Generation and verification of the ATF4 knockout specifically in proximal tubules

Gamma glutamyltransferase 1(Ggt1) is abundantly and selectively expressed in the PTs of the kidneys, particularly in PT segment 3 (Kidney Interactive Transcriptomics (KIT), https://humphreyslab.com/SingleCell/). We engineered a murine transgenic (Tg) line with a portion of the Ggt1 promoter driving CreER [22] (GCER) for tamoxifen-inducible deletion of any floxed gene in the PTs at any desired age of mice (Fig. 1A). Crossing this Tg line with ATF4fl/f/l mice resulted in the GCERA line. After tamoxifen injections on 2 consecutive days, these GCERA mice exhibited selective ATF4 deletion (GCREA∆) in the PTs (Fig. 1A, B). ATF4 protein was also expressed in PTs of WT, but not in GCREA∆ kidneys (Fig. 1C), demonstrating ATF4 specific deletion in PTs.

Fig. 1

Generation and verification of the ATF4 knockout specifically in proximal tubules. A Generation of GgtCreER; ATF4 knockout (KO) (GCERAΔ) mice. B Genotyping with DNA extracted from the kidney and liver (as negative reference) and PCR as described in the “Methods” section; C Staining of kidney tissues with ATF4 antibody. WT: wild type; GCERA: ggtCreER; ATF4fl/fl; GCERAΔ: ggtCreER; ATF4fl/fl after treatment with tamoxifen; K: kidney; L: liver

Major changes in mRNAs and proteins after proximal tubule ATF4 deletion

Since ATF4 is a transcription factor, we first asked whether ATF4 deletion in PTs resulted in any changes in transcript levels. We extracted total RNA from kidney cortices and conducted genome-wide deep mRNA sequencing (mRNA-seq). We used kidney cortices because PT cells comprise the majority (about 60–70%) of the cortices and thus, the sequencing data will be representative of PT cell mRNAs [23]. The quality of RNA-seq data from 10 samples is summarized (Supplementary Table 2). Principal component analysis (PCA) shows that transcripts of WT and GCERAΔ samples are closely clustered and that these two groups are well separated for the majority of transcripts (Fig. 2A), indicating high reproducibility and the ability to distinguish between the transcript profiles of these two groups.

Fig. 2

Deletion of ATF4 in proximal tubules caused genome-wide transcript changes. A Principal component analysis (PCA) of mRNA-seq data; B volcano plots of -log10(padj) versus log2(fold change) of gene expression in GCERAΔ over WT cortices. padj and log2(fold change) were calculated using R package DESeq2. Significantly changed genes are in blue, purple, and red. Genes significantly downregulated by two-fold are in blue. Genes significantly upregulated by two-fold are in red; C GSEA generated heatmap of top 50 features for each phenotype; D pathways generated by BioJupies; E significantly increased Slc genes by two-fold, and F significantly decreased Slc genes by two-fold

We detected 12,804 mRNAs with FPKMs of ≥ 1, 31.2% of which were significantly changed in GCERAΔ vs. WT cortices. Major mRNA changes are shown in the volcano plot (Fig. 2B). Gene Set Enrichment Analysis (GSEA) generated a heatmap of the top 50 altered mRNAs, with half of them enriched in WT and the other half enriched in GCERAΔ cortices (Fig. 2C). Gene Ontology analysis of GCERAΔ compared to WT cortices with BioJupies [24] shows that 5 out of 15 upregulated and 4 out of 15 top downregulated molecular function pathways are transporter systems (Fig. 2D). Collectively, ATF4 PT deletion caused major changes in transporter mRNAs (Fig. 2B–D). We therefore focused on transporters. Detailed analysis revealed that 288 transporter genes were detected, 50 of which were significantly increased and 75 decreased in GCERAΔ versus WT. The transcripts increased by ≥ two-fold and those decreased by ≥ two-fold in GCERAΔ compared to WT cortices are shown in Fig. 2E, F, respectively. Among transporter transcripts expressed at higher levels in GCERAΔ compared to WT cortices are Slc22a28, Slc22a30, and Slc7a13 (Fig. 2E). Among transporter transcripts with lower expression in GCERAΔ compared to WT are Slc7a12, Slc22a29, and Slc17a2 (Fig. 2F).

