RNASET2 degrades mRNAs that protect against lipotoxicity Authors: Shuiling Zhao, Stevens Bontemps, Ryan M. Nottingham, Jun Yao, Mariana Acuña, David E. Cohen, Alan M. Lambowitz, Jean E. Schaffer Published: 2026 ABSTRACT ObjectivesRNASET2 is a lysosomal RNase whose enzymatic function is required for early events in lipotoxicity. However, the endogenous RNA substrates of RNASET2 that modulate lipid-induced cell death are not known. The purpose of this study was to identify RNASET2 substrates that impact lipotoxic stress.MethodsRNA sequencing was used to identify RNAs that increase in abundance in human cells upon RNASET2 knockdown, and actinomycin D assays were used to show that RNASET2 impacted decay rates of these RNAs. We tested for the presence of these RNAs in immunoisolated lysosomes and determined the contribution of the lysosomal membrane transporter SIDT2 in delivery of these RNAs to the lysosome. A role for these RNAs in lipotoxic cell death was directly tested in loss- and gain of function analysis.ResultsRNASET2 knockdown increased steady-state abundance of UCHL3, PFN2 and PRDX3 mRNAs and prolonged their decay rate, leading to increased protein expression. These mRNAs were delivered to the lysosomal lumen by the lysosomal membrane transporter SIDT2 that mediates RNautophagy. While UCHL3 and PFN2 have not previously been implicated in lipotoxic responses, expression of these proteins protected against lipid-induced cell death.ConclusionsOur study identified specific mRNA substrates of RNASET2 and uncovered a previously unexplored function for lysosomes and RNautophagy in regulation of the response to metabolic stress. Moreover, we demonstrated that RNautophagy selectively regulates turnover of specific endogenous RNAs and thereby impacts regulation of gene expression. Introduction In type 2 diabetes, hyperlipidemia causes accumulation of excess fatty acids in non-adipose tissues that is associated with lipotoxicity in the heart, liver, retina, kidney and peripheral nerves [[1], [2], [3], [4], [5]]. Ectopic lipid accumulation initiates endoplasmic reticulum and oxidative stress that contribute to organ dysfunction [[6], [7], [8]]. Genetic screens in model systems have provided powerful approaches to dissect the early molecular events in lipotoxicity that underlie complications of diabetes and hyperlipidemia, because loci are identified on the basis of functional contributions to the pathway of interest. As anticipated, such approaches uncovered genes that regulate lipid metabolism as determinants of the lipotoxic response [9,10]. In our previous loss-of-function genetic screen in Chinese hamster ovary cells, the finding that RNASET2 is a regulator of palmitate-induced cell death was unexpected, because this is a lysosomal enzyme and its endogenous substrates are not known [11]. While many RNases function in the cytoplasm or nucleus of eukaryotic cells, RNASET2 is one of three mannose 6-phosphate modified RNases localized to the lysosomal lumen [12,13]. Biochemical studies of purified recombinant RNASET2 show this enzyme cleaves single-stranded RNA optimally at an acidic pH [14], consistent with its function within lysosomes. Since purified RNASET2 has limited sequence specificity when assayed against model RNA substrates [[14], [15], [16], [17]], delivery of RNAs to the lysosome may govern which RNAs are degraded by RNASET2. Identification of specific RNAs that serve as substrates for degradation by RNASET2 is critical for a full understanding its contribution to metabolic stress responses. In C. elegans and S. cerevisiae, T2 endoribonucleases related to RNASET2 have been shown to degrade rRNAs delivered to lysosomes by ribophagy and to cleave cytosolic tRNAs during oxidative stress [18,19]. Whether rRNAs or tRNAs are RNASET2 substrates during lipotoxicity is not known. To address this gap, we mapped the time course of rRNA degradation during lipotoxicity and employed Thermostable Group II Intron Reverse Transcriptase (TGIRT) RNA sequencing to identify candidate RNAs that increase at steady state in RNASET2 haploinsufficient cells concomitant with protection against lipotoxicity. Our experiments showed that RNASET2 degrades specific mRNAs that are delivered to the lysosome by RNautophagy. Expression of these RNASET2 substrates modulates susceptibility to lipotoxicity. Methods Cells HEK 293T cells were cultured in Dulbecco's Modified Eagle Medium containing 4500 mg/L glucose and 110 mg/L pyruvate (Thermo Fisher 11995–065) supplemented with 10% heat-inactivated FBS and 2 mM glutamine at 37 °C with 5% CO2. Lentivirus production and transduction To generate lentivirus, HEK 293T were co-transfected with shRNA plasmids and lentiviral packaging plasmid mix (Cellecta Ready-to-Use CPCP-K2A) using Lipofectamine™ Transfection Reagent (Invitrogen 18324012) following the manufacturers’ protocols. Viral supernatants were supplemented with 4 μg/ml polybrene (Sigma–Aldrich TR-1003) and used to transduce naïve HEK 293T cells for KD of RNASET2 or SIDT2, for expression of mCherry-GFP-LC3B fluorescent reporter, or for expression of Tmem192.3XHA. Cells transduced with shRNA lentivirus were selected by flow cytometry based on EGFP co-expression. All experiments with KD cells were performed within 6 passages of viral transduction and selection, a time frame over which the degree of KD was stable. See Supplementary Table 1 for plasmids. Lysosomal function To quantify lysosomal pH, actively growing cells (∼50% confluency) were incubated with 50 nM LysoTracker™ Deep Red (Thermo Fisher L12492) for 30 min and analyzed by flow cytometry. Lysosomal cargo delivery and degradation was quantified by incubating cells with 30 μg/mL DQ™ Red BSA (Thermo Fisher, D12051) and analyzed by flow cytometry. Treatment with 100 nM Bafilomycin A1 (BAF, Thermo Fisher AAJ67193XF) was used as a control for both LysoTracker (added 2 h prior to staining) and DQ Red BSA staining (added for 3 h during staining). Lipotoxicity assays Actively growing cells (∼60–65% confluency) were treated by supplementing growth media with 1 mM palmitic acid (PALM, NuChek PREP N-16A-O11-F) freshly complexed with 0.5 mM fatty acid-free BSA (Lampire Biological Laboratories 7500804) as a carrier or with 0.5 mM fatty acid-free BSA as vehicle control as previously described [20]. Floating and attached cells were recovered for all analyses. For quantification of death, cells were washed with PBS and stained with 3 μM Draq-7 (BioStatus DR70250) for 15 min at room temperature and analyzed by flow cytometry. RNA quantification RNA was extracted using TRIzol (for cells, Invitrogen 15596018) or TRIzol LS (for lysosomes, Invitrogen 10296010) and the Direct-zol RNA Miniprep kit (Zymo Research R2050). 