Harvard Stem Cell Institute

Vaccinia-related kinase 1 (VRK1) is a promising therapeutic target in gliomas and glioblastomas where VRK2 is silenced by promoter methylation, rendering VRK1 essential for accurate nuclear envelope reassembly following mitosis. Small-molecule ATP-site drug discovery for VRK1 has been hindered by the absence of robust and reproducible biochemical assays. Through virtual screening, we identified previously unreported VRK1-binding scaffolds and validated them in biochemical kinase assays, yielding an 82 nM inhibitor with high selectivity for VRK1 over VRK2. During characterization of this compound, we found that a commonly used commercial time-resolved fluorescence resonance energy transfer VRK1 activity assay is dependent on purification tag-mediated VRK1 dimerization. Leveraging the new inhibitor, we developed fluorogenic tool compounds that increase in fluorescence intensity upon binding to the active site of VRK1 and do not require artificial dimerization of VRK1. The top probe exhibits a Kd of 180 nM and is useful for ligand displacement assays using both fluorescence enhancement and time-resolved fluorescence resonance energy transfer readouts. Together, these results introduce new chemical scaffolds for targeting VRK1, define an assay artifact that has complicated VRK1 inhibitor discovery, and deliver fluorogenic tool compounds for high-throughput screening of ATP-site VRK1 inhibitors, enabling future drug discovery efforts against this emerging cancer vulnerability.

Finding good branching orders is key to solving SAT problems efficiently, but finding such branching orders is a difficult problem. Using a learning based approach to predict a good branching order before solving, therefore, has potential. In this paper, we investigate predicting branching orders using graph neural networks as a preprocessing step to conflict-driven clause learning (CDCL) SAT solvers. We show that there are significant gains to be made in existing CDCL SAT solvers by providing a good initial branching. Further, we provide three labeling methods to find such initial branching orders in a tractable way. Finally, we train a graph neural network to predict these branching orders and show through our evaluations that a GNN-initialized ordering yields significant speedups on random 3-CNF and pseudo-industrial benchmarks, with generalization capabilities to instances much larger than the training set. However, we also find that the predictions fail at speeding up more difficult and industrial instances. We attribute this to the solver's dynamic heuristics, which rapidly overwrite the provided initialization, and to the complexity of these instances, making GNN prediction hard.

At present, there are no FDA-approved orally-available bone anabolic agents to treat osteoporosis. PTH stimulates bone formation through an intracellular signaling cascade that involves the inhibition of salt-inducible kinase (SIK) isoforms 2 and 3. Therefore, direct small molecule SIK2/SIK3 inhibitors may represent a strategy to mimic PTH actions to treat different forms of osteoporosis. We previously described the synthesis and characterization of SK-124, a pharmacologic SIK2/SIK3 inhibitor that increases trabecular bone formation in eugonadal mice. However, the efficacy of this agent in osteoporosis mouse models remains unknown. Hypogonadism is an important cause of age-related bone loss. In this study, we investigated the therapeutic potential of SK-124 in a male hypogonadal bone loss model (orchiectomy, ORX) in BALB/c mice. Radiographic and histological analyses revealed that SK-124-treated ORX mice showed reduced bone loss compared to the vehicle-treated ORX mice. Serum bone turnover markers demonstrated that SK-124 treatment increased bone turnover, suggesting that SK-124 acts in a PTH-like manner in ORX mice. Bone RNA-sequencing analysis demonstrated novel pathways associated with increased bone formation in response to SK-124 treatment. These findings indicate that SK-124 prevents bone loss in a hypogonadal bone loss model and holds potential as an orally available therapeutic for treating osteoporosis due to testosterone deficiency.

