Mitochondrial sulfide promotes life span and health span through distinct mechanisms in developing versus adult treated Caenorhabditis elegans

Significance Deteriorating health across the life course is a major societal burden, and effective therapeutics are lacking. Mitochondrial decline has long been associated with age-related health loss. We show that small, clinically meaningful doses of a mitochondrion-targeting sulfur donor (AP39) extend Caenorhabditis elegans health in older age, which act by maintaining mitochondrial integrity. Adult onset of AP39 delivery, when mitochondrial and cell structural dysfunction are already manifested, also promoted healthy aging. Distinct association of health span extension with mitochondria, cytoskeletal and peroxisome molecular profiles, under regulation of the elt-6/elt-3 transcription factor regulatory circuit, further distinguished adult-onset AP39 therapy. Our results establish a framework for forward translating mitochondrial sulfide as a potentially viable healthy aging intervention in mammals.

The mitochondrial suspension was further centrifuged at 15,000 g for 3 min to wash the mitochondria. The mitochondrial pellet reformed, and after removing the supernatant, the pellet was added to 300 μl of resuspension buffer [human serum albumin, 0.5 mg/ml, 240 mM sucrose, 15 mM KH2PO4, 2 mM Mg(CH3COO)2 × 4 H2O, and 0.5 mM EDTA at pH 7.2] and centrifuged at 15,000 g for 3 min. The final mitochondrial pellet was added to 100 μl of resuspension buffer and stored on ice until analysis. To measure CS, 15 μl mitochondrial suspension was added to 185 μl homogenization buffer (50 mM KH2PO4, 1 mM EDTA, and 0.05% Triton X-100) and was homogenized using a glass pestle at 200 rpm for 2 min. The homogenate was then centrifuged at 24,000 g before the formation of 5,5'-dithiobis-(2-nitrobenzoic acid)-coenzyme A containing a thiol group (DTNB-CoA-SH) in the supernatant was measured spectrophotometrically at 412 nm to determine CS activity.

RNA interference protocols
Lifespan and healthspan experiments for the RNAi screen were performed using The Infinity Screening System (NemaLife Inc, Texas, USA)a microfluidics-based lifespan machine. Day 0 adult animals were washed from plates with liquid NGM solution (3 g NaCl, 2.5 g peptone, 1 L H2O), loaded in a 2.5 ml syringe equipped with a microtube and injected into sterilised Infinity Chips (NemaLife Inc, Texas, USA) pillar-based microfluidic devices, as previously described (2). 100 µl of food solution (liquid NGM containing 20 mg/ml bacteria, 0.6 mM IPTG, 50 µg/ml ampicillin, 100nM AP39 + 0.01% DMSO or 0.01 % DMSO) was then added to the Infinity Chips. Throughout lifespan experiments, the Infinity Chips were stored at 20 °C in a humid chamber to prevent dehydration. To remove progeny, the microfluidic devices were washed with sterile liquid NGM every day for 90s in the Infinity Screening System, for the duration of the worms' lifespan. After each wash the devices were filmed for 90s using an Apple iPod. The food and drug solution were replaced daily in each device. The videos recorded were analysed using the NemaLife software (NemaLife Inc, Texas, USA) to obtain daily numbers for live and moving worms.
For all experiments, n≈160 per condition, across 2 biological replicates.

C. elegans and bacteria preparation for RNAi screen
Animals were syncronized to L1 larval stage by gravity flotation and grown for 60 hours on NGM agar plates containing 1 mM IPTG, 50 µg/ml ampicillin and either 100nM AP39

RNA isolation for next generation sequencing
Synchronised worms were grown on NGM agar plates until young adulthood and treated with AP39 mtH2S from either L1 or day 0, as described above. On sample collection day, 100 worms were manually picked and added to 1 ml of TRIzol™ Reagent (Thermofisher using Kallisto (v0.46.1)(4) to estimate transcript-level abundances, from which gene-level counts were inferred via the tximport package for R (5). Genes with consistently low expression (count < 10 in every sample) were then removed, leaving 10,451 genes for downstream analyses.

RNA-seq data analyses
The DESeq2 package for R (6) was used to test for differential gene expression, with comparisons made between day 0 wild-type worms and wild-type/treated worms from each of the other two time points, as well as directly between conditions at each of the three time points. In each case, the 'ashr' adaptive shrinkage method was applied to log2 fold-change estimates (7) and the Benjamini-Hochberg procedure used to control for false discovery rate (FDR), with genes defined as significantly differentially expressed when FDR < 0.05. To further elucidate time-specific condition effect patterns, the 'degPatterns' algorithm of the DEGreport R package was applied to normalized counts (obtained via variance stabilizing transformation) to group by expression profile any gene differentially regulated between treatments at day 0 and/or: differentially expressed in at least one treatment at one or more time point vs. day 0 wild-types and differentially expressed directly between at least one treatment pair at one or more time points. During the clustering process, genes outlier of cluster distribution were removed, and a default minimum cluster size of 15 genes applied. Functional enrichment analysis of defined gene clusters was consequently undertaken using the gprofiler2 package for R (8), with the following data sources utilized: each Gene Ontology (GO) category (biological process, BP; cellular component, CC; molecular function, MF), the Kyoto Encyclopedia of Gene and Genomes (KEGG) pathway database, and regulatory motif matches from TRANSFAC. In each case, all genes that were tested for differential expression were input as background and the annotated statistical domain scope option employed, with enriched terms selected as those with an FDR < 0.05 enriched for at least 2 genes. Each gene cluster was also input into the Online Search Tool for Retrieval of Interacting Genes/Proteins (STRING, v11.5; (9)) to infer respective protein-protein interaction (PPI) networks, with default confidence score of > 0.4 used to define connections, as quantified using all active interaction sources excluding text-mining. Network components were ranked by node degree to deduce the most highly connected ('hub') components within each network.