1 μg RNA was reverse-transcribed using the iScript™ Reverse Transcription Supermix (Bio-Rad 1708841) or the SuperScript™ VILO™ MasterMix (Thermo Fisher 11755500). Real-time quantitative PCR (qRT-qPCR) for cellular mRNAs was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad 1725274) with data analysis using the ΔΔCt method relative to GAPDH or β-actin. ddPCR was performed for lysosomal RNA using the QuantStudio Absolute Q Digital PCR System (Thermo Fisher). Absolute Q dPCR assays were performed following the manufacturer's guide (Thermo Fisher, hUCHL3: Hs04334565_m1, hPFN2: Hs00160050_m1, hPRDX3: Hs00428953_g1). Lysosomal RNA was normalized to lysosomal protein, as determined by bicinchoninic acid protein assay (BCA Thermo Fisher 23225). See Supplementary Tables 2 and 3 for primers. For determination of RNA decay, actively growing cells were treated with 5 μg/ml Actinomycin D (Thermo Fisher A7592) for 0, 2, 4 and 8 h and RNA abundance analyzed by qRT-PCR. For UCHL3, PFN2, and PRDX3, slope of decay was calculated from best fit line for each independent experiment. Consistent with its known short half-life [21], >90% of the MYC mRNA was depleted by 2 h of actinomycin D treatment, and slope was determined from the 0- and 2 h time points only. Immunoblotting Cells were lysed using Pierce™ IP Lysis Buffer (Thermo Fisher 87787) containing cOmplete Protease Inhibitor Cocktail (Millipore Sigma11697498001), and protein quantified by BCA. Protein (10–25 μg) was separated on either 10% NuPAGE Bis-Tris gels, 4–12% NuPAGE Bis-Tris gels, or 10% SDS-PAGE gels, and transferred by iBlot (Invitrogen IB21001) to PVDF membranes. Following blocking (Tris-buffered saline with Tween-20 [TBST, 25 mM Tris pH 7.4, 2 mM KCl, 135 mM NaCl, 0.05% Tween-20] containing 5% non-fat milk), membranes were incubated with primary antibodies and appropriate horse-radish peroxidase-coupled secondary antibodies, developed using Clarity Max chemiluminescence substrate (Bio-Rad 1705060) or Immoblion Forte Western HRP substrate (Millipore Sigma WBLUF0500), and imaged with the ChemiDoc MP Imaging System (Bio-Rad12003154) with quantification using Image Lab. See Supplementary Table 4 for antibodies. Autophagy assays Following treatment with palmitate, BSA, or 250 nM Torin 1 (Selleck Chemical S2827) for 6 h and/or with 100 nM BAF for 2 h, autophagy induction was assessed by immunoblotting. To monitor autophagic flux, cells were transduced with a lentiviral mCherry-GFP-LC3B fluorescent reporter, treated with palmitate, BSA, or 250 μM Torin 1 for 6 or 10 h, fixed (4% paraformaldehyde) stained with DAPI, and analyzed by confocal microscopy. Autophagic flux was quantified by the ratio of mCherry puncta to the total of mCherry and GFP puncta. To determine the impact of autophagy on UCHL3, PFN2 and PRDX3 mRNAs, cells were treated with or without 250 nm Torin 1 for 24 h, 100 nm BAF for 2–16 h or 50 μM chloroquine (Thermo Fisher J64459.14) for 4–24 h prior to RNA extraction and quantification of gene expression by qRT-PCR. TGIRT sequencing RNA was extracted as above from RNASET2 KD and SC cells following treatment with palmitate or BSA for 8 h. Total RNA was prepared for TGIRT-seq library preparation using a two-fraction method [22], and RNA was cleaned up between steps using RNA Clean and Concentrator-5 Kit (Zymo Research R1015, RCC5). Briefly, following treatment with DNase I, rRNA was depleted (Ribocop depletion kit, Lexogen 144.24), and size fractionated into short (≤200 nt) and long RNAs fractions according to RCC5 protocol. Long RNAs were fragmented to ∼200 nucleotides (NEB Next RNA Fragmentation Module, New England Biolabs E6150S). Small and fragmented large RNAs were pooled prior to treatment with T4 Polynucleotide Kinase (Lucigen P0503K). Each step of RNA processing was evaluated and quantitated using an RNA 6000 Pico Chip mRNA assay on an Agilent 2100 Bioanalyzer. TGIRT-seq library construction was as previously described using the low salt template switching reaction buffer [23]. Libraries were evaluated by a High Sensitivity DNA chip on an Agilent 2100 Bioanalyzer, pooled, and sequenced on a single lane of an Illumina Novaseq SP-200 flowcell generating 100 bp paired-end reads. Bioinformatic analysis of TGIRT sequencing data Illumina TruSeq adapters and PCR primer sequences were trimmed from the reads with Cutadapt v2.8 (https://github.com/marcelm/cutadapt) (sequencing quality score cut-off at 20; p-value < 0.01) and reads <15-nt after trimming were discarded. To minimize mismapping, we used a sequential mapping strategy. First, reads were mapped to the human mitochondrial genome (Ensembl GRCh38 Release 93) and the Escherichia coli genome (GeneBank: NC_000913) using HISAT2 v2.1.0 (https://github.com/DaehwanKimLab/hisat2) with customized settings (-k 10 --rfg 1,3 --rdg 1,3 --mp 4,2 --no-mixed --no-discordant -- no-spliced- alignment) to identify reads derived from mitochondrial and E. coli RNAs (denoted Pass 1). Unmapped reads from Pass1 were then mapped to a collection of customized references sequences for human sncRNAs (miRNA, tRNA, Y RNA, Vault RNA, 7SL RNA, 7SK RNA) and rRNAs (2.2 kb 5S rRNA repeats from the 5S rRNA cluster on chromosome 1 (1q42, Gene- Bank: X12811) and 43 kb 45S rRNA containing 5.8S, 18S and 28S rRNAs from clusters on chromosomes 13, 14, 15, 21, and 22 (GeneBank: U13369), using HISAT2 with the following settings -k 20 --rdg 1,3 --rfg 1,3 --mp 2,1 --no-mixed --no-discordant --no-spliced-alignment -- norc (denoted Pass 2). Unmapped reads from Pass 2 were then mapped to the human genome reference sequence (Ensembl GRCh38 Release 93) using HISAT2 with settings optimized for nonspliced mapping (-k 10 --rdg 1,3 --rfg 1,3 --mp 4,2 --no-mixed --no-discordant --no-spliced- alignment) (denoted Pass 3) followed by splice aware mapping (-k 10 --rdg 1,3 --rfg 1,3 --mp 4,2 --no-mixed --no-discordant --dta) (denoted Pass 4). For reads that mapped to multiple genomic loci with the same mapping score in passes 3 to 4, the alignment with the shortest distance between the two paired ends (i.e., the shortest read span) was selected. In the case of ties (i.e., reads with the same mapping score and read span), reads mapping to a chromosome were selected over reads mapping to scaffold sequences, and in other cases, the read was assigned randomly to one of the tied choices. The filtered multiply mapped reads were then combined with the uniquely mapped reads from Passes 3–4 by using SAMtools v1.10 (https://github.com/samtools/samtools). To generate counts for different genomic features, the mapped reads were intersected with Ensembl GRCh38 Release 93 gene annotations plus the RNY5 gene and its 10 pseudogenes, which were not annotated in this release. Annotations for repeat elements were from RepBase (https://www.girinst.org/repbase/), and annotations for hairpin miRNA and mature miRNA were from miRBase (https://www.mirbase.org/). Coverage of each feature was calculated by BEDTools v2.29.2 (https://bedtools.readthedocs.io/en/latest/content/bedtools-suite.html). To avoid miscounting reads with embedded sncRNAs that were not filtered out in Pass 2 (e.g., snoRNAs), reads were first intersected with sncRNA annotations, and the remaining reads were then intersected with the annotations for protein-coding gene RNAs, lincRNAs, antisense RNAs, and other lncRNAs to get the read count for each annotated feature. Read counts were normalized by using DESeq2 (https://github.com/thelovelab/DESeq2) with sample size factors estimated based on counts from nuclear protein-coding genes. Likelihood ratio test (LRT) was then used to identify any genes that show change in expression across datasets with different treatments in different cells. Cell fractionation Following treatment with palmitate or BSA, cells were harvested by trypsinization and sequential detergent extraction was used to isolate cytosolic and membrane-bound organelles as previously described [24]. Lysosome isolations Lysosome isolation from cells expressing TMEM192-3XHA was adapted from a previously described method [25]. 