Building on the identification of ABCB5 as a marker of limbal stem cells (LSCs), this study examines CD63, a newly identified molecule co-expressed with ABCB5 in limbal epithelial cells, to define its role in maintaining corneal epithelial cell identity. RNA sequencing (RNA-seq) was performed on flow cytometry-sorted Abcb5-positive and Abcb5-negative murine corneal epithelial cells. CD63 expression in human corneal tissue was assessed by immunostaining. CD63 was silenced in cultured human limbal epithelial cells using siRNA-mediated knockdown and resulting molecular and cellular changes were analyzed by qRT-PCR, flow cytometry, RNA-seq, Western blotting, and cell proliferation assays. RNA-seq analysis revealed increased expression of LSC markers, including Krt15, Krt6b, Fgfr1, Gpha2, Ifitm3, Ifitm1, and Cd63, and decreased expression of differentiation-associated markers, such as Krt12, Gja1, and Ovol1 in Abcb5-positive cells. Immunostaining of human corneal tissue demonstrated strong CD63 expression localized to the limbal region. Knockdown of CD63 in cultured human limbal epithelial cells resulted in reduced cell proliferation and significantly decreased expression of corneal epithelium-enriched genes, including KRT12, CLU, ALDH1A1, ALDH3A1, TGFBI, and MYEOV. Notably, CD63 knockdown led to an approximately 50% reduction in expression of PAX6, a key transcriptional regulator of corneal epithelial identity. CD63 is highly expressed in the human limbus and is required for maintaining cell proliferation and the expression of corneal epithelium-specific proteins, likely through regulation of PAX6. These findings establish CD63 as a functionally important component of limbal stem cell biology and a key contributor to corneal epithelial homeostasis.

Cervical cancer (CVC) is classically understood as an HPV-driven disease in which high-risk HPV infects cervical basal epithelial cells, induces neoplastic transformation, and drives upward epithelial expansion, pathogenic events underlying the success of cytology- and HPV-based screening. With widespread HPV vaccination and screening, the WHO has launched the CVC elimination initiative and believes that CVC will be the first cancer that would be eliminated as a public health problem. However, statistics showed that CVC remains the most common gynecologic malignancy worldwide, and in the United States, where HPV testing and Pap smears have reached near-maximal implementation, the CVC mortality rate has plateaued for more than two decades. These epidemiologic trends suggest the existence of a subset of CVC that can escape current screening strategies. Here, we demonstrate that hyperactivation of YAP1 caused by disruption of Hippo-YAP signaling is sufficient to induce a subtype of HPV-independent invasive CVC that lacks surface lesions and therefore evades HPV- and cytology-based detection. Using single-cell RNA sequencing combined with high-resolution spatial transcriptomics, we define the cellular architecture, molecular pathways, and immune microenvironment underlying this invasive subtype. We show that YAP1-driven tumors adopt an EMT-high transcriptional state and selectively recruit immunosuppressive myeloid-derived suppressor cells that functionally interact with cancer cells to promote invasion and progression. Our findings suggest that targeting the disrupted Hippo signaling may offer a new strategy to prevent the subset of invasive CVC that current HPV- and Pap-based programs cannot detect, an essential step toward achieving global CVC elimination.

Lysosomal membrane integrity is essential for cellular homeostasis, and its failure drives lysosomal storage disorders (LSD) and neurodegeneration. The dipeptide L-leucyl-L-leucine methyl ester (LLOMe) is widely used to model lysosomal damage, yet its mechanism remains poorly understood. The prevailing view holds that LLOMe polymerizes into membrane-permeabilizing peptide chains within the lysosomal lumen. Using cryo-electron tomography in cultured cells and primary neurons, we visualized the structural basis of LLOMe-induced lysosomal damage. We reveal that LLOMe forms amyloid structures within lysosomes that directly interact with and rupture the limiting membrane through mechanical stress. In vitro reconstitution confirms this amyloid-mediated mechanism. These findings establish a structural paradigm for lysosomal membrane disruption and provide insights into how disease-relevant protein aggregates, implicated in neurodegeneration and LSD, may compromise lysosomal integrity.

Chronic inflammation promotes aging and age-associated diseases. While metabolic interventions can modulate inflammation, how metabolism and inflammation are connected remains unclear. Cytoplasmic chromatin fragments (CCFs) drive chronic inflammation through the cGAS-STING pathway in senescence and aging. However, CCFs are larger than nuclear pores, and how they translocate from the nucleus to the cytoplasm remains uncharacterized. Here we report that chromatin fragments exit the nucleus via nuclear egress, a membrane trafficking process that shuttles large complexes across the nuclear envelope. Inactivating critical nuclear egress proteins, the ESCRT-III or Torsin complex, traps chromatin fragments at the nuclear membrane and suppresses cGAS-STING activation and senescence-associated inflammation. Glucose limitation or metformin inhibits CCF formation through AMPK-dependent phosphorylation and autophagic degradation of ALIX, an ESCRT-III component. In aged mice, metformin reduces ALIX, CCFs, and cGAS-mediated inflammation in the intestine. Our study identifies a mechanism linking metabolism and inflammation and suggests targeting the nuclear egress of chromatin fragments as a strategy to suppress age-associated inflammation.