NSUN7 is a catalytically inactive RNA m⁵C methyltransferase essential for sperm flagellum assembly

Abstract

RNA 5-methylcytosine (m⁵C) is a widespread RNA modification mainly catalyzed by the NSUN family, whose dysfunction is linked to various diseases. Among NSUN1-7, all possessing motifs IV and VI essential for RNA m⁵C formation, NSUN7’s catalytic activity and physiological role remain unclear, despite its known association with male infertility. Here, we show that NSUN7 is predominantly enriched in adult mouse testes, particularly in elongated spermatids. Through transcriptome-wide m⁵C mass spectrometry and ultrafast bisulfite sequencing, we found that Nsun7 knockout has no impact on RNA m⁵C modification. Mechanistically, compared to NSUN1-6, NSUN7 lacks S-adenosylmethionine (SAM) binding ability due to sequence variations in motif IV, where the key SAM-binding Asp in motif IV is replaced by Leu, thereby abolishing its catalytic activity. Nsun7 deficiency impairs sperm progressive motility, accompanied by defects in axonemes and mispositioning of longitudinal columns. Single-cell RNA sequencing shows that Nsun7 knockout decreases the levels of a cohort of mRNAs related to cilium organization in elongated spermatids. Collectively, our findings resolve the long-standing question of NSUN7’s enzymatic role and reveal a non-catalytic regulatory function, providing a more complete understanding of NSUN family biology.

The NSUN family, comprising seven members, is the primary RNA m5C methyltransferase in higher eukaryotes. Here, the authors demonstrate that NSUN7 is a catalytically inactive RNA m5C protein due to sequence variations in its motif IV. Instead, the authors reveal a noncatalytic, regulatory role of NSUN7 in sperm flagellum assembly.

Introduction

Currently, more than 170 types of chemical modifications have been found on various RNA species¹, greatly expanding the functional diversity of RNA. RNA modifications not only affect the structure of RNA², but also regulate RNA localization^3,4, processing^5,6, stability^7,8, and participate in multiple biological processes, including protein translation⁹, nutrient metabolism¹⁰, and stress responses¹¹. Notably, dysregulation of RNA modifications is closely related to a range of human diseases^12,13, such as neurological disorders¹⁴, metabolic syndromes¹⁵, and male infertility¹⁶.

Among these diverse RNA modifications, 5-methylcytosine (m⁵C) is one of the most prevalent RNA modifications, and is of particular interest due to its abundance across a wide variety of RNA species, including tRNA, rRNA, and mRNA^1,17,18. RNA m⁵C is catalyzed by DNMT2, a member of the DNA methyltransferase (DNMT) family, and by the NOL1/NOP2/Sun domain (NSUN) methyltransferase family¹⁷. Both DNMT2 and the NSUN family use S-adenosylmethionine (SAM) as the methyl donor to catalyze RNA m⁵C modification, highlighting the critical role of SAM binding in the catalytic process. In higher eukaryotes, the NSUN family comprises seven members, named NSUN1-7. All seven members have the canonical NSUN domain, which contains both an RNA m⁵U-like motif (motif VI) and a DNA m⁵C-like motif (motif IV)^19,20. The Cys residue in motif VI serves as a nucleophile, and the Cys residue in motif IV acts as a general base to initiate product release^19,20. The RNA m⁵C catalytic activity of the NSUN domain also relies on the collaboration of several amino acid residues located within these two motifs, such as Asp, Ala, and Pro in motif IV^19,20. Structural and functional analyses revealed that Asp in motif IV is indispensable for RNA m⁵C formation, as substitution of Asp with Asn in motif IV abolishes both SAM binding and enzymatic activity of NSUN6, ultimately leading to autosomal recessive intellectual disability²¹.

To date, the RNA substrates of most NSUN family members have been extensively characterized. NSUN1 is responsible for a specific m⁵C modification site on cytoplasmic 28S rRNA^22,23. NSUN2 can modify mRNA, vault RNA, and cytoplasmic and mitochondrial tRNA^24–27. NSUN3 catalyzes methylation of mitochondrial tRNA^28,29. NSUN4 was described as methylating mitochondrial 12S rRNA and other mitochondrial RNAs^30,31. NSUN5 modifies cytoplasmic 28S rRNA and mRNA^32,33. NSUN6 is able to methylate cytoplasmic tRNA and mRNA^20,34–36. However, the enzymatic activity and RNA substrates of NSUN7 remain controversial^37,38.

The defects in Nsun7 result in impaired sperm motility and male infertility in mice³⁹. In humans, mutations such as the C26232T transition, T26248G transversion in exon 7, and an A11337 deletion are tightly associated with reduced sperm motility, and can serve as infertility markers in Iranian asthenospermic men^40,41, highlighting the importance of NSUN7 in spermatogenesis and sperm maturation. Moreover, NSUN7 is closely linked to the development of various cancers and neurological diseases^38,42–44. Nevertheless, the unresolved controversy surrounding the RNA m⁵C catalytic activity and RNA substrates of NSUN7 hinders progress in elucidating its physiological and pathological functions.

During spermiogenesis^45,46, spermatogonia undergo mitosis to produce primary spermatocytes, which then enter meiosis I to form secondary spermatocytes. Following meiosis II, these cells differentiate into round spermatids. Through a series of remodeling and maturation steps, round spermatids transform into elongated spermatids. Subsequently, elongated spermatids enter the epididymis, and further mature into fully functional spermatozoa. Central to the spermatozoa tail is a microtubule-based axoneme structure composed of a characteristic ‘9 + 2’ arrangement of microtubules^47,48. Structurally, spermatozoa consist of several distinct parts: the head, which contains the nucleus and is capped by the acrosome; the neck, which connects the head to the tail; and the tail, also known as the flagellum, subdivided into three main segments: the midpiece, principal piece, and endpiece⁴⁹. The midpiece houses numerous mitochondria arranged in a spiral formation, forming the mitochondrial sheath that provides energy for sperm motility. The principal piece constitutes the majority of the tail length and contains two longitudinal columns (LCs) within the fibrous sheath. Finally, the endpiece, the shortest segment of the tail, continues the extension of the axoneme. Together, these components ensure the ability of spermatozoa to navigate through the female reproductive tract and achieve fertilization. Defects in sperm structure or mitochondrial dysfunction commonly lead to sperm motility disorders and male infertility^50,51.

In this study, we showed that NSUN7 is primarily localized in adult mouse testes, especially in elongated spermatids. Through m⁵C mass spectrometry (MS) and ultrafast bisulfite sequencing analyses of different types of RNA from testicular tissues, we unexpectedly found that Nsun7 knockout has no impact on the RNA m⁵C modification level. Mechanistically, we showed that NSUN7 is incapable of binding SAM, indicating that NSUN7 is an RNA m⁵C catalytically inactive protein. This catalytic deficiency arises from sequence variations in motif IV of NSUN7 compared to the conserved residues observed in NSUN1-6, particularly the substitution of the critical SAM binding residue Asp with Leu in this region, thereby abolishing SAM binding. In Nsun7 knockout mice, the progressive motility of sperm was severely impaired, accompanied by axonemal defects and mispositioning of LCs. Moreover, single-cell RNA sequencing of testes revealed that Nsun7 knockout resulted in a significant downregulation of a cohort of mRNAs in elongated spermatids, particularly transcripts associated with cilium organization. In summary, our findings demonstrate that NSUN7 is the only RNA m⁵C catalytically inactive member in the NSUN family, and reveal its critical role in regulating sperm axoneme structure assembly and maintaining the position of LCs.

Results

NSUN7 is an evolutionarily conserved protein in vertebrates and exhibits testis-specific enrichment in the mouse

Using sequence alignment analysis (Supplementary Fig. 1), we found that NSUN7 is evolutionarily conserved among vertebrates, and exhibits high sequence similarity across species, including human, mouse, Bos taurus, Macaca fascicularis, Gallus gallus, Xenopus laevis, and Maylandia zebra. Importantly, NSUN7 proteins in different species retain the two conserved NSUN family motifs: an RNA m⁵U-like motif (motif VI) and a DNA m⁵C-like motif (motif IV).

To determine the expression profile of Nsun7, we conducted RT-qPCR using multiple tissues from adult 12-week-old mice. Our results showed that the mRNA expression of Nsun7 is predominantly enriched in mouse testes (Fig. 1A). Consistently, NSUN7 protein was clearly detected in adult mouse testes using a homemade NSUN7-specific antibody (Fig. 1B). In addition, we isolated various reproductive tissues of female mice, including the ovary, fallopian tube, uterus, and vagina. In contrast, NSUN7 protein was too low to be detected in these tissues with testis serving as the positive control (Supplementary Fig. 2). Further analysis of Nsun7 mRNA expression at different developmental stages of mouse testes showed that Nsun7 transcription begins at approximately postnatal day 28 (Fig. 1C), and this expression pattern was also confirmed at the protein level (Fig. 1D). Notably, postnatal day 28 corresponds well to the developmental stage at which elongated spermatids initially appear in mouse testes⁵², suggesting that NSUN7 may play a functional role during the late stages of spermatid differentiation. Since Nsun7 is highly expressed in adult mouse testes, all subsequent experiments used 12-week-old mice unless otherwise specified.