3–4 x 107 cells were harvested by trypsinization and cell pellets resuspended in KPBS buffer (136 mM KCl, 10 mM KH2PO4, 50 mM sucrose, pH7.2, 1X Protease cOmplete Inhibitor Cocktail) were broken in a pre-chilled cell homogenizer (Isobiotec, Heidelberg, Germany) with a 12 μm clearance ball size and 3 strokes. Following removal of cell debris (centrifugation at 1000 × g for 10 min at 4 °C), supernatants were incubated while nutating with 100 μl Pierce™ anti-HA Magnetic beads (Thermo Scientific 88837) for 30 min, followed by 2 washes with KPBS buffer supplemented with 300 mM NaCI. 10 μg of RNase A (New England Biolabs T3018L) was added during nutation and washes to eliminate RNA contamination from the cell lysate (RNAs that might bind outer surface of the lysosome), followed by the addition of 100U RNase inhibitor (Invitrogen AM2696) during elution. Lysosomes were eluted with 100 μl KPBS with 0.5% NP-40. Gain- and loss-of-function of RNASET2 targets Cells were transfected with siRNA pools (UCHL3: Horizon L-006059-00-0005; PFN2: Thermo Fisher s10379; PRDX3: Horizon L-010355-00-0005; CONTROL: Horizon D-001810-10-05) or with lentiviral plasmid for overexpression of UCHL3 or PFN2 (Supplementary Table 1) using Lipofectamine 3000 (Thermo Fisher L3000008) in Opti-Mem Medium (Thermo Fisher 31985088). Cells were treated the following day with palmitate or BSA prior to mRNA analysis (48 h following transfection) or flow cytometry quantification of cell death (48 h following palmitate vs. BSA treatment). RNA-seq RNA was extracted as above from SC and SIDT2 KD cells. Sequencing libraries were constructed using polyA enrichment and Kapa mRNA Hyperprep. NovaSeqX Plus 10B was used to generate 150-nucleotide paired-end reads. Bioinformatic analysis of RNA-seq Sequencing reads were trimmed, filtered for Phred quality (≥Q15), aligned to the human transcriptome (Emsembl version 113) using kallisto [26], and transcript counts were converted to gene counts using tximport [27]. We only retained genes that had at least 0.12 counts per million (CPM) in at least four samples. We normalized using TMM and Voom-transformed to log2CPM, which had a substantial batch effect due to replicate number. Differential expression between each knockdown vs. scrambled group was tested with replicate number as a covariate using Limma-Voom [28]. To identify genes that were changed in both knockdowns in the same direction, we tested replication with ScreenDMT [29]. We adjusted the log2CPM for replicate number prior to plotting. We corrected for multiple hypotheses using the Benjamini-Hochberg false discovery rate (FDR). The heatmap was created using ComplexHeatmap [30]. Statistics For biochemical and cell biological analyses, results are expressed as mean ± standard error (SE) for a minimum of 3 independent experiments, and each data point derives from an independent experiment. Two-tailed unpaired, paired, or ratio paired t-test, or one-way ANOVA with post-hoc correction for multiple comparisons were used as appropriate. For all tests, p < 0.05 was considered significant (adjusted p-value < 0.05 for TGIRT sequencing analysis; FDR <0.05 for RNA-seq). Data points identified as outliers based on Grubbs or Rout test were excluded. All statistical comparisons were performed using GraphPad Prism. Results RNASET2 haploinsufficiency protects against lipotoxicity in human cells Loss-of-function genetic screens have revealed essential roles of protein-coding mRNAs, long non-coding RNAs, and small nucleolar RNAs in lipotoxicity [9,10,31,32], prompting us to consider diverse RNA biotypes as potential RNASET2 substrates. To leverage the deep annotation of these RNAs in the human genome and a cell type that is susceptible to lipotoxicity, we established RNASET2 haploinsufficient human embryonic kidney (HEK) 293T cells using lentiviral shRNA transduction and flow cytometry selection. We generated stable populations of 293T cells with knockdown (KD) of RNASET2 mRNA compared to cells transduced with a control (SC) shRNA (Figure 1A). KD decreased abundance of the 36 kDa RNASET2 precursor protein and its 27 and 19 kDa peptide products (Figure 1B).Figure 1Haploinsufficiency of RNASET2 in human cells confers lipotoxicity resistance. HEK 293T cells were transduced with one of two lentiviral shRNA targeting RNASET2 (KD1 or KD2) or non-targeting shRNA (SC). (A) RNASET2 expression relative to GAPDH by RT-qPCR. Means ± standard error (SE) for n = 4 independent experiments. (B) Representative immunoblot of RNASET2 in KD and SC cells with quantification of RNASET2 (precursor and 2 processed peptides) relative to β-ACTIN. Means ± SE for n = 3. (C, D) Representative plots of lysosomal pH assessed by LysoTracker Deep Red staining (C) or lysosomal activity by DQ-BSA staining (D), analyzed by flow cytometric analysis in SC and KD cells. SC cells treated with 100 nM bafilomycin A1 (BAF) served as control. Graphs show means ± SE for n = 3–4 independent experiments. (E) Representative plots of cell death (Draq-7 positive) in KD1 and SC cells following treatment with palmitate complexed to BSA (PALM+) vs. BSA carrier alone (PALM−) for 48 h. Bar graph shows means ± SE for n = 3 independent experiments. p-values by one-way ANOVA. Haploinsufficiency of RNASET2 in human cells confers lipotoxicity resistance. HEK 293T cells were transduced with one of two lentiviral shRNA targeting RNASET2 (KD1 or KD2) or non-targeting shRNA (SC). (A) RNASET2 expression relative to GAPDH by RT-qPCR. Means ± standard error (SE) for n = 4 independent experiments. (B) Representative immunoblot of RNASET2 in KD and SC cells with quantification of RNASET2 (precursor and 2 processed peptides) relative to β-ACTIN. Means ± SE for n = 3. (C, D) Representative plots of lysosomal pH assessed by LysoTracker Deep Red staining (C) or lysosomal activity by DQ-BSA staining (D), analyzed by flow cytometric analysis in SC and KD cells. SC cells treated with 100 nM bafilomycin A1 (BAF) served as control. Graphs show means ± SE for n = 3–4 independent experiments. (E) Representative plots of cell death (Draq-7 positive) in KD1 and SC cells following treatment with palmitate complexed to BSA (PALM+) vs. BSA carrier alone (PALM−) for 