Lkb1 orchestrates γδ T-cell metabolic and functional fitness to control IL-17-mediated autoimmune hepatitis.

γδ T cells play a crucial role in immune surveillance and serve as a bridge between innate and adaptive immunity. However, the metabolic requirements and regulation of γδ T-cell development and function remain poorly understood. In this study, we investigated the role of liver kinase B1 (Lkb1), a serine/threonine kinase that links cellular metabolism with cell growth and proliferation, in γδ T-cell biology. Our findings demonstrate that Lkb1 is not only involved in regulating γδ T lineage commitment but also plays a critical role in γδ T-cell effector function. Specifically, T-cell-specific deletion of Lkb1 resulted in impaired thymocyte development and distinct alterations in γδ T-cell subsets in both the thymus and peripheral lymphoid tissues. Notably, loss of Lkb1 inhibited the commitment of Vγ1 and Vγ4 γδ T cells, promoted the maturation of IL-17-producing Vγ6 γδ T cells, and led to the occurrence of fatal autoimmune hepatitis (AIH). Notably, clearance of γδ T cells or blockade of IL-17 significantly attenuated AIH. Mechanistically, Lkb1 deficiency disrupted metabolic homeostasis and AMPK activity, accompanied by increased mTORC1 activation, thereby causing overactivation of γδ T cells and enhanced apoptosis. Interestingly, activation of AMPK or suppression of mTORC1 signaling effectively inhibited IL-17 levels and attenuated AIH in Lkb1-deficient mice. Our findings highlight the pivotal role of Lkb1 in maintaining the homeostasis of γδ T cells and preventing IL-17-mediated autoimmune diseases, providing new insights into the metabolic programs governing the subset determination and functional differentiation of thymic γδ T cells.

SLAPP: Subgraph-level attention-based performance prediction for deep learning models.

The intricacy of the Deep Learning (DL) landscape, brimming with a variety of models, applications, and platforms, poses considerable challenges for the optimal design, optimization, or selection of suitable DL models. One promising avenue to address this challenge is the development of accurate performance prediction methods. However, existing methods reveal critical limitations. Operator-level methods, proficient at predicting the performance of individual operators, often neglect broader graph features, which results in inaccuracies in full network performance predictions. On the contrary, graph-level methods excel in overall network prediction by leveraging these graph features but lack the ability to predict the performance of individual operators. To bridge these gaps, we propose SLAPP, a novel subgraph-level performance prediction method. Central to SLAPP is an innovative variant of Graph Neural Networks (GNNs) that we developed, named the Edge Aware Graph Attention Network (EAGAT). This specially designed GNN enables superior encoding of both node and edge features. Through this approach, SLAPP effectively captures both graph and operator features, thereby providing precise performance predictions for individual operators and entire networks. Moreover, we introduce a mixed loss design with dynamic weight adjustment to reconcile the predictive accuracy between individual operators and entire networks. In our experimental evaluation, SLAPP consistently outperforms traditional approaches in prediction accuracy, including the ability to handle unseen models effectively. Moreover, when compared to existing research, our method demonstrates a superior predictive performance across multiple DL models.

Efficient and accurate compound scaling for convolutional neural networks.

Designing efficient and accurate network architectures to support various workloads, from servers to edge devices, is a fundamental problem as the use of Convolutional Neural Networks (ConvNets) becomes increasingly widespread. One simple yet effective method is to scale ConvNets by systematically adjusting the dimensions of the baseline network, including width, depth, and resolution, enabling it to adapt to diverse workloads by varying its computational complexity and representation ability. However, current state-of-the-art (SOTA) scaling methods for neural network architectures overlook the inter-dimensional relationships within the network and the impact of scaling on inference speed, resulting in suboptimal trade-offs between accuracy and inference speed. To overcome those limitations, we propose a scaling method for ConvNets that utilizes dimension relationship and runtime proxy constraints to improve accuracy and inference speed. Specifically, our research notes that higher input resolutions in convolutional layers lead to redundant filters (convolutional width) due to increased similarity between information in different positions, suggesting a potential benefit in reducing filters while increasing input resolution. Based on this observation, the relationship between the width and resolution is empirically quantified in our work, enabling models with higher parametric efficiency to be prioritized through our scaling strategy. Furthermore, we introduce a novel runtime prediction model that focuses on fine-grained layer tasks with different computational properties for more accurate identification of efficient network configurations. Comprehensive experiments show that our method outperforms prior works in creating a set of models with a trade-off between accuracy and inference speed on the ImageNet datasets for various ConvNets.