Fig. 1. Characterization of the mouse NSUN7.

A RT-qPCR analysis of Nsun7 mRNA levels in tissues from 12-week-old mice. B Western blot analysis of NSUN7 protein levels in multiple tissues from 12-week-old mice. The experiment was repeated three times with similar results. C Mouse Nsun7 mRNA levels in postnatal testicular tissues isolated on specified days via RT-qPCR analysis. D Western blot analysis of NSUN7 protein levels in developing testes at specific postnatal days. The experiment was repeated three times with similar results. E A schematic diagram illustrating the knockout strategy for mouse Nsun7. Dual sgRNAs were targeted to exons 3 and 4, respectively. Dark blue boxes represent coding regions. F Western blot showing the expression of NSUN7 in WT, Nsun7^+/−, and Nsun7^−/− testes. The experiment was repeated three times with similar results. G Immunofluorescence showing the cellular localization of NSUN7 in WT and Nsun7^−/− testis sections. Scale bar, 20 μm. The experiment was repeated three times with similar results. Data in (A, C) are presented as the mean ± SD for three independent experiments. Source data are provided as a Source Data File.

To investigate the subcellular localization of NSUN7, we expressed the gene encoding C-terminally Flag-tagged NSUN7 (NSUN7-Flag) in HEK293T cells and performed immunolabeling. Immunofluorescence analysis revealed that NSUN7 is primarily localized in the cytoplasm, with minor nuclear presence (Supplementary Fig. 3A). To understand the physiological function of NSUN7 and explore its role in sperm development, we generated a Nsun7 knockout mouse model using the CRISPR-Cas9 system (Fig. 1E), and subsequently evaluated NSUN7 protein levels in wild-type (WT), heterozygous mutant (Nsun7^+/−), and homozygous mutant (Nsun7^−/−) mouse testes (Fig. 1F). To examine the subcellular localization of NSUN7 in adult mouse testes, we performed immunofluorescence staining and observed that NSUN7 protein was localized in the testicular lumen (Fig. 1G). Further analysis showed that NSUN7 colocalized with the microtubule protein acetylated tubulin (Ac-Tub) (Supplementary Fig. 3B), suggesting that NSUN7 specifically localizes to the flagella of elongated spermatids.

Thus, our results showed that NSUN7 is predominantly enriched in adult mouse testis, with specific localization to the flagella of elongated spermatids, suggesting a potential role for NSUN7 in the late stages of sperm development.

Knockout of Nsun7 does not affect m⁵C levels on different types of RNA in the mouse testes assayed by RNA MS

As a member of the NSUN family, NSUN7 harbors the canonical NSUN domain. Considering that the NSUN domain of NSUN7 retains the two catalytic Cys residues essential for RNA m⁵C formation (Supplementary Fig. 1), we intended to investigate whether NSUN7 possesses RNA m⁵C methyltransferase activity and to decipher its impact on RNA m⁵C modification. We isolated different types of RNA from adult WT and Nsun7^−/− mouse testes, and detected m⁵C modification levels through UPLC-MS/MS analysis. Total RNA samples were fractionated into two different pools: RNA smaller than 200 nt and RNA larger than 200 nt. We subsequently isolated 18S and 28S rRNA from RNA larger than 200 nt (Fig. 2A). For each RNA pool, we quantified the m⁵C levels by UPLC-MS/MS and normalized their abundance to the abundance of unmodified nucleoside A. Unexpectedly, no significant differences were observed in m⁵C modification levels in total RNA, RNA smaller than 200 nt, and RNA larger than 200 nt between WT and Nsun7^−/− mouse testes. Moreover, we also found that knockout of Nsun7 had no significant effect on the m⁵C modification level of 28S rRNA from testes (Fig. 2B). For 18S rRNA, we did detect m⁵C modifications in either WT or Nsun7^−/−mouse testes (Fig. 2B). On the other hand, we enriched mRNA from testicular total RNA using oligo dT beads to examine m⁵C modification levels (Fig. 2C). However, the m⁵C levels in enriched mRNA samples were below the detection limit of our MS analysis (Fig. 2D), likely due to the inherently low abundance of m⁵C modifications on mRNA.

Fig. 2. UPLC-MS/MS analysis of m⁵C levels in different RNA pools.

A Schema showing the isolation of different RNA types from testicular tissues, including RNA smaller than 200 nt (<200 nt), RNA larger than 200nt (>200 nt), 18S rRNA, and 28S rRNA. B UPLC-MS/MS analysis of m⁵C levels in different RNA pools from WT and Nsun7^−/− mouse testes. KO, Nsun7^−/− mice. C Schematic showing purification of mRNA from testes total RNA via oligo (dT)-conjugated beads. D Mass chromatograms of A (Q1/Q3 = 268.1/136.2) and m⁵C (Q1/Q3 = 258.1/126.1) of mRNA isolated from WT and Nsun7^−/− testes. Target peaks are indicated by triangles. The standard represents a reference sample containing A and m⁵C nucleosides. The experiment was repeated three times with similar results. E Schema showing the isolation of RNA smaller than 200 nt or RNA larger than 200 nt from testicular haploid spermatids. F UPLC-MS/MS analysis of m⁵C levels in different RNA pools from WT and Nsun7^−/− haploid spermatids. KO, Nsun7^−/− mice. All mice were 12 weeks of age. nd, not detected. Data in (B, F) are presented as the mean ± SD for three independent experiments. p values were determined by a two-tailed unpaired Student’s t-test in (B, F). ns not significant. Source data are provided as a Source Data File.

Given that NSUN7 is mainly localized in the elongated spermatids of testes (Fig. 1G), and that Nsun7 mRNA is probably highest in round and elongated spermatids⁵³, we isolated haploid spermatids including round and elongated spermatids from adult WT and Nsun7^−/− mouse testes using a method based on the principle of STA-PUT^54,55, and extracted intracellular RNAs to assess the effect of NSUN7 on RNA m⁵C levels. Total RNA samples from haploid spermatids were fractionated into two pools: RNA smaller than 200 nt and RNA larger than 200 nt (Fig. 2E). m⁵C modification levels were detected in these RNA fractions from WT and Nsun7^−/− haploid spermatids. Consistent with the results from whole testes, knockout of Nsun7 had no obvious influence on RNA m⁵C modifications in either RNA fraction from haploid spermatids (Fig. 2F).

Together, these RNA MS results show that knockout of Nsun7 does not affect m⁵C modification levels in different types of RNA from mouse testes or testicular haploid spermatids.

UBS-sequencing reveals that knockout of Nsun7 has no impact on testicular mRNA m⁵C modification nor on tRNA or rRNA

Given the inherent limitations of UPLC-MS/MS in detecting low-abundance m⁵C modification on mRNA⁵⁶, to verify the impact of NSUN7 on mRNA m⁵C modification, we isolated mRNA from adult WT and Nsun7^-/- mouse testes and performed ultrafast bisulfite sequencing (UBS-seq)⁵⁷ to profile m⁵C modification sites and levels on mRNA (Fig. 3A). In WT testes, we identified 251 high-confidence mRNA m⁵C sites (Fig. 3B; Supplementary Data 2), with a median methylation level of 0.1 (Fig. 3B), consistent with the low abundance of mRNA m⁵C modification previously indicated by our UPLC-MS/MS results. Notably, comparative analysis revealed that knockout of Nsun7 neither led to the emergence of additional modification sites, nor significantly altered mRNA m⁵C modification levels, suggesting that knockout of Nsun7 had no effect on m⁵C modification sites or their modification levels (Fig. 3C). Furthermore, we extracted the top ten mRNAs with the highest sequencing coverage, including Asrgl1, Insl3, Tcp10c, Ubb-ps, Pabpc1, Ldhal6b, Spata18, Ybx3, Atp8b3, Hk1. In all cases, the sequencing results showed no obvious changes in RNA m⁵C levels on these mRNAs (Fig. 3D).

Fig. 3. Transcriptome-wide profiling of RNA m⁵C by UBS-seq in WT and Nsun7^-/- mouse testes.