48 h. Bar graph shows means ± SE for n = 3 independent experiments. p-values by one-way ANOVA. To determine the impact of RNASET2 KD on lysosomal integrity, we analyzed lysosomal acidification using LysoTracker Deep Red staining. In KD1 cells with 43% decrease in RNASET2, lysosomal acidity was unchanged (Figure 1C). By contrast, in KD2 cells with 81% decrease in RNASET2, lysosomal acidity was decreased, suggesting that near depletion of RNASET2 perturbs lysosomal integrity. Furthermore, we used Dye-Quenched Bovine Serum Albumin (DQ-BSA) to assess delivery of cargo through the endocytic pathway to the lysosome and luminal proteolytic activity [33]. DQ-BSA staining was decreased in KD2 but not KD1 cells (Figure 1D), indicating that near depletion of RNASET2 also impairs lysosomal function. These results are consistent with reports that knockout of lysosomal nucleases, PLD3 and DNASE2A, alters lysosomal function [34,35]. In light of these findings, we used KD1 as a model of RNASET2 haploinsufficiency in further analyses. To determine the impact of RNASET2 knockdown on the response of 293T cells to lipotoxicity, we quantified RNASET2 expression and cell death. As previously observed in Chinese hamster ovary cells [11], exposure to lipotoxic concentrations of palmitate increased RNASET2 expression in SC cells (Fig. S1), although increase was only observed at 48 h of treatment. At each time point analyzed, RNASET2 expression in KD1 cells was significantly less than SC cells. To assess sensitivity to lipotoxicity, KD1 and SC cells were treated with palmitate or BSA as a control and lipotoxicity was quantified by flow cytometric analysis. KD cells were significantly protected from cell death (Figure 1E). Thus, similar to previous observations in rodent fibroblasts and myoblasts [11], haploinsufficiency of RNASET2 in HEK 293T cells protects against palmitate-induced lipotoxicity. Targeting of RNASET2 to the lysosomal lumen suggests that its RNA substrates may also be delivered to the lysosomal lumen, likely by autophagy. Previous studies have shown that macroautophagy can be induced or suppressed by lipotoxicity, depending on cell or tissue type and species of lipid [36]. Treatment of HEK 293T cells with palmitate increased ATG14 phosphorylation and increased the ratio of LC3-II/LC3-I, similar to the response to Torin 1 treatment (Fig. S2A and S2B) and consistent with macroautophagy induction. Addition of bafilomycin A1 (BAF) to palmitate-treated cells further increased the LC3-II/LC3-I ratio, consistent with increased autophagic flux. To monitor the arrival of autophagic cargo at the lysosome, we expressed an mCherry-GFP-LC3B reporter in HEK 293T cells. This reporter exhibits increased GFP and mCherry puncta upon autophagy induction, but GFP fluorescence relative to mCherry fluorescence decreases when the autophagosomes fuse with lysosomes. Following 6 and 10 h of palmitate treatment, increased ratio of red/total puncta provided evidence for arrival of autophagic cargo at the lysosomal lumen (Fig. S2C). This timing of autophagic induction is consistent with a process that could potentially deliver RNA substrates to the lysosomal lumen and impact susceptibility to early events in lipotoxicity [11]. In an RNASET2-deficient zebrafish model, lysosomes in the central nervous system are engorged with undigested rRNA, indicating that RNASET2 may have a role in rRNA degradation [37]. Furthermore, many cellular stresses upregulate ribophagy [38]. We analyzed the time course of rRNA degradation during palmitate-induced lipotoxicity in RNASET2 KD and SC cells. At 24 h following palmitate treatment (but not at earlier timepoints), 18S rRNA was significantly decreased in SC cells but not in RNASET2 KD cells (Fig. S2D). There were no changes in 28S rRNA (Fig. S2E). These findings indicate that ribophagy is a late event in palmitate-induced lipotoxicity. Given that RNASET2 haploinsufficiency manifests as increased antioxidant capacity at baseline and resistance to lipid-induced oxidative stress as early as 2 h following palmitate treatment [11], our findings suggest that RNASET2 substrates other than rRNAs determine susceptibility to lipotoxicity. Selective mRNAs increase in RNASET2 haploinsufficient cells Based on the hypothesis that RNASET2 substrates that contribute to lipotoxic cell death would be increased in KD1 compared to SC cells, we isolated total RNA from untreated (BSA control) and palmitate-treated (8 h) cells. Libraries were constructed from ribo-depleted RNA using TGIRT to capture RNAs of diverse sizes and biotypes [23,39,40]. TGIRT-seq revealed a profile typical of cellular RNAs, with reads corresponding to protein-coding genes and tRNAs being the most abundant biotypes followed by other small non-coding RNAs (sncRNAs). The remaining small proportion of reads mapped to other biotypes such as long non-coding RNAs (lncRNA, including lincRNA, pseudogenes, and antisense RNAs), repeat elements, and residual rRNAs (Figure 2A). Among non-tRNA sncRNAs, the most abundant were 7SL RNA and 7SK RNA along with snoRNAs and snRNAs. Vault RNAs and miRNAs were relatively low compared to other sncRNA biotypes (Figure 2B). KD cells did not differ significantly from SC cells in abundance of nuclear or mitochondrial tRNAs, tRNA fragments, or snoRNAs.Figure 2Impact of RNASET2 KD on gene expression. TGIRT-seq of ribo-depleted RNA from RNASET2 knockdown (KD1) and SC cells treated with PALM vs. BSA carrier (n = 4 independent experiments). (A, B) Stacked bar graphs showing the percentage of reads in each sample assigned to rRNA, repeats, mitochondrial tRNA (MT), small non-coding RNA (sncRNA), tRNA, lncRNA, antisense, lincRNA, pseudogene, and protein coding RNA (A) and for sncRNA subcategories other than tRNAs (B), including vault RNA (VT RNA), small nuclear RNA (snRNA) and small nucleolar RNA (snoRNA). (C) Heatmap of genes identified by DESeq2 analysis as significantly differentially expressed (Padj ≤ 0.05). A likelihood ratio test was used to identify any genes that were significantly expressed in any of the groups compared to the rest, with tiles colored by z-score of DESeq2-normalized counts. Color scale for z-score and color code for RNA biotype are at bottom of heatmap. (D–G) Four distinguishable groups of genes and features were identified and color-coded based on the clustering and expression pattern (groups 1 to 4, with number of genes and features in each group in parentheses). Box plots of average z-scores of DESeq2-normalized counts in each experimental group for each differentially expressed gene. Thin colored lines connect data points for the same gene in each experimental group. A thick black line in each plot shows a trend line fitted by polynomial regression. (H) Heatmap of genes in Group 4. Impact of RNASET2 KD on gene expression. TGIRT-seq of ribo-depleted RNA from RNASET2 knockdown (KD1) and SC cells treated with PALM vs. BSA carrier (n = 4 independent experiments). (A, B) Stacked bar graphs showing