METTL3-mediated m6A methylation orchestrates mRNA stability and dsRNA contents to equilibrate γδ T1 and γδ T17 cells.

γδ T cells make key contributions to tissue physiology and immunosurveillance through two main functionally distinct subsets, γδ T1 and γδ T17. m6A methylation plays critical roles in controlling numerous aspects of mRNA metabolism that govern mRNA turnover, gene expression, and cellular functional specialization; however, its role in γδ T cells remains less well understood. Here, we find that m6A methylation controls the functional specification of γδ T17 vs. γδ T1 cells. Mechanistically, m6A methylation prevents the formation of endogenous double-stranded RNAs and promotes the degradation of Stat1 transcripts, which converge to prevent over-activation of STAT1 signaling and ensuing inhibition of γδ T17. Deleting Mettl3, the key enzyme in the m6A methyltransferases complex, in γδ T cells reduces interleukin-17 (IL-17) production and ameliorates γδ T17-mediated psoriasis. In summary, our work shows that METTL3-mediated m6A methylation orchestrates mRNA stability and double-stranded RNA (dsRNA) contents to equilibrate γδ T1 and γδ T17 cells.

Transcription factor TOX maintains the expression of Mst1 in controlling the early mouse NK cell development.

Rationale: TOX is a DNA-binding factor required for the development of multiple immune cells and the formation of lymph nodes. However, the temporal regulation mode of TOX on NK cell development and function needs to be further explored. Methods: To investigate the role of TOX in NK cells at distinct developmental phases, we deleted TOX at the hematopoietic stem cell stage (Vav-Cre), NK cell precursor (CD122-Cre) stage and late NK cell developmental stage (Ncr1-Cre), respectively. Flow cytometry was used to detect the development and functional changes of NK cell when deletion of TOX. RNA-seq was used to assess the differences in transcriptional expression profile of WT and TOX-deficient NK cells. Published Chip-seq data was exploited to search for the proteins directly interact with TOX in NK cells. Results: The deficiency of TOX at the hematopoietic stem cell stage severely retarded NK cell development. To a less extent, TOX also played an essential role in the physiological process of NKp cells differentiation into mature NK cells. Furthermore, the deletion of TOX at NKp stage severely impaired the immune surveillance function of NK cells, accompanied by down-regulation of IFN-γ and CD107a expression. However, TOX is dispensable for mature NK cell development and function. Mechanistically, by combining RNA-seq data with published TOX ChIP-seq data, we found that the inactivation of TOX at NKp stage directly repressed the expression of Mst1, an important intermediate kinase in Hippo signaling pathway. Mst1 deficient at NKp stage gained the similar phenotype with Toxfl/flCD122Cre mice. Conclusion: In our study, we conclude that TOX coordinates the early mouse NK cell development at NKp stage by maintaining the expression of Mst1. Moreover, we clarify the different dependence of the transcription factor TOX in NK cells biology.

mTORC1 signaling pathway integrates estrogen and growth factor to coordinate vaginal epithelial cells proliferation and differentiation.

The mouse vaginal epithelium cyclically exhibits cell proliferation and differentiation in response to estrogen. Estrogen acts as an activator of mTOR signaling but its role in vaginal epithelial homeostasis is unknown. We analyzed reproductive tract-specific Rptor or Rictor conditional knockout mice to reveal the role of mTOR signaling in estrogen-dependent vaginal epithelial cell proliferation and differentiation. Loss of Rptor but not Rictor in the vagina resulted in an aberrant proliferation of epithelial cells and failure of keratinized differentiation. As gene expression analysis indicated, several estrogen-mediated genes, including Pgr and Ereg (EGF-like growth factor) were not induced by estrogen in Rptor cKO mouse vagina. Moreover, supplementation of EREG could activate the proliferation and survival of vaginal epithelial cells through YAP1 in the absence of Rptor. Thus, mTORC1 signaling integrates estrogen and growth factor signaling to mediate vaginal epithelial cell proliferation and differentiation, providing new insights into vaginal atrophy treatment for post-menopausal women.