A Schematic diagram of the principle of ultrafast bisulfite RNA sequencing. B Violin plot showing the m⁵C levels at 251 detected m⁵C sites in mRNA from WT and Nsun7^−/− mouse testes (n = 3 per group, biological replicates). The horizontal line indicates the median value. C Volcano plot showing no significant difference in the m⁵C levels in mRNA from WT and Nsun7^−/− mouse testes (n = 3 per group, biological replicates). D m⁵C levels of the top ten mRNAs with the highest sequencing coverage. E, F m⁵C modifications were detected at position 3438 (E) and position 4099 (F) in 28S rRNA from WT and Nsun7^−/− mouse testes. G Box plot showing the m⁵C levels at all 192 detected m⁵C sites in small RNAs from WT and Nsun7^−/− mice testicular tissues (n = 3 per group, biological replicates). The central line represents the median; the box extends from the first quartile (Q1, 25th percentile) to the third quartile (Q3, 75th percentile); the whiskers extend to the minimum and maximum values within 1.5× interquartile range (IQR = Q3−Q1) from the box edges. H Volcano plot showing no significant difference in the m⁵C levels in small RNAs from WT and Nsun7^−/− mouse testes (n = 3 per group, biological replicates). I m⁵C modification sites and levels in tRNA-Glu-TTC-1-1 from WT and Nsun7^−/− mouse testes. J Scatter plot showing the m⁵C levels at all nine detected m⁵C sites in mitochondrial RNAs from testicular tissues of WT and Nsun7^−/−mice (n = 3 per group, biological replicates). K m⁵C levels in mitochondrial 12S rRNA from WT and Nsun7^−/− mouse testes. L Eight m⁵C modification sites and their levels in six mitochondrial tRNAs from WT and Nsun7^−/− mouse testes. All mice were 12 weeks of age. Data in (D, E, F, I, K, and L) are presented as the mean ± SD for three independent experiments. p values were determined by a two-sided Fisher’s exact test (B, C, G, H, J) or two-tailed unpaired Student’s t-test (D, L). ns not significant. Source data are provided as a Source Data File.

To comprehensively validate the impact of NSUN7 on RNA m⁵C modifications across the transcriptome, we separately isolated RNA larger than 200 nt and RNA smaller than 200 nt from WT and Nsun7^−/− mouse testes, and conducted UBS-seq analyses. In large RNAs, the highest sequencing coverage was observed for 28S rRNA, which exhibited m⁵C modifications at positions 3438 and 4099 (Fig. 3E, F and Supplementary Data 3). Notably, Nsun7 deficiency did not affect the m⁵C levels at either site (Fig. 3E, F).

In small RNAs, we identified 192 high-confidence m⁵C sites in WT testes, with a median modification level of 0.8, consistent with the high m⁵C levels in small RNAs detected by UPLC-MS/MS. Remarkably, Nsun7 knockout neither generated additional modification sites nor significantly affected the m⁵C levels of these 192 sites (Fig. 3G, H and Supplementary Data 4). The small RNA with the highest sequencing coverage was tRNA-Glu-TTC-1-1, exhibiting nearly 100% m⁵C modification at positions 49 and 50, which showed no difference in Nsun7^−/− mouse testes compared to WT controls (Fig. 3I).

For mitochondrial RNA, we identified a total of nine m⁵C modification sites, located on 12S rRNA and six tRNAs. Knockout of Nsun7 did not affect the m⁵C modification sites or their modification levels (Fig. 3J and Supplementary Data 5). We detected m⁵C modification at position 842 on 12S rRNA (position 911 at mitochondrial genome), which remained unchanged in Nsun7^−/− mouse testes compared to WT controls (Fig. 3K). Additionally, we detected eight m⁵C modification sites across six mitochondrial tRNAs (mt-tRNA-Ser2, mt-tRNA-His, mt-tRNA-Leu1, mt-tRNA-Leu2, mt-tRNA-Met, mt-tRNA-Pro), exhibiting distinct modification levels. We found that knockout of Nsun7 had no impact on m⁵C modification levels on these mitochondrial tRNAs (Fig. 3L).

In summary, combining UPLC-MS/MS and UBS-seq on comprehensive transcriptome-wide profiling of m⁵C modifications, our results demonstrate that NSUN7 is dispensable for m⁵C modifications in mouse testes across various RNA species, as neither the number of methylation sites nor their modification levels are affected by Nsun7 knockout.

NSUN7 is catalytically inactive for RNA m⁵C methylation due to a lack of SAM-binding capability

The seven NSUN members, NSUN1-7, all contain the canonical NSUN domain, consisting of two conserved motifs IV and VI. Our previous work on the structure of NSUN6/tRNA/SAM complex (PDB:5WWS) demonstrated that the amino acid residues “DAPCS” in motif IV are critical for RNA m⁵C formation²⁰. Interestingly, “DAPCS” in motif IV is highly conserved in NSUN1-6, while it is replaced with “LPRCS” in NSUN7 (Fig. 4A), suggesting that NSUN7 may have a different catalytic mechanism or enzymatic activity.

Fig. 4. Mouse NSUN7 fails to bind SAM.

A Sequence alignment of the mouse NSUN protein family. The conserved motifs IV and VI are highlighted and marked with boxes. B Schematic representation of mouse NSUN7. NTD, N-terminal domain. The intrinsically disordered regions of mouse NSUN7 were predicted by IUPred3 software. C Structural model of AlphaFold-predicted mouse NSUN7. NTD and NSUN domains are marked. D The critical Asp or Leu residue in motif IV and its role in recognition, SAM, and the targeted cytosine in the structural superposition of NSUN6/tRNA/SAM (PDB:5WWS) (cyan) and AlphaFold-predicted mouse NSUN7 structure (deep blue). E SDS-PAGE analysis of the purified NSUN domain of mouse NSUN7, as indicated. The experiment was repeated three times with similar results. F Purified mouse NSUN domain was analyzed by gel filtration chromatography on a Superdex 200 increase size exclusion chromatography column. The elution volumes of the mouse NSUN domain and the standard protein are marked above the graph. The experiment was repeated three times with similar results. G Detection of the SAM-binding affinity of the purified NSUN domain measured by ITC, using buffer as a negative control. The experiment was repeated three times with similar results.

Sequence analysis and structure prediction demonstrated that the 1–188 amino acid residues form the N-terminal domain of mouse NSUN7²⁰, while 208–518 amino acid residues constitute the NSUN domain (Fig. 4B, C). The C-terminal extension domain is predicted to be intrinsically disordered (Fig. 4B) by IUPred3 software⁵⁸. Our previous study showed that the side chain carboxyl group of Asp in motif IV forms hydrogen bonds with the N3 and N4 nitrogen atoms of the targeted cytosine (C), as well as with the amino group of SAM²⁰. These interactions precisely orient the base in the active site and facilitate bond formation by transient protonation of the endocyclic N3 of C (Fig. 4B). When Asp is mutated to Ala, NSUN6 fails to bind SAM and completely loses its methyltransferase activity²⁰. Furthermore, a missense mutation of Asp to Asn in motif IV abolishes SAM-binding capacity, resulting in a loss of enzymatic activity of NSUN6 and contributing to autosomal recessive intellectual disability²¹, indicating the indispensable role of Asp in motif IV for RNA m⁵C catalysis. In mouse NSUN7, the corresponding substitution of this Asp with Leu in motif IV may disrupt these precise interactions (Fig. 4D), potentially altering the enzymatic activity of NSUN7 compared to other members of the family.

To verify this hypothesis, we attempted to purify the full-length mouse NSUN7 protein to test its ability to bind SAM. However, despite optimizing various conditions, we found it difficult to obtain mouse NSUN7 protein in its full-length form. We therefore purified the NSUN domain of mouse NSUN7, which was obtained with high purity and confirmed to exist as a monomer in solution (Fig. 4E, F). Subsequently, we performed ITC to evaluate its SAM-binding capability. Intriguingly, we found that the NSUN domain of mouse NSUN7 does not bind SAM at all (Fig. 4G).

To investigate whether the substitution of Asp with Leu in motif IV is the major cause of NSUN7 lacking SAM-binding capacity, we reverted Leu378 to Asp in the NSUN domain of mouse NSUN7 and assessed the SAM-binding capacity of this mutant protein. Both wild-type and L378D mutant NSUN domains of mouse NSUN7 were purified with high purity (Supplementary Fig. 4A), and their SAM-binding capacities were examined using ITC (Supplementary Fig. 4B–D). During the initial injections of SAM into the L378D mutant NSUN domain, distinct exothermic signals were observed. Subsequently, at the 5th and 8th injections, distinct exothermic and endothermic peaks were detected, which probably resulted from SAM-induced conformational changes in the protein during titration. In the subsequent SAM titrations, the exothermic signals gradually weakened and eventually stabilized. In contrast, the wild-type NSUN domain showed no detectable heat changes throughout the titration. These results suggest that the L378D mutant NSUN domain is capable of binding SAM.

A similar substitution of Asp with Leu in motif IV was also observed in human NSUN7 (Supplementary Fig. 5A–D). Sequence alignment and AlphaFold structure prediction showed that the NSUN domain of human NSUN7 comprises 200–513 amino acid residues (Supplementary Fig. 5B, C)⁵⁹. We then purified the NSUN domain of human NSUN7 (Supplementary Fig. 5E, F) and tested its SAM-binding affinity. Consistent with the result of mouse NSUN7, the NSUN domain of human NSUN7 also failed to bind SAM (Supplementary Fig. 5G).

Collectively, our results indicate that although NSUN7 maintains the canonical NSUN domain, sequence variations in motif IV and the critical SAM-binding residue change from Asp to Leu render it incapable of SAM binding and consequently inactive for RNA m⁵C methylation, suggesting that NSUN7 is an RNA m⁵C catalytically inactive member of the NSUN family.