the percentage of reads in each sample assigned to rRNA, repeats, mitochondrial tRNA (MT), small non-coding RNA (sncRNA), tRNA, lncRNA, antisense, lincRNA, pseudogene, and protein coding RNA (A) and for sncRNA subcategories other than tRNAs (B), including vault RNA (VT RNA), small nuclear RNA (snRNA) and small nucleolar RNA (snoRNA). (C) Heatmap of genes identified by DESeq2 analysis as significantly differentially expressed (Padj ≤ 0.05). A likelihood ratio test was used to identify any genes that were significantly expressed in any of the groups compared to the rest, with tiles colored by z-score of DESeq2-normalized counts. Color scale for z-score and color code for RNA biotype are at bottom of heatmap. (D–G) Four distinguishable groups of genes and features were identified and color-coded based on the clustering and expression pattern (groups 1 to 4, with number of genes and features in each group in parentheses). Box plots of average z-scores of DESeq2-normalized counts in each experimental group for each differentially expressed gene. Thin colored lines connect data points for the same gene in each experimental group. A thick black line in each plot shows a trend line fitted by polynomial regression. (H) Heatmap of genes in Group 4. DESeq2 analysis followed by likelihood ratio test was used to identify differentially expressed genes between cell types (SC vs. KD) and between treatments (BSA control vs. palmitate). Genes that had similar significantly different expression patterns (adjusted p-value <0.05) after any of the four treatments (SC, SC–PALM, KD, KD-PALM) were clustered and revealed four distinct expression patterns in a heatmap (Figure 2C). Group 1 (Figure 2D) contained genes whose transcripts were underrepresented in palmitate-treated cells (SC–PALM vs. SC and KD-PALM vs. KD). Group 2 (Figure 2E) contained genes whose transcripts were underrepresented in KD cells, regardless of treatment. Group 3 (Figure 2F) contained genes whose transcripts were overrepresented in palmitate-treated cells, regardless of RNASET2 expression. Group 4 (Figure 2G) contained genes whose transcripts were overrepresented in cells with RNASET2 KD, regardless of palmitate treatment (SC vs. KD and SC–PALM vs KD-PALM). We considered the 24 RNAs in Group 4 that were increased in KD cells, regardless of treatment, as potential RNASET2 substrates (Figure 2H). These included 22 protein-coding genes and 2 pseudogenes. Among the protein-coding genes, 5 have known roles in regulating cell death (ANP32B, C3orf38, RNF6, TMBIM6, and UCHL3) [[41], [42], [43], [44], [45]], and 3 have functions in producing reactive oxygen species or regulating oxidative stress responses (CYBA, PNPO and PRDX3) [[46], [47], [48]]. In independent analyses, we confirmed that steady-state abundance of UCHL3, PFN2, and PRDX3 mRNAs were significantly increased in KD1 cells both at baseline and following palmitate treatment (Figure 3A–C). This led to increased protein abundance for UCHL3 and PFN2 in KD1 cells under both basal and lipotoxic conditions (Figure 3D, E). KD of RNASET2 caused a trend for increase in PRDX3 protein at baseline, but not following palmitate treatment (Figure 3F). The increase in UCHL3 and PFN2 mRNAs was confirmed independently in cells transduced with another shRNA targeting RNASET2 (KD3, Fig. S3A–C). PRDX3 mRNA was not increased (Fig. S3D), possibly because the decrease in RNASET2 was more modest. Taken together, the timing of changes in UCHL3, PFN2 and PRDX3 mRNAs is consistent with altered antioxidant capacity at baseline and protection at the early stages of lipotoxicity.Figure 3RNASET2 KD increases abundance of specific mRNAs and their proteins. Analysis of RNASET2 knockdown (KD1) and control (SC) cells. (A–C) UCHL3 (A), PFN2 (B), and PRDX3 (C) mRNA abundance relative to GAPDH by RT-qPCR following treatment with PALM (PALM+) vs. BSA carrier (PALM−) for 24 h. Means ± SE for n = 9–11 independent experiments. (D–F) Representative immunoblots of UCHL3 (D), PFN2 (E), and PRDX3 (F) following treatment with PALM vs. BSA carrier for 24 h. Quantification of means ± SE for n = 4–5 independent experiments. P-Values by paired t-test (A-C) or ratio paired t-test (D-F). RNASET2 KD increases abundance of specific mRNAs and their proteins. Analysis of RNASET2 knockdown (KD1) and control (SC) cells. (A–C) UCHL3 (A), PFN2 (B), and PRDX3 (C) mRNA abundance relative to GAPDH by RT-qPCR following treatment with PALM (PALM+) vs. BSA carrier (PALM−) for 24 h. Means ± SE for n = 9–11 independent experiments. (D–F) Representative immunoblots of UCHL3 (D), PFN2 (E), and PRDX3 (F) following treatment with PALM vs. BSA carrier for 24 h. Quantification of means ± SE for n = 4–5 independent experiments. P-Values by paired t-test (A-C) or ratio paired t-test (D-F). We hypothesized that the impact of RNASET2 knockdown on mRNA abundance resulted from changes in mRNA decay rates. Following actinomycin D treatment to inhibit RNA polymerase II, the decay of UCHL3, PFN2, and PRDX3 mRNAs was significantly prolonged in KD1 cells compared to SC cells (Figure 4A–C). MYC, a mRNA known to be short-lived [21], served as a control and was unchanged (Figure 4D). Together, these findings support a model in which RNASET2 expression decreases the abundance of UCHL3, PFN2, and PRDX3 proteins by degrading their mRNAs.Figure 4RNASET2 KD prolongs decay of specific mRNAs. RNASET2 KD (KD1) and SC cells were treated with actinomycin D for indicated times. RT-qPCR quantification of UCHL3 (A), PFN2 (B), and PRDX3 (C), with MYC (D) as a control. For each experiment, slope of decay was plotted. Means ± SE for n = 3–8 independent experiments. p-values by unpaired t-test. RNASET2 KD prolongs decay of specific mRNAs. RNASET2 KD (KD1) and SC cells were treated with actinomycin D for indicated times. RT-qPCR quantification of UCHL3 (A), PFN2 (B), and PRDX3 (C), with MYC (D) as a control. For each experiment, slope of decay was plotted. Means ± SE for n = 3–8 independent experiments. p-values by unpaired t-test. mRNAs degraded by RNASET2 are targeted to the lysosome For RNASET2 to act on a substrate mRNA, the enzyme must physically contact its substrate. Mannose 6-phosphate modification of RNASET2 and an acidic pH optimum are most consistent with localization of mammalian RNASET2 in the lysosomal lumen [13,14]. Nonetheless, the related yeast T2 ribonuclease, RNY1, which normally localizes in the vacuole, is released into the cytosol in response to oxidative stress [19]. To test whether lipotoxic stress causes relocalization of RNASET2 from the lysosome to the cytosol of mammalian cells, we fractionated soluble and membrane components of 293T cells under basal and lipotoxic conditions, probing for lysosomal (LAMP1), ER (CALX, PDI), and soluble (HSP90) proteins as controls (Figure 5A). As expected, the endoplasmic reticulum marker PDI was associated with the membrane fraction of untreated cells but was released from the ER lumen to the cytosol following treatment with