huntington’s disease are condemned to progressive demise of the CNS by virtue of nerve cell death. It is likewise the case that no effective treatments exist for injuries to the brain or peripheral nervous system, including traumatic or concussive brain injury, spinal cord injury, or peripheral nerve injury. Any chemical having the capacity to safely impede nerve cell death in the context of these varied diseases or injuries would offer the opportunity for transformative impact in modern medicine. Previously, we performed a target-agnostic, in vivo screen in search of chemicals that might enhance hippocampal neurogenesis in adult mice (Pieper et al., 2010, Cell). The screen was simple in concept. We selected 1,000 drug-like chemicals from the 250,000 compounds in the University of Texas Southwestern Medical Center (UTSWMC) high-throughput screening center. The compounds were selected to preserve chemical diversity, enhance the representation of chiral molecules, and avoid untoward chemical properties such as reactive moieties. The 1,000 molecules were randomly pooled into groups of ten, and each pool was administered directly into the left ventricle of two adult mice. Intracranial delivery was facilitated by stereotactic positioning of a canula fed directly by an Alzet minipump containing the mixture of ten chemicals. The drug mixture was administered over a week-long period at concentrations anticipated to deliver mid-nanomolar levels of the ten test compounds. Daily injections of the thymidine analog, bromodeoxyuridine (BrdU), were coadministered in order to monitor the formation of new hippocampal nerve cells. Following compound administration, animals were sacrificed such that brain tissue could be recovered, sectioned, and stained with antibodies to BrdU. This 2-year screen led to the discovery of a handful of pools that enhanced neurogenesis in both test mice that had been exposed to the pool.

Breakdown of the active pools allowed the individual testing of each of the ten chemicals in the pool, leading to the discovery of eight proneurogenic compounds (Pieper et al., 2010, Cell). Among the eight proneurogenic chemicals, pharmacological testing gave evidence that only one of the compounds had favorable pharmacological properties. Pool seven (P7) contained an aminopropyl carbazole as its active, third compound (C3). When administered to mice via intraperitoneal, intravenous, or oral routes, the P7C3 compound revealed favorable half-life, volume of distribution, and brain penetration. It was also found that P7C3 could be safely administered to mice and rats for prolonged periods at concentrations well above those required to stimulate hippocampal neurogenesis, giving evidence that the molecule was not overtly toxic to rodents. Although it was initially anticipated that proneurogenic compounds would act by stimulating the mitotic birth of newborn nerve cells in the subgranular layer of the dentate gyrus, P7C3 revealed no such activity. The 2-fold enhanced level of BrdU-positive neurons observed over a week-long dose of P7C3 was absent when animals were pulsed with BrdU for only 24 hr. More strikingly, when BrdU was pulsed for only 1 day and animals were subsequently administered P7C3 for 1 month, we observed a far larger enhancement in BrdU-positive hippocampal neurons (500%). Instead of stimulating the mitotic division of neuronal stem cells, these observations gave evidence that P7C3 mitigated the death of newborn neurons. Under the conditions of our study, only 10%–20% of newborn neurons survive the month-long differentiation process to become properly wired hippocampal neurons. Prolonged administration of P7C3 significantly mitigated the death of newborn neurons, such that upward of half of the cells survive the month-long “differentiation gauntlet” taking place between stem cell mitosis and terminal nerve cell differentiation.