Nsun7 deficiency causes a sperm progressive motility defect and male infertility

Considering the correlation between NSUN7 and male infertility, we generated adult WT, Nsun7^+/−, and Nsun7^−/− mice to explore the role of NSUN7 in male reproduction. Adult Nsun7^−/− mice appeared normal and healthy, and were indistinguishable from their WT and heterozygous littermates (Fig. 5A). The body and testis weights of Nsun7^−/− mice were similar to those of littermate controls (Fig. 5B–D). To test fertility, we housed individual males (WT and Nsun7^−/−) with WT females, and individual females (Nsun7^−/−) with WT males. Strikingly, we found that all tested Nsun7^−/− male mice were completely sterile, while Nsun7^−/− female mice exhibited normal fertility (Fig. 5E), highlighting a specific requirement for NSUN7 in male reproduction. To examine whether Nsun7 deficiency caused any defect in spermatogenesis, we performed histological analysis of testes and epididymides. Similar to WT, all germ cell populations (spermatogonia, spermatocytes, round spermatids, and elongated spermatids) appeared normal in Nsun7^−/− testes (Fig. 5F). In addition, numerous spermatozoa could be observed in Nsun7^−/− epididymides (Fig. 5G). Thus, these findings indicated that NSUN7 has no impact on the process of spermatogenesis.

Fig. 5. Knockout of Nsun7 results in sperm progressive motility defects and infertility in male mice.

A, B Representative images of WT, Nsun7^+/−, Nsun7^−/− mice (A) and testes (B). C, D Mice weight (C) and testes weight (D) of WT, Nsun7^+/−, Nsun7^−/− mice. E All tested homozygous Nsun7^−/− males were infertile, but females were fertile. F Periodic acid Schiff staining of testis sections from WT and Nsun7^−/− mice. Scale bar, 100 μm. The experiment was repeated three times with similar results. G H&E staining of paraffin cauda epididymis sections from WT and Nsun7^−/− mice. Scale bar, 100 μm. The experiment was repeated three times with similar results. H Immunofluorescence staining of acrosome and mitochondrial sheath of sperm from cauda epididymis of WT and Nsun7^−/− mice. Acrosome, red; mitochondrial sheath, green; nucleus, blue. Scale bar, 5 μm. The experiment was repeated three times with similar results. I–K Sperm counts (I), motility (J), and progressive motility (K) analyses of sperm from the cauda epididymis of WT and Nsun7^−/− mice. L Sperm from Nsun7^−/− mice showed a reduced ability to fertilize wild-type oocytes in vitro. All mice were 12 weeks of age. Data in (C, D, I, J, and K) are presented as the mean ± SD for three independent experiments. p values were determined by a one-way ANOVA followed by Tukey’s multiple comparisons test (C, D) or two-tailed unpaired Student’s t-test (I, J, K). ^*p < 0.05; ^**p < 0.01; ^***p < 0.001; ns not significant. Source data are provided as a Source Data File.

To reveal the cause of infertility in Nsun7^−/− male mice, we initially examined spermatozoa collected from the cauda epididymides. However, no obvious abnormalities were observed in sperm morphology, acrosome structure, or mitochondrial sheath organization in Nsun7^−/− sperm compared to WT (Fig. 5H). Subsequently, we examined sperm count and motility using computer-assisted sperm analysis (CASA). Although sperm counts and the proportion of motile sperm in Nsun7^−/− mice were similar to those in controls (Fig. 5I, J), the percentage of progressively motile sperm dramatically decreased in Nsun7^−/− samples (Fig. 5K), suggesting that spermatozoa from Nsun7^−/− cauda epididymides showed severely impaired progressive motility. To analyze the fertilizing capacity of Nsun7^−/− spermatozoa, we conducted in vitro fertilization (IVF) assays using WT oocytes. Quantitative analysis revealed severe impairment of fertilization competence in Nsun7^−/− sperm, with only 22 out of 271 oocytes successfully fertilized (8.1% fertilization rate; Fig. 5L and Supplementary Fig. 6). These results indicated that Nsun7^−/− male mice are infertile due to severely impaired sperm progressive motility.

Together, our results demonstrate that Nsun7 is a critical gene for the development of sperm progressive motility and male fertility.

Nsun7 deficiency disrupts axonemal integrity and causes mispositioning of LCs in spermatozoa

To assess the effect of NSUN7 in sperm, we conducted ultrastructural detection of epididymal mature spermatozoa from WT and Nsun7^−/− mice using transmission electron microscopy (TEM) (Fig. 6A). We found that both WT and Nsun7^−/− sperm heads exhibited condensed structures and showed no obvious differences (Fig. 6B). These observations suggested that Nsun7 knockout does not affect the structure of the sperm head.

Fig. 6. Knockout of Nsun7 affects sperm microtubule assembly and the position of LCs.

A Schematic diagram of the structure and cross-sections of the midpiece, principal piece, and endpiece of spermatozoa. B Representative images of transmission electron micrographs of sperm heads from WT and Nsun7^−/− mice. Scale bar, 500 nm. C Representative images of transmission electron micrographs of the cross-section of midpiece, principal piece, and endpiece of sperm from WT and Nsun7^−/− mice. Scale bar, 200 nm. The experiment was repeated three times with similar results. D Percentages of normal and abnormal microtubule structure in midpiece, principal piece, and endpiece of spermatozoa from WT (n = 3) and Nsun7^−/− (n = 3) mice. Number of WT axonemes = 114 observed in midpiece, and number of Nsun7^−/− axonemes = 106 observed in midpiece; number of WT axonemes = 125 observed in principal piece, and number of Nsun7^−/− axonemes = 113 observed in principal piece; number of WT axonemes = 143 observed in endpiece, number of Nsun7^−/− axonemes = 162 observed in endpiece. E Representative images of transmission electron micrographs of sperm tails from WT and Nsun7^−/− mice. White arrows indicate LCs. Scale bar, 200 nm. The experiment was repeated three times with similar results. F Percentages of LCs distribution from WT (n = 3) and Nsun7^−/− (n = 3) mice. Number of WT LCs = 70 observed, number of Nsun7^−/− LCs = 61 observed. All mice were 12 weeks of age.

Considering that the maintenance of mitochondrial structure and function is critical for spermatozoa motility, we next investigated whether Nsun7 deficiency could affect mitochondrial function. Mitochondrial membrane potential (MMP), a sensitive biomarker reflecting mitochondrial status, was measured using JC-1 staining and revealed no significant differences between Nsun7^−/− and WT spermatozoa (Supplementary Fig. 7A). Since mitochondria are the predominant source of reactive oxygen species (ROS), we also measured ROS levels using 7-di-chlorodihydrofluorescein diacetate (DCFH-DA) staining. Notably, ROS production in Nsun7^−/− spermatozoa had no significant changes compared to WT (Supplementary Fig. 7B). Together, these results showed that knockout of Nsun7 does not impair mitochondrial function in mature sperm.

We further inspected the flagella cross-sections of WT and Nsun7^−/− spermatozoa. In the midpiece, well-defined outer dense fibers (ODFs) and the axoneme consisting of the typical ‘9 + 2’ microtubules were both observed in WT and Nsun7^−/− spermatozoa, whereas the Nsun7^−/− spermatozoa displayed more axonemal defects, including missing microtubules (Fig. 6C). Notably, the incidence of axonemal defects increased progressively from the midpiece to the principal piece and endpiece regions (Fig. 6D).

Additionally, we also observed mispositioning of LCs within the fibrous sheath of Nsun7^−/− spermatozoa (Fig. 6E). In WT sperm, LCs are typically attached to outer microtubule doublets 3 and 8. However, in Nsun7^−/− sperm, only 25% of LCs showed normal attachment, while approximately 72% of LCs were connected to microtubule doublets 2 and 8 (Fig. 6F). Microtubules provide necessary mechanical support through their organized arrangement, serving as the foundation for flagellar movement. LCs further stabilize this structure, helping maintain the proper alignment and function of the microtubules. As both axonemal microtubules and LCs are essential for flagellar stability and motility⁶⁰, these observed defects likely contribute to the severely impaired progressive motility of Nsun7^−/− spermatozoa.

Thus, our results demonstrate that knockout of Nsun7 does not affect mitochondrial function, but leads to severe flagellar microtubule defects and mispositioned longitudinal columns, which are critical for normal sperm motility.

Nsun7 depletion downregulates a cohort of mRNAs associated with cilium organization in elongated spermatids

To investigate the molecular mechanisms of Nsun7 in sperm axonemal assembly, we performed single-cell RNA sequencing (scRNA-seq) of adult testes from WT and Nsun7^−/− mice, given the testis-specific enrichment of NSUN7. A complex testicular cell type was unsupervisedly visualized using the uniform manifold approximation and projection (UMAP) algorithm (Fig. 7A), and ten cell types were identified according to the expression of canonical cell-type-specific markers as reported previously (Fig. 7B)⁶¹. These included six somatic cells, Leydig, Sertoli cells, macrophage cells, smooth muscle cells, other immune cells, and endothelium cells, and four germ cells, including spermatogonia, spermatocytes, round spermatids, and elongated spermatids.