palmitate [49]. RNASET2 remained associated with membrane-bound organelles under both control and palmitate-treated conditions and was not released into the cytosol. This indicates that under lipotoxic stress, RNASET2 substrates are likely degraded in the lysosomal lumen.Figure 5mRNAs accumulate in lysosomes of RNASET2 KD cells. (A) Representative immunoblots of HEK 293T soluble and membrane fractions for lysosomes (LAMP1), ER (CALX, PDI), cytosol (HSP90), and RNASET2 following treatment with PALM (PALM+) vs. BSA carrier (PALM−) for 24 h. (B) Representative immunoblots of cell lysate (input) and isolated lysosomes from RNASET2 KD (KD1) and SC cells. (C–E) Digital droplet PCR quantification of UCHL3 (C), PFN2 (D), and PRDX3 (E) in lysosomes isolated from KD1 vs. SC cells following treatment with PALM (PALM+) vs. BSA carrier (PALM−). Means ± SE for n = 4. p-values by ratio paired t-test. mRNAs accumulate in lysosomes of RNASET2 KD cells. (A) Representative immunoblots of HEK 293T soluble and membrane fractions for lysosomes (LAMP1), ER (CALX, PDI), cytosol (HSP90), and RNASET2 following treatment with PALM (PALM+) vs. BSA carrier (PALM−) for 24 h. (B) Representative immunoblots of cell lysate (input) and isolated lysosomes from RNASET2 KD (KD1) and SC cells. (C–E) Digital droplet PCR quantification of UCHL3 (C), PFN2 (D), and PRDX3 (E) in lysosomes isolated from KD1 vs. SC cells following treatment with PALM (PALM+) vs. BSA carrier (PALM−). Means ± SE for n = 4. p-values by ratio paired t-test. We hypothesized that RNASET2 substrates targeted to lysosomes for degradation would accumulate in the lysosomal lumen in the setting of RNASET2 KD. To test this, we stably expressed HA-tagged TMEM192 in RNASET2 KD1 and SC cells and used the HA epitope-tag to isolate lysosomes for further analysis. Immunoblotting confirmed that lysosomes isolated by this method were intact, as evidenced by enrichment of not only LAMP1, but also luminal HEXA and RNASET2 (Figure 5B). As expected, the 36 kDa precursor form of RNASET2, which is observed in input that contained ER, was absent from the lysosomal fraction. Using digital droplet PCR to quantify target RNAs and normalizing RNA copies for lysosomal protein recovered, UCHL3, PFN2, and PRDX3 mRNAs were increased in the lysosomes of RNASET2 KD cells under both basal and lipotoxic conditions (Figure 5C–E). Overall, these data support a model in which specific mRNAs are delivered to the lysosomal lumen for degradation by RNASET2. RNASET2 substrates are delivered to lysosomes RNautophagy Autophagic pathways deliver cellular macromolecules to the lysosomal lumen for degradation via autophagosomes, chaperone-mediated autophagy, or microautophagy, processes that can be inhibited by the mammalian target of rapamycin (mTOR) kinase [[50], [51], [52]]. Treatment of 293T cells with Torin 1 to inhibit mTOR and induce autophagy decreased steady-state abundance of UCHL3, PFN2, and PRDX3 mRNAs, consistent with a potential role for an autophagic pathway in degradation of these RNAs (Figure 6A–C). However, treatment with BAF or chloroquine to block fusion of autophagosomes with lysosomes did not cause accumulation of these targets (Figure 6D, E), indicating that macroautophagy was unlikely to be responsible for delivery of these mRNAs to RNASET2 in the lysosomal lumen.Figure 6SIDT2 KD increases UCHL3, PFN2 and PRDX3 mRNAs and proteins. (A–C) RT-qPCR quantification of UCHL3 (A), PFN2 (B) and PRDX3 (C) mRNAs following treatment with Torin 1 for 24 h. Means ± SE for n = 3 independent experiments. (D,E) RT-qPCR quantification of UCHL3, PFN2 and PRDX3 mRNAs in cells treated with 100 nm BAF for 2–16 h (D) or 50 μm chloroquine for 4–24 h (E). (F-L) HEK 293T cells were transduced with lentiviral shRNAs targeting SIDT2 (KD1, KD2) or non-targeting (SC) shRNA. mRNA for SIDT2 (F), UCHL3 (G), PFN2 (H), and PRDX3 (I) were quantified by RT-qPCR. Representative immunoblots for UCHL3 (J), PFN2 (K), PRDX3 (L) and β-ACTIN protein with quantification of means ± SE for n = 3–4. p-values by unpaired t-test (A-C), one-way ANOVA (F–I), or ratio paired t-test (J-L). SIDT2 KD increases UCHL3, PFN2 and PRDX3 mRNAs and proteins. (A–C) RT-qPCR quantification of UCHL3 (A), PFN2 (B) and PRDX3 (C) mRNAs following treatment with Torin 1 for 24 h. Means ± SE for n = 3 independent experiments. (D,E) RT-qPCR quantification of UCHL3, PFN2 and PRDX3 mRNAs in cells treated with 100 nm BAF for 2–16 h (D) or 50 μm chloroquine for 4–24 h (E). (F-L) HEK 293T cells were transduced with lentiviral shRNAs targeting SIDT2 (KD1, KD2) or non-targeting (SC) shRNA. mRNA for SIDT2 (F), UCHL3 (G), PFN2 (H), and PRDX3 (I) were quantified by RT-qPCR. Representative immunoblots for UCHL3 (J), PFN2 (K), PRDX3 (L) and β-ACTIN protein with quantification of means ± SE for n = 3–4. p-values by unpaired t-test (A-C), one-way ANOVA (F–I), or ratio paired t-test (J-L). RNautophagy is a form of microautophagy for lysosomal delivery of RNAs via the multi-transmembrane protein, systemic RNAi-defective (SID) transmembrane family member 2 (SIDT2) [53,54]. SIDT2 predominantly localizes to the lysosomal limiting membrane where it functions as an RNA transporter to move target RNAs from the cytosol into the lysosomal lumen [54]. Although LAMP2C serves as a receptor on the cytosolic face of the lysosome for SIDT2's RNA cargo, only SIDT2 has been shown to be essential for this pathway [54,55]. To test whether RNautophagy regulates putative RNASET2 substrate mRNAs, we treated HEK 293T cells with two different lentiviral shRNAs to knockdown SIDT2 mRNA by 68 (KD1) and 66% (KD2), respectively (Figure 6F). We were unable to quantify impact on SIDT2 protein levels with commercially available antibodies due to lack of antibody specificity of these reagents. Nonetheless, SIDT2 KD increased steady-state abundance of mRNAs for UCHL3, PFN2, and PRDX3 mRNAs (Figure 6G–I). This also led to increases in UCHL3, PFN2, and PRDX3 proteins (Figure 6J-L), indicating that these mRNAs remained available for translation in the cytosol. Thus, SIDT2 regulated expression of UCHL3, PFN2, and PRDX3. To probe the mechanism by which SIDT2 regulated UCHL3, PFN2, and PRDX3, we analyzed mRNA decay rates in SIDT2 KD cells. Similar to findings with RNASET2 KD, treatment with two different shRNAs targeting SIDT2 slowed the decay of UCHL3 and PFN2 mRNAs (Figure 7A, B). There was a trend toward slower decay for PRDX3 that did not reach significance (Figure 7C). Decay of MYC was unchanged by SIDT2 KD (Figure 7D). To assess whether SIDT2 and RNASET2 function in the same pathway, we knocked down SIDT2 in cells with RNASET2 KD. Lysosomal accumulation of UCHL3, PFN2, and PRDX3 mRNAs observed with RNASET2 KD was abrogated by simultaneous KD of SIDT2 (Figure 7E–G). Together, our findings are consistent with a model of selective delivery of mRNAs to the lysosomal lumen by SIDT2, upstream of degradation by RNASET2.Figure 7SIDT2 KD prolongs decay rate of RNASET2 targets and decreases these RNAs in