To investigate whether the ATF4-deletion associated changes in transcripts could be correlated with changes in proteins, we conducted proteome-wide proteomics studies. With filtration of the peptides ≥ 2, the total number of proteins detected in the cortices was 5471, with 2012 proteins (38.7%) significantly changed. Like the PCA of mRNAs, the PCA of proteins shows close clustering in WT and GCERAΔ cortices, respectively, and these two groups are well separated (Fig. 3A). The volcano plot (Fig. 3B) shows that transporters represent the major class of proteins that exhibit changes, also indicated as top-ranked protein enrichments in the heatmap generated by the GSEA (Fig. 3C). Detailed changes in some proteins are shown in Fig. 3D, E.

Fig. 3

Deletion of ATF4 in proximal tubules caused proteome wide protein changes. A Principal component analysis (PCA) of proteomics data; B volcano plots of -log10(padj) versus log2(fold change) of protein levels of GCERAΔ over WT. padj and log2(fold change) were calculated using R package DESeq2. Significantly changed proteins are in blue, purple, and red. Proteins significantly downregulated by two-fold are in blue. Proteins significantly upregulated by two-fold are in red; C GSEA generated heatmap of top 50 features for each phenotype; D significantly increased Slc proteins, and E significantly decreased Slc proteins

PT ATF4 deletion affects transporters for inorganic cations/anions and amino acids/oligopeptides in the kidney cortex

Both transcriptomics and proteomics studies demonstrate that the major changes caused by ATF4 deletion are in transporters. Thus, we uploaded all transporter mRNAs and proteins into the DeepVenn program [25] and performed correlation analysis. Almost all detected transporter proteins are translated from the detected mRNAs (Fig. 4A), and 80% of the shared mRNAs are positively correlated with their corresponding proteins (Fig. 4B).

Fig. 4

Deletion of ATF4 significantly affected amino acid transporters (AATs) and Slc22 family transporters at both mRNA and protein levels. A Venn diagram of all detected transporter mRNAs and proteins; B correlation between shared 123 transporter mRNAs and proteins; C pathways generated by Ingenuity Pathway Analysis (IPA) of all transporter mRNAs. Positive Z score is in red, negative Z-scores are in blue; D significantly increased AATs; E significantly decreased AATs; F significantly increased Slc22 family transporters; G significantly decreased Slc22 family transporters. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.0001

Since many types of transporters are expressed in the kidney [26], we conducted Ingenuity Pathway Analysis (IPA) of all 288 transporters. The pathways for transport of inorganic cations/anions and amino acids/oligopeptides exhibited the greatest changes (Fig. 4C). Therefore, we examined these amino acid transporters (AATs) in more detail. Of 45 detected AATs, Slc7a13 was the only one that showed increases at both mRNA and protein levels in GCERAΔ compared to WT (Fig. 4D). Slc7a13 (formerly AGT-1) is localized on the apical side of the PT [27]; Slc7a13 transports glutamate (Glu), aspartate (Asp), and cystine from urine to blood [28]. Compared to the number of mRNAs/proteins increased by the ATF4 deletion, many more AAT mRNAs/proteins were decreased in GCERAΔ compared to WT (Fig. 4E). The mRNA (Slc7a12) showing the greatest decrease was not detected in our proteomics analysis.

Additionally, some members of the Slc22 family were changed by ATF4 deletion. Twenty-one Slc22 family genes were detected. Six of eight transcripts increased in GCERAΔ compared to WT were also increased at the protein level in GCERAΔ compared to WT (Fig. 4F). One of these, Slc22a2, is located on the PT basolateral side [29]. One of Slc22a2’s substrates is creatinine [30], and another substrate, notably, is cisplatin [31]. Increased expression of Slc22a2 in the ATF4 knockout PTs could potentially lead to transport of more creatinine from the blood to the urine, increasing urine creatinine. Among the Slc22 family transcripts decreased in GCERAΔ compared to WT, only Slc22a8 showed lower mRNA and protein levels (Fig. 4G).

PT ATF4 deletion alters metabolite profiles in serum, kidney, and urine

Since ATF4 deletion in PTs changed large numbers of proteins, some of which carry out metabolic reactions, we hypothesized that the PT ATF4 deletion could alter renal metabolism. We took an untargeted approach, profiling all metabolites in all three compartments, the blood/serum, kidney, and urine, by LC/MS. PCA shows that the metabolites in serum are closely clustered in WT vs. GCERAΔ cortices and that these two groups are separated by principal component-1 (PC-1) of 42.3% of metabolites (Fig. 5). Greater separations between GCERAΔ and WT are found in kidney (Fig. 5E) and urine (Fig. 5I). These PCAs demonstrate that there are metabolite differences between WT and GCERAΔ in all three compartments.