Fig. 7. Knockout of Nsun7 results in a decrease in a cohort of mRNAs associated with microtubule assembly.

A UMAP plot of cells from WT and Nsun7^−/− mice testes, revealing ten clusters of cell types, including spermatogonia, spermatocytes, round spermatids, elongated spermatids, Sertoli cells, Leydig cells, immune cells, endothelium cells, and smooth muscle cells. B Dot plot showing the canonical marker genes of the ten clusters of cell types. C Violin plot showing the Nsun7 expression levels across ten clusters of cell types. D Volcano plot of differentially expressed gene analysis results from scRNA-seq in elongated spermatids from WT and Nsun7^−/− mice testes. The dots for significantly downregulated genes were colored in light blue, genes that do not change were marked in gray, and significantly upregulated genes were marked in red. E Gene ontology analysis of differentially expressed genes and their associated terms in biological process, cellular component, and molecular function in elongated spermatids from WT and Nsun7^−/− mice testes. Cilium organization-related genes (Cfap20, Cfap52, Cfap45, Cdc20, Wdr35, Atp6v1d, Adad1, Ccdc39, Aurka, Cep126, Tuba3a, Tuba3b) are all among the downregulated genes and marked accordingly, with genes (Psmf1, Sord, Capza3, Mlf1) that do not change used as controls. (F) RT-qPCR analysis of the above cilium organization-related genes mRNA levels in elongated spermatids, with genes (Psmf1, Sord, Capza3, Mlf1) used as controls. All mice were 12 weeks of age. Data in (F) are presented as the mean ± SD for three independent experiments. p values were adjusted for multiple testing using the Bonferroni method (D) or determined by a two-tailed unpaired Student’s t-test (F). ^*p < 0.05; ^**p < 0.01; ^***p < 0.001; ns not significant. Source data are provided as a Source Data File.

We then investigated whether knockout of Nsun7 induced changes in cellular or molecular events in different cell populations. scRNA-seq revealed that Nsun7 mRNA is expressed in spermatocytes, round spermatids, and elongated spermatids (Fig. 7C), consistent with published datasets⁵³. We identified differentially expressed genes (DEGs) in these three Nsun7-expressing germ cell types (14 DEGs in spermatocytes, 12 DEGs in round spermatids, and 129 DEGs in elongated spermatids, Supplementary Data 6). Notably, elongated spermatids exhibited the largest number of DEGs.

In Nsun7^−/− elongated spermatids, 128 genes were significantly downregulated, and 1 gene was significantly upregulated compared to WT (Fig. 7D and Supplementary Data 6). Gene Ontology (GO) enrichment analysis of these 128 DEGs showed that knockout of Nsun7 leads to downregulation of a series of genes that were most enriched in “cilium organization” biological process and “motile cilium” cellular component (Fig. 7E), aligning with the observed structural defects in sperm flagella. Intriguingly, among the enriched cilium organization-associated genes, the expression levels of Tuba3a and Tuba3b showed the most significant changes (Fig. 7D and Supplementary Data 6). To verify the above results, we isolated elongated spermatids from WT and Nsun7^−/− testes, and detected the expression levels of DEGs through RT-qPCR analysis. Our results showed that the mRNA levels of several well-studied genes involved in cilium organization, including Cfap20⁶², Cfap52⁶³, Cfap45⁶⁴, Wdr35⁶⁵, Tuba3a^66,67, and Tuba3b^66,67 were certainly downregulated in Nsun7^−/− elongated spermatids, while control genes Psmf1, Sord, Capza3, and Mlf1 showed no changes (Fig. 7F).

To further validate the role of NSUN7 in sperm flagella assembly, we performed data-independent acquisition (DIA) proteomic MS analysis to examine global translational changes in haploid spermatids from Nsun7^−/− testes (Supplementary Data 7). Compared to WT controls, 26 significantly downregulated and 19 upregulated proteins were identified in Nsun7^−/− haploid spermatids (Supplementary Fig. 8A). GO analysis of genes corresponding to the downregulated proteins showed significant enrichment in cytoskeleton-related pathways (Supplementary Fig. 8B). The cytoskeleton, comprising actin filaments, microtubules, intermediated filaments, and septins, plays a fundamental role in sperm flagella organization and cellular functions including cell motility and intracellular trafficking⁶⁸. The downregulation of cytoskeletal proteins in Nsun7^−/− haploid spermatids further supports a pivotal role for NSUN7 in sperm flagella assembly.

To further explore the molecular characteristics of NSUN7 and potential mechanisms underlying its role, we examined the effect of RNA on NSUN7. Treatment of mouse testis lysates with RNase A markedly decreased the endogenous NSUN7 protein level (Supplementary Fig. 9A). Consistently, ectopic expression of human NSUN7 in cultured cells showed that RNase A treatment reduced NSUN7 protein abundance (Supplementary Fig. 9B). Our results suggest that the association of NSUN7 with RNA contributes to maintaining its protein stability.

We next investigated how disease-associated mutations in NSUN7 affect its properties and function. Therefore, we generated plasmids carrying the reported human male infertility-associated point mutation p.S308A (corresponding to C26232T and T26248G in exon 7), and a deletion (A11337 deletion in exon 4), which introduces a frameshift resulting in a truncated protein of 156 amino acids^40,41, and expressed them in cultured cells. The p.S308A mutant exhibited markedly reduced NSUN7 protein level compared with the wild-type protein, whereas no detectable expression was observed for the A11337-deletion mutant (Supplementary Fig. 9C). Our results suggest that these infertility-associated mutations reduce NSUN7 protein levels and may affect its function. Collectively, our study reveals that NSUN7 is specifically enriched in adult mouse testes, particularly in elongated spermatids. NSUN7 is the only RNA m⁵C catalytically inactive member of the NSUN family due to sequence variations in motif IV, especially the key SAM-binding Asp in motif IV is replaced by Leu in NSUN7, thereby abolishing SAM binding. Nsun7 deficiency impairs sperm progressive motility, thus leading to male infertility. While mitochondrial function is not affected, Nsun7 knockout causes defects in axoneme structure and mispositioning of longitudinal columns. Thus, NSUN7 plays a critical role in regulating sperm axoneme structure and maintaining the position of LCs by modulating the expression of cilium organization-associated genes in elongated spermatids (Fig. 8).

Fig. 8.

Nsun7 encodes an RNA m⁵C catalytically inactive protein whose deficiency disrupts sperm flagellum assembly, resulting in male infertility.

Discussion

NSUN7 is an RNA m⁵C catalytically inactive protein conserved across vertebrates

In this study, we demonstrate that NSUN7 does not contribute to RNA m⁵C methylation in adult mouse testes, as Nsun7 knockout had no impact on m⁵C modification levels of different RNA types. Although both mouse and human NSUN7 retain the canonical NSUN domains, sequence variations in motif IV and critical SAM-binding residue change from Asp to Leu render them losing the SAM-binding capability and consequently inactive for RNA m⁵C methylation, suggesting that both mouse and human NSUN7 function as RNA m⁵C catalytically inactive proteins.

Through sequence alignment, we observed that NSUN7 is conserved in vertebrates, including Homo sapiens, Mus musculus, B. taurus, M. fascicularis, G. gallus, X. laevis, and M. zebra. All orthologs possess the canonical NSUN domains and two conserved motifs IV and VI. Of note, the Asp in motif IV is mostly replaced by Leu, and in some cases by Val or Thr in NSUN7 across vertebrates. Considering that neither Val nor Thr possesses the carboxyl side chain like Asp, these substitutions might affect the binding of NSUN7 to SAM and the targeted site C. Given the indispensable role of critical SAM-binding residue Asp in motif IV for RNA m⁵C formation^20,21, we speculated that NSUN7 is an RNA m⁵C catalytically inactive protein in vertebrates.

NSUN7 is crucial for regulating sperm axoneme assembly and maintaining the position of LCs

Human reproductive health is crucial for population continuity. Nonetheless, approximately 15% of couples globally face challenges with infertility, struggling to conceive. Among these cases, male factor infertility is responsible for approximately 40%, and its prevalence continues to rise^69,70. Alarmingly, the causes and underlying pathological mechanisms remain unexplained in nearly 50% of clinical male factor infertility cases⁶⁹, highlighting the urgent need to unravel the pathological mechanisms and develop dependable diagnostic tools and treatment strategies. The deficit in Nsun7 results in male infertility³⁹. Additionally, the polymorphism of NSUN7 is closely related to asthenospermia in human populations, including Iranian men, and it can serve as a promising marker for infertility identification^40,41, underscoring the significance of NSUN7 in spermatogenesis and sperm maturation. Here, we found that NSUN7 is predominantly enriched in adult mouse testes, particularly in the flagellum tail of elongated spermatids. Nsun7 deficiency resulted in severely reduced progressive sperm motility, potentially due to defects in sperm axoneme structure and mispositioning of LCs, while mitochondrial function remained intact. Additionally, knockout of Nsun7 led to decreased expression levels of genes related to cilium organization in elongated spermatids. To investigate the molecular mechanism underlying this regulation, we attempted immunoprecipitation (IP) and RNA-immunoprecipitation (RIP) using both commercial and homemade NSUN7 antibodies. Unfortunately, none of these antibodies were available for detecting the targeted NSUN7 protein, precluding us from identifying its RNA or protein interaction partners. This technical limitation hampers further mechanistic dissection of how Nsun7 deficiency leads to decreased expression of cilium organization-associated genes in elongated spermatids. Approaches such as examining mRNA stability, localization, or exploring signaling pathways, especially cytoskeleton dynamics of microtubule-organizing centers (MTOC) during spermiogenesis, could shed light on this in follow-up studies.

