AAV vector design, authentication, and preparation

Detailed information of all AAVs used in this study is listed in Figure 1—source data 1, along with the authentication information. cDNAs of mouse Txnip, Hif1a, Hk2, Ldha, Ldhb, Slc2a1, Bsg1, Cpt1a, Oxct1, Mpc1, and Mpc2, and human Nrf2, were purchased from GeneCopoeia (Rockville, MD). Mouse Vegf164 cDNA (Robinson and Stringer, 2001) was synthesized through Integrated DNA Technologies (Coralville, IA). We obtained the following plasmids as gifts from various depositors through Addgene (Watertown, MA): Hk1, Pfkm, and Pkm2 (William Hahn and David Root; #23730, #23728, #23757), Pkm1 (Lewis Cantley and Matthew Vander Heiden; #44241), H2BGFP (Geoff Wahl; #11680), mitoRFP (i.e., DsRed2-mito, Michael Davidson; #55838), GFP-Txnip (Clark Distelhorst; #18758), W3SL (Bong-Kiun Kaang; #61463), 3xFLAG (Thorsten Mascher, #55180), and PercevalHR and pHRed (Gary Yellen; #49082, #31473). The cDNA of mouse RdCVF was a gift from Leah Byrne and John Flannery (UC Berkeley, CA). iGlucoSnFR was provided under a Material Transfer Agreement by Jacob Keller and Loren Looger (Janelia Research Campus, VA). RedO promoter was provided as a gift, and SynPVI (also known as ProA7) and SynP136 (also known as ProA1) promoters were provided under a Material Transfer Agreement, from Botond Roska (IOB, Switzerland). The Best1 promoter was synthesized by a lab member, Wenjun Xiong, using Integrated DNA Technologies based on the literature (Esumi et al., 2009). Mutated Txnip, dominant-negative HIF1α (Jiang et al., 1996), and RO1.7 promoter (Ye et al., 2016) were created from the Hif1a and RedO plasmids correspondingly in house using Gibson assembly. Of note, we found that the RedO promoter is stronger than SynP136 or SynPVI promoters, but less specific. RedO has a low level of expression in some rods. SynP136 and SynPVI drive expression that is exclusive to cones, that is, no rod expression in keeping with the observation of Roska Lab (Jüttner et al., 2019). SynPVI (0.5 kb) is shorter than SynP136 (2 kb), and is thus better for packaging insert genes that have a large size.

All of the new constructs in this study were cloned using Gibson assembly. For example, AAV-RedO-Txnip was cloned by replacing the EGFP sequence of AAV-RedO-EGFP at the NotI/HindIII sites, with the Txnip sequence, which was PCR-amplified from the cDNA vector adding two 20 bp overlapping sequences at the 5′- and 3′-ends. All of the AAV plasmids were amplified using Stbl3 Escherichia coli (Thermo Fisher Scientific). The sequences of all AAV plasmids were verified with directed sequencing and restriction enzyme digestion. The key plasmids were verified with Next-Generation complete plasmid sequencing (MGH CCIB DNA Core), which is able to capture the full sequence of the ITR regions. The genome sequence of critical AAVs (i.e., AAV8-RedO-Txnip.C247S and AAV8-RedO-Txnip.S308A) was verified with PCR and directed sequencing.

All of the vectors were packaged in recombinant AAV8 capsids using 293 T cells and purified with iodixanol gradient as previously described (Grieger et al., 2006; Xiong et al., 2015). The titer of each AAV batch was determined using protein gels, comparing virion protein band intensities with a previously established AAV standard. The concentration of our AAV production usually ranged from 2 × 10¹² to 3 × 10¹³ gc/mL. Multiple batches of key AAV vectors (e.g., four batches of AAV8-RedO-Txnip and three batches of AAV-RedO-siLdhb^(#2)) were made and tested in vivo to avoid any potential batch effects.