Dig

Abstract: The incredible generative ability of large-scale text-to-image (T2I) models has demonstrated strong power of learning complex structures and meaningful semantics. However, relying solely on text prompts cannot fully take advantage of the knowledge learned by the model, especially when flexible and accurate controlling (e.g., structure and color) is needed. In this paper, we aim to ``dig out" the capabilities that T2I models have implicitly learned, and then explicitly use them to control the generation more granularly. Specifically, we propose to learn low-cost T2I-Adapters to align internal knowledge in T2I models with external control signals, while freezing the original large T2I models. In this way, we can train various adapters according to different conditions, achieving rich control and editing effects in the color and structure of the generation results. Further, the proposed T2I-Adapters have attractive properties of practical value, such as composability and generalization ability. Extensive experiments demonstrate that our T2I-Adapter has promising generation quality and a wide range of applications. Our code is available at https://github.com/TencentARC/T2I-Adapter.

Abstract: Herpes simplex virus (HSV) entry requires the interaction of glycoprotein D (gD) with a cellular receptor such as herpesvirus entry mediator (HVEM or HveA) or nectin-1 (HveC). However, the fusion mechanism is still not understood. Since cholesterol-enriched cell membrane lipid rafts are involved in the entry of other enveloped viruses such as human immunodeficiency virus and Ebola virus, we tested whether HSV entry proceeds similarly. Vero cells and cells expressing either HVEM or nectin-1 were treated with cholesterol-sequestering drugs such as methyl-beta-cyclodextrin or nystatin and then exposed to virus. In all cases, virus entry was inhibited in a dose-dependent manner, and the inhibitory effect was fully reversible by replenishment of cholesterol. To examine the association of HVEM and nectin-1 with lipid rafts, we analyzed whether they partitioned into nonionic detergent-insoluble glycolipid-enriched membranes (DIG). There was no constitutive association of either receptor with DIG. Binding of soluble gD or virus to cells did not result in association of nectin-1 with the raft-containing fractions. However, during infection, a fraction of gB but not gC, gD, or gH associated with DIG. Similarly, when cells were incubated with truncated soluble glycoproteins, soluble gB but not gC was found associated with DIG. Together, these data favor a model in which HSV uses gB to rapidly mobilize lipid rafts that may serve as a platform for entry and cell signaling. It also suggests that gB may interact with a cellular molecule associated with lipid rafts.

Abstract: While CD28 functions as the major T cell costimulatory receptor, a number of other T cell molecules have also been described to induce T cell costimulation. Here, we investigated the mechanisms by which costimulatory molecules other than CD28 contribute to T cell activation. Non-CD28 costimulatory molecules such as CD5, CD9, CD2, and CD44 were present in the detergent-insoluble glycolipid-enriched (DIG) fraction/raft of the T cell surface, which is rich in TCR signaling molecules and generates a TCR signal upon recruitment of the TCR complex. Compared with CD3 ligation, coligation of CD3 and CD5 as an example of DIG-resident costimulatory molecules led to an enhanced association of CD3 and DIG. Such a DIG redistribution markedly up-regulated TCR signaling as observed by ZAP-70/LAT activation and Ca2+ influx. Disruption of DIG structure using an agent capable of altering cholesterol organization potently diminished Ca2+ mobilization induced by the coligation of CD3 and CD5. This was associated with the inhibition of the redistribution of DIG although the association of CD3 and CD5 was not affected. Thus, the DIG-resident costimulatory molecules exert their costimulatory effects by contributing to an enhanced association of TCR/CD3 and DIG.