It is noteworthy that our results show that Nsun7 is highly expressed at postnatal day 28, which corresponds well to the developmental stage at which elongated spermatids initially appear in mouse testes⁵². It seems possible that NSUN7 may play a role during spermiogenesis. However, due to current limitations in antibody specificity and detection sensitivity, it remains challenging to precisely determine the subcellular localization of NSUN7 during spermiogenesis. Future studies may require improving the specificity and sensitivity of the NSUN7 antibody to perform immuno-electron microscopy or single-molecule localization microscopy to study its precise localization.

During the preparation of this manuscript, a study reported that NSUN7 binds to specific mRNAs, suggesting NSUN7 possesses RNA-binding capability⁷¹. Our results show that RNase treatment has an impact on NSUN7 protein stability, further indicating that the association of NSUN7 with RNA contributes to maintaining its protein stability. Nevertheless, it remains uncertain whether the reduced expression of cilium organization-associated genes results from this potential RNA-related role or from the severe structural defects observed in the axoneme and longitudinal columns. The severe structural defects in the axoneme and longitudinal columns might disrupt cellular processes needed for the proper storage, localization, or stability of these specific mRNAs, which may trigger a stress response affecting mRNA levels broadly or specifically^72–74. Therefore, distinguishing cause from effect remains challenging and awaits future studies employing advanced single-molecule and high-resolution RNA–protein interaction detection approaches.

The effect of RNA modifications and modifying enzymes on male reproduction

RNA modifications and associated modifying enzymes can regulate spermatogenesis and the maintenance of sperm function⁷⁵. Patients diagnosed with asthenozoospermia or teratozoospermia showed disrupted levels of RNA modifications in semen samples, including m¹G, m⁵C, m²G, and m¹A. Notably, m⁵C levels showed the most pronounced increase in patients with teratozoospermia compared with normal individuals, indicating the prominent role of m⁵C in sperm function⁷⁶. Moreover, DNMT2 and the NSUN family, as RNA m⁵C-modifying enzymes, also play vital roles in male reproduction. DNMT2 regulates the profile of sperm tRNA-derived small RNAs (tsRNAs) and mediates the transgenerational inheritance of paternal metabolic disorders to offspring through m⁵C modifications on sperm tsRNAs^77,78. NSUN2 localizes to the chromatid body in round spermatids and affects RNA processing within the chromatid body²⁶. Loss of Nsun2 specifically blocks meiotic progression of germ cells into the pachytene stage, resulting in defective spermatogenesis²⁶. Interestingly, mouse Shtap (sperm head and tail-associated protein), which plays a role in proper maintenance of sperm function, is embedded within the Nsun4 locus⁷⁹. Moreover, m⁵C-mediated splicing events significantly upregulate specific isoforms of MAEL (maelstrom spermatogenic transposon silencer), which is likely to promote spermatogenesis⁸⁰. However, in contrast to other family members, our study demonstrates that NSUN7 does not affect RNA m⁵C levels in different types of RNA in the adult mouse testes, indicating that NSUN7 regulates sperm flagellum assembly in a manner independent of RNA m⁵C modification. The precise molecular pathways through which NSUN7 exerts its function remain to be elucidated.

Methods

All animals were housed and treated according to the procedures approved by the Animal Ethics Committee of ShanghaiTech University, and all animal experiments were done in compliance with ethical guidelines and the approved protocols (ethics no: 20231030001).

Materials

Cytidine (C), adenosine (A), guanosine (G), uridine (U), m⁵C, 5’-guanosine monophosphate (GMP), ammonium acetate (NH₄OAc), Tris-base, β-mercaptoethanol (β-Me), benzonase, pyrophosphate, and phosphodiesterase I were purchased from Sigma-Aldrich Co. LLC. (St. Louis, MO, USA). Tris-HCl, tryptone, yeast extract, MgCl₂, NaCl, ATP, CTP, GTP, UTP, and isopropyl-D-thiogalactoside (IPTG) were purchased from Sangon Biotech (Shanghai, China). DNA fragment rapid purification, plasmid extraction kits, HiScript III RT SuperMix for qPCR, and ChamQ Universal SYBR QPCR Master Mix were purchased from Vazyme Biotech (Nanjing, China). KOD-plus mutagenesis kit and KOD-plus-neo Kit were from TOYOBO. Dynabeads protein G, lipofectamine 2000, and RNAiMAX transfection reagent, TRIzol, Bacterial alkaline phosphatase, Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) solution pH 7.0, T4 DNA ligase, 4’,6-diamidino-2-phenylindole (DAPI), ribonuclease inhibitor, all restriction endonucleases, polyvinylidene fluoride (PVDF) membranes and chemiluminescent substrates were obtained from Thermo Scientific (Waltham, MA, USA). EZ-trans was purchased from Life-ilab (Shanghai, China). Tissue-Tek O.C.T., bovine serum albumin (BSA), Collagenase I, and DNase I were purchased from Yeasen (Shanghai, China). TYH (Toyoda, Yokoyama, Hoshi) capacitation medium was purchased from Aibei (Nanjing, China). Enhanced immunostaining permeabilization buffer, immunostaining blocking buffer, and hematoxylin and eosin staining kit were purchased from Beyotime (Shanghai, China). Qproteome nuclear protein kits and Ni²⁺-NTA Superflow resin were purchased from Qiagen Inc. (Hilden, Germany). SAM was purchased from New England BioLabs, Inc. PCR primers were synthesized by BioSune (Shanghai, China).

Antibodies were obtained from different companies. The antibodies used in this study were as follows: Peroxidase AffiniPure Goat Anti-Mouse/Rabbit IgG (H + L) (33201ES60/33101ES60, Yeasen, Shanghai, China), anti-β-actin antibody (AC004, ABclonal, Wuhan, China), anti-Lamin A/C antibody (4777, Cell Signaling Technology, Danvers, USA), anti-α-Tubulin antibody (3873, Cell Signaling Technology), anti-acetylated Tubulin (T7451, Sigma-Aldrich), anti-Flag antibody (14793, Cell Signaling Technology), anti-GAPDH antibody (AC002, ABclonal), Alexa Fluor 488 AffiniPure Goat Anti-Mouse/Rabbit IgG (H + L) (33206ES60/33106ES60, Yeasen). Anti-NSUN7 antibody was generated through the N-terminal 1–188 amino acid residues of NSUN7 as an antigen to immunize rabbits.

Cell culture

HEK293T cells were purchased from the cell resource center of the Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China. The cells were authenticated by the provider using STR profiling. They were cultured in a 37 °C incubator with 5% CO₂ in Dulbecco’s modified Eagle’s medium (high glucose) (Corning, USA) supplemented with 10% fetal bovine serum (Gibco, USA) and were routinely checked for mycoplasma via PCR. The cells used in this study were confirmed to be mycoplasma-free. The viable cell numbers were counted by trypan blue staining assays.

Animals

All animals were housed and treated according to the procedures approved by the Animal Ethics Committee of ShanghaiTech University, and all animal experiments were done in compliance with ethical guidelines and the approved protocols (Ethics No: 20231030001). The C57BL/6 J mice were maintained in specific-pathogen-free conditions with a 12/12-h light/dark cycle, 22–26 °C, 40–70% relative humidity, below 60 decibels, and with food and water ad libitum. CRISPR/Cas9 technology was used to generate Nsun7 KO mice by targeting exons 3 and 4 of the Nsun7 gene in the C57BL/6 J background. The sgRNA sequences for generating Nsun7 KO mice were as follows:

sgRNA1: TTATGAATCCTGCATTCTTA

sgRNA2: GATCCAAGGTATGAACATGT

After zygote microinjection and embryo transplantation, genomic DNA from the tail tip of F0 mice was sequenced, and positive founders were selected for mating with wild-type C57BL/6 J mice. Genomic DNA from F1 pups was sequenced to confirm germline transmission of the edited chromosome and absence of off-target hits. Genotyping of the Nsun7 KO mice was performed by sequencing cloned PCR products based upon the following primers:

Nsun7-identify-F: GCGTCCTGGGAATCAAACCA

Nsun7-identify-R: TGGGGACACCTTCAAGACATCA

Protein expression and purification

The coding sequence of Mus musculus NSUN7 (NM_001400914.1) was amplified from cDNA, which was obtained by RT-PCR from total RNA extracted from adult mouse testicular tissues. Then, the sequence was inserted into pET22b, along with a DNA sequence encoding a C-terminal His₆-tag. Two separated truncated forms containing the genes encoding only 1–188 or 208–518 amino acid residues of NSUN7 were constructed according to the protocol provided with the KOD-plus mutagenesis kit. The constructs were verified by Sanger sequencing, and expressed in E. coli Rosetta (DE3). Recombinant NSUN7 or truncated NSUN7 proteins were purified by affinity chromatography on Ni-NTA Superflow resin. In short, the wet cells (~3 g) were suspended in 15 mL of 10 mM imidazole in buffer A (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM DTT, and 10% (v/v) glycerol) and sonicated on ice. The crude extracts were centrifuged at 40,000×g for 40 min to remove the debris and insoluble fractions. The supernatant was mixed gently with 0.5 mL Ni-NTA resin for 30 min at 4 °C and then washed with 50 mL of 20 mM imidazole in buffer A to remove nonspecific binding proteins. Binding proteins were eluted in 8 mL of 250 mM imidazole in buffer A. The eluted protein was concentrated and purified by gel filtration on a Superdex 200 increase 10/300 GL column (GE Healthcare, USA). The protein concentrations were determined using UV absorbance at 280 nm, and the molar absorption coefficient was calculated according to the sequence of each protein⁸¹.