Fig. 5

Deletion of ATF4 in proximal tubules caused changes in metabolite profiles in the serum, kidney, and urine. A, E, I Principal component analysis (PCA) of metabolites in the serum, kidney, and urine; B, F, J volcano plots of -log10(padj) versus log2(fold change) of metabolites of GCERAΔ over WT. Significantly reduced metabolites are in blue; significantly increased metabolites are in red. C, G, K Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of metabolites in the serum, kidney, and urine generated by MetaboAnalyst; D, H, L Small Molecule Pathway database (SMPDB) pathways of metabolites in the serum, kidney, and urine generated by MetaboAnalyst

The serum volcano plot shows that all significantly changed metabolites (23/400), except one, are increased in GCERAΔ compared to WT (Fig. 5B). One of the metabolites increased in GCERAΔ cortices is Asp. More metabolites (85/652) were altered in the kidneys of GCERAΔ vs WT than in serum or urine (Fig. 5F). Contrary to the changes in serum, all significantly changed metabolites (41/413) in urine are reduced in GCERAΔ compared to WT, except for malic acid and urea (Fig. 5J). Urine creatinine is higher in GCERAΔ compared to WT. Overall, ATF4 deletion in PTs was associated with increased concentrations of metabolites in serum and decreased concentrations of metabolites in urine (Fig. 5B, J), suggesting that ATF4 deletion causes enhanced uptake of metabolites from urine to serum as well as increased excretion of wastes such as urea and creatinine.

Pathway analysis using the MetaboAnalyst KEGG database [32] shows that ATF4 deletion has a significant impact on many metabolic pathways. Among those are Phe-Tyr-Trp synthesis and Ala-Asp-Glu metabolism (Fig. 5C, G, K). Pathway analysis based on the Small Molecule Pathway Database (SMPDB) [33] shows similar impacts on pathways such as Ala metabolism, Asp metabolism, and the malate-Asp shuttle in GCERAΔ vs WT (Fig. 5D, I, L).

PT ATF4 deletion increased creatine, an energy source for muscle partially supplied by the kidney, in serum from GCERAΔ mice, but decreased creatine in urine (Fig. 6A). In contrast, creatinine, a chemical waste product of creatine, was decreased in serum but was significantly increased in the kidneys and urine of GCERAΔ compared to WT (Fig. 6B). Urea, excreted in the urine, is a breakdown product of proteins and/or amino acids [20], whereas uric acid is primarily a breakdown product of nucleotides such as DNA or RNA [34]. In mouse, uric acid is oxidized to allantoin and excreted in urine [35], which is different from what occurs in humans. ATF4 deletion in PTs caused increased excretion of waste products from proteins and/or amino acids (Fig. 6C) and a trend toward a reduction in waste products from nucleotides (Fig. 6D, E).

Fig. 6

Deletion of ATF4 in proximal tubules affects excretion of wastes and reabsorption of nutrients. A–M, metabolites in the serum, kidney, and urine analyzed by metabolomics studies. N–U, markers of renal functions measured by spectrometry. Transporters in red and blue are increased and reduced in GCERAΔ compared to WT, respectively. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.0001

The kidney is one of the tissues in which the highest consumption of glutamine (Gln) occurs [36]. Gln is the most abundant, naturally occurring non-essential amino acid [36]. Gln is essential for many cellular functions, including the synthesis of nucleotides and non-essential amino acids [37]. ATF4 deletion in PTs reduced Gln levels in the kidney by almost four-fold and increased Gln in serum by two-fold (Fig. 6I). A similar trend occurred for Glu (Fig. 6H). Glu can be converted to Ala and Asp, which are precursors for the synthesis of several other amino acids [38]. Increased Glu probably contributes to the increased Asp in GCERAΔ blood compared to WT (Fig. 6F). Asp can also be produced by hydrolysis of asparagine, which is also increased in serum but decreased in the urine of GCERAΔ compared to WT (Fig. 6G). PT ATF4 deletion also affected arginine levels (Fig. 6J), but not those of proline or Ala (Fig. 6K–M). Additional amino acids affected by PT ATF4 deletion are shown in Supplementary Fig. 1.