lysosomes. (A-D) SIDT2 KD (KD1 or KD2) and SC cells were treated with actinomycin D for indicated times. RT-qPCR quantification of UCHL3 (A), PFN2 (B), PRDX3 (C) and MYC (D). Means ± SE for n = 4 independent experiments. For each experiment, slope of decay was also plotted. (E-G) Digital droplet PCR quantification of UCHL3 (E), PFN2 (F), and PRDX3 (G) in lysosomes isolated from cells with RNASET2 knockdown (KD1), cells with both RNASET2 (KD1) and SIDT2 (KD1) knockdown, and control (SC) cells. Means ± SE for n = 4 independent experiments. p-values by one-way ANOVA. SIDT2 KD prolongs decay rate of RNASET2 targets and decreases these RNAs in lysosomes. (A-D) SIDT2 KD (KD1 or KD2) and SC cells were treated with actinomycin D for indicated times. RT-qPCR quantification of UCHL3 (A), PFN2 (B), PRDX3 (C) and MYC (D). Means ± SE for n = 4 independent experiments. For each experiment, slope of decay was also plotted. (E-G) Digital droplet PCR quantification of UCHL3 (E), PFN2 (F), and PRDX3 (G) in lysosomes isolated from cells with RNASET2 knockdown (KD1), cells with both RNASET2 (KD1) and SIDT2 (KD1) knockdown, and control (SC) cells. Means ± SE for n = 4 independent experiments. p-values by one-way ANOVA. SIDT2's role in delivery of RNASET2 substrates to RNASET2, raised the possibility that SIDT2 expression might impact susceptibility to lipotoxicity. However, using two independent shRNAs to knockdown SIDT2, we did not observe protection against palmitate-induced cell death (Fig. S4A). To gain insights into differences between SIDT2 KD and RNASET2 KD, we performed RNAseq analysis of cells with SIDT2 knockdown. In two independent lines, SIDT2 KD significantly increased the abundance of 331 mRNAs (Fig. S4B and S4C). This much broader impact of SIDT2 KD on mRNA abundance and the function of SIDT2 in transporting DNA as well as RNA across the lysosomal membrane [56] may explain the divergence in lipotoxicity phenotypes following knockdown of these different proteins. RNASET2 substrates impact sensitivity to lipotoxicity Prior studies have shown that sequestration of excess palmitate in lipid droplets, cellular lipid export, fatty acid oxidation, chemical chaperones that reduce ER stress, and antioxidants mitigate lipotoxic cell death and tissue damage [20,49,[57], [58], [59]]. The antioxidant PRDX3 and other peroxiredoxin family members afford protection in cell culture and animal models of lipotoxicity [[60], [61], [62], [63], [64]]. UCHL3, a deubiquitinating enzyme [65], and PFN2, an actin binding protein that regulates cytoskeletal dynamics and membrane trafficking [66,67], have not previously been implicated to function in the response to lipotoxicity. To explore the links between UCHL3, PFN2, and lipotoxicity, we tested the impact of loss- and gain-of-function of these proteins on palmitate-induced cell death. siRNA-mediated KD of UCHL3 or PFN2 increased palmitate-induced cell death, but the effects were not additive (Figure 8A, B). Overexpression of UCHL3 or PFN2 significantly reduced palmitate-induced cell death, although simultaneous overexpression did not have additive effects (Figure 8C, D). Lack of additivity of these effects suggests these proteins act in the same pathway and that the impact of either manipulation alone leads to maximal effect. Additionally, siRNA-mediated KD of UCHL3, PFN2, and PRDX3 abrogated protection from lipid-induced cell death in RNASET2 KD cells (Fig. S5). Together, our findings point to a role for UCHL3 and PFN2 in protecting cells in the face of lipid overload and suggest a model in which selective degradation of these mRNAs in the lysosome contributes to the progression of lipotoxic injury.Figure 8PFN and UCHL3 regulate lipotoxic cell death. (A, B) HEK 293T cells were transfected with siRNA targeting UCHL3 and/or PFN2 or non-targeting (CTRL) siRNA. UCHL3 and PFN2 mRNA abundance relative to GAPDH by RT-qPCR (A). Means ± SE for n = 3. Cell death (Draq-7+) following treatment with PALM vs. BSA carrier for 48 h (B). Means ± SE for n = 4–5. (C, D) HEK 293T were transduced with lentivirus for expression of UCHL3, PFN2 or EGFP as control (CTRL). Representative immunoblots for UCHL3 and PFN2 with quantification of expression (C). Means ± SE for n = 3. Cell death (Draq-7+) following treatment with palmitate vs. BSA control for 48 h (D). Means ± SE for n = 3 independent experiments. p-values by one-way ANOVA. PFN and UCHL3 regulate lipotoxic cell death. (A, B) HEK 293T cells were transfected with siRNA targeting UCHL3 and/or PFN2 or non-targeting (CTRL) siRNA. UCHL3 and PFN2 mRNA abundance relative to GAPDH by RT-qPCR (A). Means ± SE for n = 3. Cell death (Draq-7+) following treatment with PALM vs. BSA carrier for 48 h (B). Means ± SE for n = 4–5. (C, D) HEK 293T were transduced with lentivirus for expression of UCHL3, PFN2 or EGFP as control (CTRL). Representative immunoblots for UCHL3 and PFN2 with quantification of expression (C). Means ± SE for n = 3. Cell death (Draq-7+) following treatment with palmitate vs. BSA control for 48 h (D). Means ± SE for n = 3 independent experiments. p-values by one-way ANOVA. Discussion Processing and degradation of cellular RNAs is orchestrated by distinct endo- and exonucleases for different classes of RNAs through processes that are spatially organized within specific cellular compartments. The observation that haploinsufficiency of the lysosomal enzyme RNASET2 protects against lipotoxicity suggests that degradation of RNA promotes cell demise in response to metabolic stress and adds to a growing appreciation of the importance of lysosomes in this process. Herein, we show that mRNAs encoding UCHL3, PFN2, and PRDX3 are selectively trafficked to the lysosome where they are degraded by RNASET2 under both basal and lipotoxic conditions. Levels of RNASET2 expression determine abundance of these mRNAs and their encoded proteins, which protect against lipid-induced cell death. These findings provide new insights into control of gene expression by selective mRNA degradation and identify endogenous mRNA substrates for the RNASET2 that impact cell physiology. Lysosomes serve as hubs for degrading macromolecules through processes that are coupled to cellular needs and metabolic state. RNASET2, along with other RNases targeted to the lysosomal lumen, degrades rRNA cargo during ribophagy when ribosomes are delivered to the lysosome via macroautophagy [38]. Our data support a role for RNASET2 in degrading rRNA late during the progression of lipotoxicity. However, the observation that haploinsufficiency of RNASET2 increases basal antioxidant capacity and functions at the earliest stages of lipotoxicity, prior to degradation of rRNAs [11], prompted our exploration for additional non-rRNA substrates. Although TGIRT sequencing provides excellent quantification for diverse RNA biotypes [39], upon RNASET2 KD we observed significant increases only in mRNAs, but not in tRNAs or small nucleolar RNAs, other classes of RNAs implicated in eukaryotic