Isothermal titration calorimetry (ITC)

ITC measurements were performed at 25 °C, using an ITC200 MicroCalorimeter (MicroCal Inc., USA). Experiments included 20 injections of 2 μL of SAM (1 mM) into the sample cell containing 80 μM NSUN7 208–518 amino acid residues truncated protein. The protein and SAM were kept in the same buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.5 mM TCEP). SAM was titrated in the same buffer that was used as a control. Binding isotherms were fitted by nonlinear regression using Origin Software version 7.0 (MicroCal Inc.). The ITC data were fitted to a one-site binding model using Origin Software version 7.0 (MicroCal Inc.).

Western blotting

Samples were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to 0.45 μm or 0.22 μm PVDF membranes, and incubated with different antibodies. After blocking with 5% (w/v) non-fat dried milk, the membranes with targeted proteins were incubated with the corresponding primary antibodies overnight at 4 °C. Membranes were then washed three times with PBST, and incubated with HRP-conjugated secondary antibody at room temperature (RT) for 30 min. After washing three times with PBST, the membranes were treated with the chemiluminescent substrates (EpiZyme, Shanghai, China), and imaging was performed using Amersham ImageQuant 800 (GE Healthcare).

Real-time quantitative PCR (RT-qPCR)

Total RNA was isolated using TRIzol according to the manufacturer’s instructions. The cDNA synthesis using total RNA as the template was performed with HiScript III RT SuperMix. RT-qPCR was performed using the relative standard curve method in QuantStudio 7 (Life Technology, USA) with ChamQ Universal SYBR QPCR Master Mix as the dsDNA fluorescence dye. The reactions were performed under the following conditions: 95 °C for 2 min; 40 cycles of 95 °C for 30 sec, 62 °C for 20 sec, and 72 °C for 30 sec; and a melting curve from 50 °C to 95 °C. The primers used for the corresponding mRNA in RT-qPCR assays are listed in Supplementary Data 1.

Quantitative analysis of tRNA modification using UPLC-MS/MS

The 200 ng of specific RNAs were hydrolyzed with 0.5 μL benzonase, 0.5 μL phosphodiesterase I, and 0.5 μL bacterial alkaline phosphatase in a 100 μL solution including 4 mM NH₄OAc at 37 °C overnight. After complete hydrolysis, the products were dissolved in acetonitrile, and subsequently applied to Ultra-Performance Liquid Chromatography-Mass Spectrometry/MS (UPLC-MS/MS). The nucleosides were separated on an Atlantis HILIC Silica column (2.1 × 150 mm, 3 μM) and then detected by a triple-quadruple mass spectrometer (AB Sciex QTRAP 6500+) in the positive ion multiple reaction-monitoring (MRM) mode. The nucleosides were quantified using the nucleoside-to-base ion mass transitions of 268.1 to 136.2 (A), 284.1 to 152.2 (G), 244.1 to 112.1 (C), 245.0 to 113.1 (U), and 258.1 to 126.1 (m⁵C). Quantification was performed by comparison to the standard curve obtained from pure nucleoside standards running in the same batch. The relative abundance of m⁵C was calculated as the ratio of m⁵C to the total A content.

Immunofluorescence microscopy

HEK293T cells were transfected with the pcDNA3.1-Nsun7-Flag plasmid. After transfection for 24 h, the cells were fixed in 4% paraformaldehyde (PFA) for 30 min and then permeabilized in 0.2% Triton X-100 for 5 min on ice. After washing with phosphate-buffered saline (PBS), fixed cells were blocked in PBS plus 0.1% Triton X-100 buffer containing 5% BSA and incubated with mouse anti-Flag antibodies at a 1:400 dilution overnight at 4 °C. The cells were then immunolabeled with AlexaFluor 488-conjugated goat anti-mouse IgG in PBS with a 1:1000 dilution for 2 h and the nuclear counterstain DAPI for 5 min at RT.

Mouse testes were fixed in 4% PFA overnight at 4  °C. After being washed in phosphate-buffered saline, the testes were perfused in 30% sucrose, and embedded in Tissue-Tek O.C.T. Compound. Tissue sections were cut at 10 μm and mounted on silanized slides. Then, the sections were permeated with enhanced immunostaining permeabilization buffer for 30 min at RT. After three washes with PBS, nonspecific binding was blocked by incubating the sections in blocking buffer for 30 min at RT. The sections were then incubated with rabbit anti-NSUN7 antibody at a 1:100 dilution overnight at 4 °C. The cells were then immunolabeled with AlexaFluor 488-conjugated goat anti-rabbit IgG in PBS with a 1:1000 dilution for 2 h. All fluorescent images were taken and analyzed using an LSM880 confocal microscope (Zeiss, Germany).

Histology

Mouse testes and epididymides were dissected and fixed in Bouin’s solution (Leica, Germany) overnight at 4 °C on a rotator, then the testes were cut into two pieces, after washing twice with phosphate-buffered saline, and dehydrated in a gradient series of different concentrations ethanol, followed by two xylene washes to remove ethanol, paraffin embedding, and sectioning at 5 μm thickness. For the histological analysis of testes or epididymides, sections were stained with hematoxylin and eosin or periodic acid-schiff after dewaxing and rehydration.

IVF

IVF was performed following a protocol from Jackson Laboratory (http://cryo.jax.org/ivf.html). In brief, ovarian stimulation was performed using standard gonadotropin-releasing hormone agonist downregulation and gonadotropin stimulation of the ovaries. Oocytes from 12-week-old mice were obtained using ultrasound-guided, transvaginal aspiration, and incubated in HTF medium (AiBei, Nanjing, China) supplemented with 15% heat-deactivated maternal sera for 4 h prior to sperm insemination. Insemination and culture were performed using standard “microdrop” techniques.

CASA

Sperm were harvested by dissecting cauda epididymides in 37 °C pre-warmed Enriched Krebs-Ringer bicarbonate medium (EKRB medium; 120.1 mM NaCl, 4.8 mM KCl, 25.2 mM NaHCO₃, 1.2 mM KH₂PO₄, 1.2 mM MgSO₄, 1.3 mM CaCl₂, supplemented with 11.1 mM glucose, 2 mM glutamine, 1× essential amino acids, 1× nonessential amino acids, 100 μg/mL streptomycin, 100 U/mL penicillin). Then, sperm-containing medium was dropped into the calibrated slide for veterinary semen analysis (Leja).

TEM

Cauda epididymal spermatozoa were incubated in TYH medium to allow dispersion, collected into a 2.0-mL tube, and then fixed in 2.5% glutaraldehyde. After primary fixation, samples were washed with 0.1 M phosphate-buffered saline, and postfixed with 1% osmium tetroxide. Dehydration was carried out in a gradient series of ethanol, and infiltrated with acetone. Ultimately, samples were embedded in Epon 812. Ultrathin sections with 70 nm thickness were prepared by ultramicrotome (Leica UC7, Leica Biosystems, Germany), and counterstained with uranyl acetate and lead citrate. Images were taken with a Zeiss GeminiSEM460 Microscope (Germany).

Isolation of mouse testicular haploid spermatids

We isolated haploid spermatids from mouse testes with a simple protocol⁸² based on the underlying principle of STA-PUT velocity sedimentation^54,55. Testes were decapsulated, and gently teased apart with forceps to obtain the seminiferous tubules. Then, the seminiferous tubules were incubated with 1 mg/mL collagenase I in 1× Krebs at 37 °C in a water bath for 10–15 min. The sedimented tubules were further digested with 0.05% trypsin containing DNase I, and gently pipetted up and down for 5–15 min before adding DMEM containing 10% FBS. Cells were pelleted by centrifugation at 500×g for 5 min, and then resuspended in 1× Krebs, and filtered with a 40 μm cell strainer to obtain a monodisperse suspension. The single-cell suspension was loaded into a 50 mL tube, which contained a discontinuous gradient from 5% to 1% BSA in 1× Krebs from bottom to top with 5 mL per layer, and then allowed to sediment by gravity for 1.5–2 h at 4 °C. The haploid spermatids were carefully collected in ~1 mL fractions in separate 1.5 mL tubes for a total of 4–5 tubes, starting from the top of the BSA gradient. Cell purity was checked under a DIC light microscope based on distinctive diameters and morphological features, and validated under an immunofluorescence microscope based on staining patterns of DAPI.

