pH-responsive and hyaluronic acid-functionalized metal-organic frameworks for therapy of osteoarthritis

Abstract

Drug therapy of osteoarthritis (OA) is limited by the short retention and lacking of stimulus-responsiveness after intra-articular (IA) injection. The weak acid microenvironment in joint provides a potential trigger for controlled drug release systems in the treatment of OA. Herein, we developed an pH-responsive metal−organic frameworks (MOFs) system modified by hyaluronic acid (HA) and loaded with an anti-inflammatory protocatechuic acid (PCA), designated as MOF@HA@PCA, for the therapy of OA. Results demonstrated that MOF@HA@PCA could smartly respond to acidic conditions in OA microenvironment and gradually release PCA, which could remarkably reduce synovial inflammation in both IL-1β induced chondrocytes and the OA joints. MOF@HA@PCA also down-regulated the expression of inflammatory markers of OA and promoted the expression of cartilage-specific makers. This work may provide a new insight for the design of efficient nanoprobes for precision theranostics of OA.

Effects of Parameter Norm Growth During Transformer Training: Inductive Bias from Gradient Descent

“…On the other hand, the pre-norm transformer reaches 66.35 on Wikitext-2 and 26.16 on PTB, slightly outperformingWang et al (2019). This is consistent with previous findings(Xiong et al, 2020) showing advantages of pre-norm over post-norm.…”

“…Interestingly, the homogeneity arguments do not work out if we instead consider the post-norm transformer architecture (Xiong et al, 2020). learning rate η and weight decay λ by training a variety of transformer language models on Wikitext-2 for 1 epoch.…”

“…We consider two variants of the transformer: pre-norm and post-norm, which vary in the relative order of layer normalization and residual connections (cf. Xiong et al, 2020). Every model exhibits norm growth over training.…”

“…Let LN(•) denote a layer norm followed by a learned affine transformation. Here we will consider the pre-norm transformer variant (Xiong et al, 2020), meaning that LN comes before the residual connection wherever it appears. 17 We will also assume that there are no biases, making all affine transformations into strict linear transformations.…”

Do Transformer Modifications Transfer Across Implementations and Applications?

“…We use the original Transformer model as our baseline model with two modifications: First, we apply layer normalization before the selfattention and feedforward blocks instead of after. This small change has been unanimously adopted by all current Transformer implementations because it leads to more effective training Xiong et al, 2020). Secondly, we use relative attention with shared biases (as used in Raffel et al (2019)) instead of sinusoidal positional embeddings, which makes it easier to train the model.…”