oxidative and lipotoxic stress responses [19,32]. RNASET2 KD not only increased the steady-state cellular abundance of UCHL3, PFN2 and PRDX3 mRNAs, but it also prolonged the lifetime of these mRNAs, observations most consistent with their degradation by RNASET2. These effects were observed in cells with RNASET2 KD prior to palmitate treatment, consistent with a role for RNASET2 in susceptibility to lipotoxicity. It is well appreciated that lipophagy is an important mechanism for clearance of excess lipids to restore metabolic balance during lipid overload states [68]. Our study uncovers a previously unexplored role for selective lysosomal mRNA degradation in the response to lipotoxcity. Multiple pathways can deliver RNA to lysosomes for degradation [69]. Our results provide strong support for RNautophagy as the mechanism for delivery of mRNAs to RNASET2. First, although treatment of HEK 293T cells with palmitate induced macroautophagy, and Torin 1 decreased the levels of UCHL3, PFN2 and PRDX3, BAF and chloroquine treatment had no impact on cellular abundance of these mRNAs. If these mRNAs were cargo delivered to the lysosome by macroautophagy, inhibiting fusion of autophagosomes with lysosomes containing RNASET2 should have increased their abundance [70]. Second, the large effect of SIDT2 knockdown alone on abundance and decay of UCHL3, PFN2, and PRDX3 mRNAs indicates that RNautophagy has a major impact on turnover of these RNAs. The lysosomal membrane protein SIDT2 functions to facilitate RNA uptake into isolated lysosomes independent of ATG5, LC-3, and HSPA8, which distinguish RNautophagy from macroautophagy and chaperone-mediated autophagy [54,55]. RNautophagy has been best characterized with respect to transport of bulk RNA and model RNA substrates [54,71]. Our findings provide evidence that RNautophagy regulates selective degradation of endogenous mRNAs and alters gene expression. The greater magnitude of effect of SIDT2 KD compared to RNASET2 KD on these mRNAs may reflect the redundancy of RNases that are targeted to the lysosomes via mannose 6-phosphate modification (RNASE1, RNASE6) and other exonucleases capable of degrading RNA that also traffic to the lysosome (PLD3, PLD4) [12,72]. A study reported during preparation of this manuscript found lysosomal enrichment of 7SL Signal Recognition Particle RNAs, Y RNAs, and mRNAs bearing 5′TOP motifs in cells with complete knockout of RNASET2 or combined knockout of RNASET2, PLD3, and RNASE4 [73]. Our data suggest that complete KO of RNASET2 could have deleterious consequences for lysosome function. Moreover, this report analyzed RNA enrichment values, normalizing lysosomal read counts by whole-cell read counts. This likely limited discovery of mRNAs for which degradation in the lysosome impacted total cellular mRNA abundance. The mechanism by which RNAs are delivered to the surface of the lysosome is not known. Prior studies have established that RNA binds directly to the cytosolic domains of LAMP2C and SIDT2, and binding to LAMP2C exhibits a preference for poly-G(15) over comparable lengths of poly-A, poly-U and poly-C [54,71]. Although the 3′ UTR of the PFN2 transcript contains stretches of 5, 6, and 7 G's, PRDX3 and UCHL3 lack poly-G sequences greater than 4 nucleotides. Recognition and targeting of mRNAs to the LAMP2C/SIDT2 complex at the lysosome surface may involve specific sequence features, epigenetic modifications, and/or mRNA binding proteins [74,75]. Analysis of these transcripts using the Discriminative Regular Expression Motif Elicitation (DREME) Tool from the MEME-suite did not identify significant motifs [76]. Future studies will be needed to uncover mechanisms that license mRNAs as RNautophagy substrates, and the power of both experimental and bioinformatic approaches will be facilitated by discovery of additional mRNAs that are degraded by this pathway. Our study shows that the RNautophagy pathway can cause remodeling of the proteome — KD of either RNASET2 or SIDT2 increases abundance of UCHL3, PFN2, and PRDX3 proteins. In the case of SIDT2 KD, the corresponding mRNAs are unable to access the lysosomal lumen and remain in the cytoplasm where they can be translated. It is less clear how RNASET2 KD increases protein abundance. However, a recent study provided evidence that SIDT2 functions bidirectionally to facilitate movement of RNAs across the lysosomal membrane [77]. We hypothesize that when RNASET2 substrates accumulate in lysosomes of RNASET2 KD cells, they can be trafficked via SIDT2 back to the cytoplasm for translation. Lipotoxicity is a pathological stimulus that incites oxidative and ER stress and ultimately results in cell death. Changes in mRNA abundance in hepatocytes, cardiomyocytes, and pancreatic beta-cells exposed to lipotoxic conditions reflects cellular responses that serve to protect against cellular damage and others that promote cellular demise [[78], [79], [80]]. Regulation of gene transcription at the level of epigenetic modifications and altered function of transcription factors is well understood to contribute to changes in gene expression in lipotoxicity [[81], [82], [83]]. Our findings establish that decreased gene expression early during lipid overload can be driven by enhanced degradation of specific mRNAs. While expression of the anti-oxidant peroxiredoxin proteins have been shown to afford protection in cell culture and animal models of lipotoxicity [[60], [61], [62], [63], [64]], our study uncovers roles for UCHL3 and PFN2 as regulators of lipotoxic cell death. CRediT authorship contribution statement Shuiling Zhao: Writing – review & editing, Writing – original draft, Investigation, Formal analysis, Conceptualization. Stevens Bontemps: Writing – review & editing, Investigation. Ryan M. Nottingham: Writing – review & editing, Methodology, Formal analysis. Jun Yao: Writing – review & editing, Visualization, Methodology, Formal analysis. Mariana Acuña: Writing – review & editing, Investigation. David E. Cohen: Writing – review & editing, Supervision, Formal analysis. Alan M. Lambowitz: Writing – review & editing, Supervision, Methodology, Formal analysis. Jean E. Schaffer: Writing – review & editing, Writing – original draft, Supervision, Project administration, Funding acquisition, Conceptualization. Financial support This work was supported by grants from 10.13039/100000002NIH to JES (5R01DK064989, 5R01DK137511), to DEC (5R01DK048873, 5R01DK056626, 2R01DK103046, 5R01DK133988), and to AML (R35GM136216), and by the 10.13039/100007878Joslin Diabetes Research Center Flow Cytometry and Molecular Phenotyping and Genotyping Cores (5P30DK036836). Declaration of competing interest This work has not previously been published and is not under consideration for publication elsewhere. A.M.L. is an inventor on patents owned by the University of Texas at Austin for TGIRT enzyme and other stabilized reverse transcriptase fusion proteins and methods for non-retroviral reverse transcriptase template switching.