PT ATF4 deletion affects glutamine metabolism in primary tubule cells

To confirm the effects of PT ATF4 deletion on nutrient metabolism, we isolated primary tubule cells from cortices of GCERAΔ and WT mice (Fig. 7A) and conducted [U13C5]glutamine tracing in those cells. [U13C5]glutamine uptake and metabolism were stopped at 2 h and 24 h, respectively, and intracellular metabolites were analyzed by LC/MS. Almost all detected intracellular glutamine contained 13C (Fig. 7C), and all five carbons were 13C (Fig. 7D). However, [U13C5]glutamine level was much lower in GCERAΔ than in WT cells, most likely resulting from lower expression of glutamine transporters (Slc38a1,2,4,7) (Fig. 4E). Glutamine was converted to glutamate, which could be metabolized to different metabolites (Fig. 6B). One of the major pathways is metabolism to α-ketoglutarate (α-KG), which subsequently goes through the TCA cycle. While there is more α-KG with 13C (Fig. 7G) in GCERAΔ cells, the amount of α-KG with 5 13Cs was lower in GCERAΔ cells (Fig. 7H), i.e., α-KG directly generated from glutamine was lower in GCERAΔ than WT. Alpha-KG was subsequently metabolized to 4-C metabolites in the TCA cycle. All 4-C metabolites were lower in GCERAΔ than WT cells (Fig. 7I–N, Q,R). Aspartate converted from oxaloacetate was also lower in GCERAΔ than WT cells (Fig. 7O, P). Glutamate can be converted to proline. Though the contribution of incorporated proline to the total proline is minimal, the incorporated proline from the glutamine tracer was much lower in GCERAΔ than WT (Fig. 7S, T).

Fig. 7

Deletion of ATF4 in proximal tubules affects glutamine metabolism in primary tubular cells. A, diagram of isolation of primary tubule cells from kidney cortices; B, diagram of uptake and metabolism of [U13C5]glutamine; C–T, abundances of metabolites incorporated or non-incorporated with 13C isotopes and their corresponding fractional incorporations at 2 h and 24 h after cells were fed with [U13C5]glutamine. Transporters in blue in B are reduced in GCERAΔ compared to WT. α-KG, α-ketoglutarate. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.0001. Orange stars indicate significance for comparisons of abundances incorporated with 13C isotopes between GCERAΔ and WT; black stars indicate significance for comparisons of abundances non-incorporated with 13C isotopes between GCERAΔ and WT

Glutamine transporters (Slc38a1,2,4) are located on the basolateral sides of proximal tubules [28], taking glutamine from the blood into cells. The lower intracellular abundance of glutamine metabolites suggests that fewer nutrients are inside PT cells to be filtered out and that more nutrients are retained in the circulation in the GCERAΔ than WT. These tracing results are consistent with metabolomics data (Fig. 6F–K).

ATF4 deletion affects kidney function

To assess if the effects of ATF4 PT deletion are manifested at the physiological level, we measured kidney functions by placing mice in metabolic cages and recording their water intake and urine output. We collected their urine and blood for measurements of albumin and creatinine in both serum and urine, and of blood urea nitrogen (BUN) in serum. We also included two additional groups (GCER and GCERA) as controls to assess if the transgenes themselves exerted any effects on renal functions. As anticipated, there were no significant differences among the WT, GCER, and GCERA cohorts for all measurements (Fig. 7). While we observed no differences in water intake (Fig. 7N) and urine output (Fig. 7P), urinary creatinine was doubled in GCERAΔ compared to WT (Fig. 7Q). Conversely, serum creatinine was lower in GCERAΔ vs WT (Fig. 7S). These changes in urinary and serum creatinine are consistent with those obtained from metabolomics studies (Fig. 7B). While slightly lower urine volumes in the GCERAΔ mice could contribute to their higher urine creatinine levels, increased export of creatinine, likely resulting from higher expression of the creatinine transporter (Slc22a2) [30] (Fig. 4F) on the basolateral side of the PT [29, 39], could explain the higher urine creatinine in GCERAΔ compared to WT. We observed no differences in albumin in either the urine or blood among these four groups (Fig. 7P, R). PT ATF4 deletion did not cause changes in BUN (Fig. 7T) nor in the BUN to creatinine ratio (Fig. 7U), which is also consistent with the serum urea levels measured from the metabolomics studies (Fig. 7C).