Separate different types of RNA

Total RNA from testes or haploid spermatids was extracted using TRIzol, and precipitated overnight with ethanol at −20 °C. mRNA can be captured from total RNA using beads conjugated with oligo-dT (Vazyme, China). For the separation of RNA larger than 200 nt or smaller than 200 nt, RNA Clean & Concentrator-5 (Zymo, USA) was utilized according to the manufacturer’s protocol. In order to obtain 18S and 28S rRNA, total RNA was separated by agarose gel electrophoresis, and recovered with Gel RNA Recovery kit (Zymo, USA) according to the manufacturer’s protocol.

UBS-seq library construction and sequencing

Bisulfite treatment was performed using UBS-2 buffer as described in ref. ⁵⁷. In brief, 45 µL of UBS-2 reagent was preheated to 98 °C for 5 min in a PCR instrument. Subsequently, approximately 0.1–1 μg of RNA dissolved in 5 µL of water was added to the reagent and thoroughly mixed. The mixture was then incubated at 98 °C for 9 min with the lid temperature set to 105 °C. After cooling to RT, desulphonation was performed according to the manufacturer’s protocol using the EZ RNA Methylation Kit (Zymo, USA). The quantity of converted RNA was measured using a Qubit.

For mRNA UBS-seq library preparation, the Stranded mRNA-seq Lib Prep Module for Illumina (ABclonal, China) was used. For large RNA UBS-seq library preparation, the converted RNA samples were fragmented into small RNAs by heat treatment in the presence of 5 mM MgCl₂, and then subjected to end-repair using T4 Polynucleotide Kinase (PNK). Finally, the libraries were constructed using the VAHTS Small RNA Library Prep Kit for Illumina V2 (Vazyme, China). For small RNA UBS-seq library preparation, the converted RNA samples underwent end-repair with T4 PNK and subsequently processed into sequencing libraries using VAHTS Small RNA Library Prep Kit for Illumina V2 (Vazyme, China). Libraries were all sequenced at Majorbio (China). Briefly, all libraries underwent quality control assessment using an Agilent TapeStation. Those libraries that passed quality control were sequenced on the NovaSeq X plus (Illumina, USA), generating paired-end 150 bp reads. Each group had three biological replicates. Comparative analysis of site-specific RNA m⁵C level between WT and Nsun7^−/− group was performed two-sided Fisher’s exact test to calculate p value. Differences were considered to be statistically significant when p value < 0.05.

UBS-seq data analysis

The UBS-seq data were analyzed using the previously described pipeline³⁶, with some modifications. In brief, we performed trimming of adapter sequences and low-quality reads with cutadapt⁸³ and trimmomatic⁸⁴. Paired reads and unpaired reads were then mapped separately to the C–T, G–A transformed mouse genome (GRCm39) with HISAT2⁸⁵. Unmapped and multiple-mapped reads were further mapped to a C-to-T converted transcriptome by Bowtie2⁸⁶. The transcript coordinates were adapted to match the genomic locations based on the mouse Ensembl GTF annotation. We analyzed the paired reads and unpaired reads, and merged them into the total BAM files. For large RNA UBS-seq data analysis, in addition to GRCm39, alignment references included the BK000964.3 and other rRNA variants from the SILVA database (https://www.arb-silva.de/). For small RNA UBS-seq data analysis, transcript coordinates were genomically aligned using a composite annotation system incorporating the annotations from Ensemble, RNAcentral⁸⁷, Genomic tRNA database⁸⁸, piRbase⁸⁹, and miRbase⁹⁰. We set the criteria for potential m⁵C modification site calling to have a coverage of C + T above 15 and a base quality score above 30. We used a binomial model to assess the p value for each potential m⁵C site.

Proteomic profiling by DIA-MS

Proteins were extracted from wild-type or Nsun7 KO mouse testis samples using SDT lysis buffer (4% SDS, 100 mM DTT, 100 mM Tris-HCl, pH 8.0). Samples were boiled and further ultrasonicated. Undissolved cellular debris was removed by centrifugation, and the supernatant was collected and quantified with a BCA Protein Assay Kit (Beyotime, China). Protein digestion was performed with the FASP method described in ref. ⁹¹. LC-MS/MS was performed on an Orbitrap Astral mass spectrometer coupled with a Vanquish Neo UPLC system (Thermo Fisher Scientific, USA). The DIA method consisted of a survey scan from 380-980 m/z at resolution 240,000 with AGC target of 500% and 5 ms injection time. The DIA MS/MS scans were acquired by Astral from 150–2,000 m/z with a 2 m/z isolation window and with AGC target of 500% and 3 ms injection time. Normalized collision energy was set to 25, and cycle time was 0.6 sec. The spectra of full MS scan and DIA scan were recorded in profile and centroid type, respectively. p values were determined by a two-tailed unpaired Student’s t-test. Proteins with log2(fold change) ≥ 1.0 (upregulation) or ≤ −1.0 (downregulation), accompanied by a p value < 0.05, were considered significantly differentially expressed.

Preparation of single-cell suspensions

The fresh wild-type or Nsun7 KO 12-week-old mouse testis samples were minced, and enzymatically digested using the MACS Tumor Dissociation Kit (Miltenyi, China) for 30 min with agitation, according to the manufacturer’s instructions. The dissociated cells were subsequently passed through a 70 μm and 40 μm cell-strainer and centrifuged at 300 g for 10 min. The pelleted cells were then suspended in red blood cell lysis buffer (Tiangen, China), and washed with 0.04% BSA in DPBS. Finally, the cell concentration was adjusted to 1000 cells/μL (viability ≥ 80%) in all samples.

Single-cell RNA library construction, sequencing, and data processing

Single-cell RNA-seq libraries were prepared with DNBelab C Series High-throughput Single-Cell RNA Library (MGI, China) according to the manufacturer’s instructions. Libraries were purified and quantified using the Qubit ssDNA kit (Thermo Fisher Scientific, USA) and Qsep100 (Bioptic, China), and subsequently sequenced on DNBelab C4/DNBelab TaiM4 Series Single-Cell Library Prep Set (MGI, China). PISA was used to calculate the gene expression of cells and create a gene × cell matrix for each library.

The “Seurat” R package was then used to perform single-cell RNA-seq analyses. Count matrices were merged using Seurat version 5.2.1⁹², and a percentage of mitochondrial genes was calculated. Cells containing fewer than 300 identified genes or more than 20% of reads arising from mitochondrial genes were removed. For normalization and feature selection, the Seurat functions “NormalizeData”, “FindVariableFeatures” (n.features = 2000), and “ScaleData” were used.

Two-dimensional representations were generated by using the outputs from PCA (Principal Component Analysis) and Harmony as inputs to the “RunUMAP” functions. Differential expression gene analysis was performed using the nonparametric two-sided Wilcoxon rank-sum test. Cell types were allocated to each cluster using the abundance of known marker genes as described previously⁶¹. p values of cell clusters between the WT and Nsun7^−/− group were calculated using the nonparametric two-sided Wilcoxon rank-sum test. Adjusted p values were used with the Bonferroni correction. Prior to differential gene analysis across cell clusters, we filtered for genes expressed in > 10% of the WT group (pct.1 > 0.1). From the filtered genes, only those with adjusted p value < 0.05, average log2(fold change) ≥ 1.5 (upregulation) or ≤−1.5 (downregulation) were considered significantly differentially expressed.

Statistics and reproducibility

Statistical analyses and data visualization were performed using GraphPad Prism version 10.4.1 and R version 4.3.0. Image visualization was performed using ImageJ (version 1.54). Error bars represent mean ± SD for three independent experiments. Comparisons between two groups were analyzed using a two-tailed unpaired Student’s t-test or a two-sided Fisher’s exact test, as indicated. For more groups, statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparisons test. Statistical significance was defined as p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.L. and R.-J.L. conceived and designed the research. J.L., W.-Y.Z., Z.W., Y.-Y.Z., and M.-M.H. performed the experiments. Q.-Y.L., Y.-J.Z., X.-Y.W., H.L., and M.-F.L. contributed to data analysis. J.L., W.-Y.Z., and R.-J.L. wrote the paper. R.J.L. supervised this research. All authors have read and approved the manuscript.

Peer review

Peer review information

Nature Communications thanks Shum Winnie, who co-reviewed with Shuo Shi, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are available from the corresponding authors upon request. Raw sequencing files of UBS-seq and scRNA-seq in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession: GSE297485 and GSE297919. Source data for the Figures and Supplementary Figs. are provided as a Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no conflict of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67233-2.