Having discovered the aminopropyl carbazole chemical in an unbiased, in vivo screen and having found that it protects newborn neurons from death, we used methods of medicinal chemistry to improve the potency and pharmacological properties of P7C3 (MacMillan et al., 2011, JACS, Naidoo et al., 2014, J Med Chem, and Pieper et al., 2014, Chem Soc Rev). Using these improved, active derivatives of P7C3, we have observed neuroprotective activity in animal models of Parkinson’s disease (De Jesús-Cortés et al., 2012, PNAS), amyotrophic lateral sclerosis (Tesla et al., 2012, PNAS), as well as concussive injury to the rodent brain (Yin et al., 2014, Cell Rep). To uncover P7C3's mechanism of action, an active derivative of P7C3 was modified to contain both a benzophenone for photocrosslinking and an alkyne for CLICK chemistry. This derivative was found to bind nicotinamide phosphoribosyltransferase (NAMPT), the rate-limiting enzyme involved in the conversion of nicotinamide into nicotinamide adenine dinucleotide (NAD). Administration of active P7C3 chemicals to cells treated with doxorubicin, which induces NAD depletion, led to a rebound in intracellular levels of NAD and concomitant protection from doxorubicin-mediated toxicity. Active P7C3 variants likewise enhanced the activity of the purified NAMPT enzyme, providing further evidence that they act by increasing NAD levels through its NAMPT-mediated salvage (Wang, Han, et al., 2014, Cell).

IV:  Low Complexity Domains, Interacellular Puncta and a Solid State Conceptualization of Information Flow from Gene to Message to Protein:

Quite by accident the McKnight lab discovered that a biotinylated isoxazole (b-isox) chemical crystallizes when exposed to cold aqueous buffer. The molecular surface of these crystals displays a repetitive array of peaks and troughs, with the troughs being 4.6Å in width. When exposed to cell lysates, the b-isox crystals co-precipitate hundreds of regulatory proteins associated with transcription and RNA biogenesis. We have come to understand that regulatory proteins are co-precipitated by virtue of the fact that their disordered, low complexity domains can transition from random coil to extended ß-sheets that fit perfectly into the 4.6Å troughs of the b-isox crystals. Proceeding from this serendipitous discovery, we initiated detailed studies on the low complexity domains associated with RNA and DNA regulatory proteins. When expressed as either mCherry or GFP fusion proteins, these low complexity domains polymerize into homogenous fibers. As a result, solutions incubated at a high protein concentration gel in a manner analogous to what happens when soluble actin protein is allowed to polymerize. These small gel droplets have been turned into a simple assay in order to study what proteins and RNAs might bind to our hydrogel droplets. As reported in two back-to-back Cell papers published in 2012, (Kato et al., 2012, Cell; Han et al., 2012, Cell) we believe that what we are studying in these polymerization reactions may be at the heart of various forms of intracellular puncta that are not membrane invested, including nuclear speckles, RNA granules and transcription factories.

More recently we have found that the C-terminal domain (CTD) of RNA polymerase, a disordered region containing 52 heptad repeats of the sequence YSPTSPS, binds to hydrogel droplets composed of the LC domains of any of the three FET proteins (Fused in sarcoma, Ewings sarcoma, or TAF15). This binding reaction is reversed by cyclin dependent kinase mediated phosphorylation of the CTD (Kwon etal, 2013, Cell). We speculate that the process of information transfer from gene (transcription factories) to messenger RNA (nuclear speckles) to the eventual sites of protein synthesis (via RNA granules) may represent a continuous, “cradle to grave” relay through fibrous polymers. Most recently, we found that serine:arginine (SR) repeats assciated with regulatory proteins controlling pre-messenger RNA splicing bound hydrogel droplets composed of fibrous polymers of the low-complexity domain of heterogeneous ribonucleoprotein A2 (hnRNPA2). Hydrogel binding was reversed upon phosphorylation of the SR domain by CDC2-like kinases 1 and 2 (CLK1/2). Mutated variants of the SR domains changing serine to glycine (SR-to-GR variants) also bound to hnRNPA2 hydrogels but were not affected by CLK1/2. When expressed in mammalian cells, these variants bound nucleoli.

The translation products of the sense and antisense transcripts of the expansion repeats associated with the C9orf72 gene altered in neurodegenerative disease encode GRn and PRn repeat polypeptides. Both peptides bound to hnRNPA2 hydrogels independent of CLK1/2 activity. When applied to cultured cells, both peptides entered cells, migrated to the nucleus, bound nucleoli, and poisoned RNA biogenesis, which caused cell death. This most recent study was published in Science (Kwon et al., 2014, Science).