Effective scaling and a flexible task interface enable large language models to excel at many tasks. PaLI (Pathways Language and Image model) extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pretrained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train the largest ViT to date (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-ofthe-art in multiple vision and language tasks (such as captioning, visual questionanswering, scene-text understanding), while retaining a simple, modular, and scalable design.

No abstract

Activator of G protein Signaling 1 (AGS1) and Ras homologue enriched in striatum (Rhes) define a new group of Ras-like monomeric G proteins whose signaling properties and physiological roles are just beginning to be understood. Previous results suggest that AGS1 and Rhes exhibit distinct preferences for heterotrimeric G proteins, with AGS1 selectively influencing Galphai and Rhes selectively influencing Galphas. Here, we demonstrate that AGS1 and Rhes trigger nearly identical modulation of N-type Ca(2+) channels (Ca(V)2.2) by selectively altering Galphai-dependent signaling. Whole-cell currents were recorded from HEK293 cells expressing Ca(V)2.2 and Galphai- or Galphas-coupled receptors. AGS1 and Rhes reduced basal current densities and triggered tonic voltage-dependent (VD) inhibition of Ca(V)2.2. Additionally, each protein attenuated agonist-initiated channel inhibition through Galphai-coupled receptors without reducing channel inhibition through a Galphas-coupled receptor. The above effects of AGS1 and Rhes were blocked by pertussis toxin (PTX) or by expression of a Gbetagamma-sequestering peptide (masGRK3ct). Transfection with HRas, KRas2, Rap1A-G12V, Rap2B, Rheb2, or Gem failed to duplicate the effects of AGS1 and Rhes on Ca(V)2.2. Our data provide the first demonstration that AGS1 and Rhes exhibit similar if not identical signaling properties since both trigger tonic Gbetagamma signaling and both attenuate receptor-initiated signaling by the Gbetagamma subunits of PTX-sensitive G proteins. These results are consistent with the possibility that AGS1 and Rhes modulate Ca(2+) influx through Ca(V)2.2 channels under more physiological conditions and thereby influence Ca(2+)-dependent events such as neurosecretion.

Neurokinin (NK) 1 receptors and Ca V 2.3 calcium channels are both expressed in nociceptive neurons, and mice lacking either protein display altered responses to noxious stimuli. Here, we examined modulation of Ca V 2.3 through NK1 receptors expressed in human embryonic kidney 293 cells. We find that NK1 receptors generate complex modulation of Ca V 2.3. In particular, weak activation of these receptors evokes mainly stimulation of Ca V 2.3, whereas strong receptor activation elicits profound inhibition that overlaps with channel stimulation. Unlike R-type channels encoded by Ca V 2.3, L-type (Ca V 1.3), Ntype (Ca V 2.2), and P/Q-type (Ca V 2.1) channels are inhibited, but not stimulated, through NK1 receptors. Pharmacological experiments show that protein kinase C (PKC) mediates stimulation of Ca V 2.3 through NK1 receptors. The signaling mechanisms underlying inhibition were explored by expressing proteins that buffer either G␣ q/11 (regulator of G protein signaling protein 3T and carboxyl-terminal region of phospholipase C-␤1) or G␤␥ subunits (transducin and the carboxyl-terminal region of bovine G-protein-coupled receptor kinase). A fast component of inhibition was attenuated by buffering G␤␥, whereas a slow component of inhibition was reduced by buffering G␣ q/11 . When both G␤␥ and G␣ q/11 were simultaneously buffered in the same cells, inhibition was virtually eliminated, but receptor activation still triggered substantial stimulation of Ca V 2.3. We also report that NK1 receptors accelerate the inactivation kinetics of Ca V 2.3 currents. Altogether, our results indicate that NK1 receptors modulate Ca V 2.3 using three different signaling mechanisms: a fast inhibition mediated by G␤␥, a slow inhibition mediated by G␣ q/11 , and a slow stimulation mediated by PKC. This new information concerning R-type calcium channels and NK1 receptors may help in understanding nociception, synaptic plasticity, and other physiological processes.

No abstract