The kinase PDK1 regulates regulatory T cell survival via controlling redox homeostasis.

Rationale: Regulatory T cells (Treg cells) play an important role in maintaining peripheral tolerance by suppressing over-activation of effector T cells. The kinase PDK1 plays a pivotal role in conventional T cell development. However, whether PDK1 signaling affects the homeostasis and function of Treg cells remains elusive. Methods: In order to evaluate the role of PDK1 in Treg cells from a genetic perspective, mice carrying the floxed PDK1 allele were crossbred with Foxp3Cre mice to efficiently deleted PDK1 in Foxp3+ Treg cells. Flow cytometry was used to detect the immune cell homeostasis of WT and PDK1fl/flFoxp3Cre mice. RNA-seq was used to assess the differences in transcriptional expression profile of WT and PDK1-deficient Treg cells. The metabolic profiles of WT and PDK1-deficient Treg cells were tested using the Glycolysis Stress Test and Mito Stress Test Kits by the Seahorse XFe96 Analyser. Results: PDK1 was essential for the establishment and maintenance of Treg cell homeostasis and function. Disruption of PDK1 in Treg cells led to a spontaneous fatal systemic autoimmune disorder and multi-tissue inflammatory damage, accompanied by a reduction in the number and function of Treg cells. The deletion of PDK1 in Treg cells destroyed the iron ion balance through regulating MEK-ERK signaling and CD71 expression, resulting in excessive production of intracellular ROS, which did not depend on the down-regulation of mTORC1 signaling. Inhibition of excessive ROS, activated MEK-Erk signaling or overload Fe2+ could partially rescue the survival of PDK1-deficient Treg cells. Conclusion: Our results defined a key finding on the mechanism by which PDK1 regulates Treg cell survival via controlling redox homeostasis.

Examining two sets of introgression lines reveals background-independent and stably expressed QTL that improve grain appearance quality in rice (Oryza sativa L.).

KEY MESSAGE: A novel QTL cluster for appearance quality on Chr07 was identified using reciprocal introgression populations in different locations in China. Two secondary F 2 populations validated QTL with significant effect on appearance quality. Appearance quality (AQ) is the main determinants of market value of rice. Identification of QTL affecting AQ is the prerequisite for efficient improvement of AQ through marker-assisted selection (MAS). Two sets of reciprocal introgression lines derived from indica Minghui 63 and japonica 02428 were used to dissect the stability of QTL affecting five AQ traits, including grain length, grain width, length to width ratio, percentage of grains with chalkiness, and degree of endosperm chalkiness using 4568 bin genotype produced from 58,000 SNPs across five different environments. A total of 41 and 30 main-effect QTL were identified in MH63 and 02428 backgrounds, respectively. Among them, 9 background-independent QTL (BI-QTL) were found. There were also 13 and 10 stable-expressed QTL (SE-QTL) across at least two environments in MH63 and 02428 backgrounds, respectively. Two important BI- and SE-QTL regions (BISERs) including BISER-I harboring qPGWC5, qDEC5, qGW5.1, and qLWR5 on chromosome 5 and BISER-II harboring qGL7, qLWR7, qPGWC7, and qDEC7 on chromosome 7 were identified. The BISER-II was newly reported and validated by two secondary F2 populations in the reciprocal backgrounds. Among 59 epistatic QTL (E-QTL) detected in this study, there were only four SE- but no BI-E-QTL detected in different environments, indicating that genetic background has stronger effect on AQ traits than the environmental factors, especially for percentage of grains with chalkiness (PGWC) and degree of endosperm chalkiness (DEC) with lower heritability. BISER-I and BISER-II harboring many BI- and SE-QTL with favorable alleles from slender grain rice are much important for improvement of rice AQ by MAS.