All plants used in this study were grown at the Changchun Agricultural Station, Northeast Institute of Geography and Agroecology, Chinese Academy of Science (CAS), China. A Gmpgl3-1 mutant was isolated from an M₂ population induced by ethylmethanesulfonate (EMS). The Gmpgl3-1 mutant was backcrossed to Williams 82 five times from 2014 to 2018 to purify the genetic background of the Gmpgl3-1 mutant. For protoplast isolation, Arabidopsis seeds of the Columbia ecotype (Col-0) were surface sterilized, vernalized, and then sown on 1/2-strength Murashige and Skoog (MS) media until the seedlings reached the four-leaf stage. Then, the seedlings were grown in pots containing peat moss and vermiculite (1/1, v/v) in a growth chamber under 150 μmol m⁻² s⁻¹ irradiance and a 14 h dark/10 h light photoperiod at 25°C, and the relative humidity was maintained at 60–75%. Leaves were collected from 3- to 4-week-old seedlings for transfection assays.

Three F₂ populations derived from a cross between the Gmpgl3-1 mutant and the Chinese soybean cultivar Hedou 12 were used to map the GmTic110a gene. DNA from 50 F₂ individuals with the Gmpgl3-1 mutant phenotype and 50 F₂ individuals with the wild-type phenotype were bulked into mutant and wild-type pools, respectively. Insertion–deletion (INDEL) markers for preliminary mapping were used according to a previously described method (Song et al., 2015). New molecular markers for fine mapping were generated; these markers are listed in Supplementary Table S4. The candidate genomic regions were identified via BSA of the F₂ population at a depth of approximately 30× using an Illumina HiSeq 2000 device (Illumina Inc., San Diego, CA, United States). The Genome Analysis Toolkit (GATK, version 3.8) was used to detect single-nucleotide polymorphisms (SNPs; Mckenna et al., 2010). Genomic regions in which Δ(SNP index) was >0.5 were selected as candidate regions.

GmTic110a homologs were identified by querying the GmTic110a sequence in the NCBI¹ database via the BLASTP program. Multiple sequence alignments were performed using ClustalX version 2.0 (Larkin et al., 2007) and were manually corrected. The obtained sequence was used as input to construct an unrooted phylogenetic tree with the neighbor-joining algorithm via the Molecular Evolutionary Genetics Analysis version 7.0 (MEGA 7.0) phylogenetic program (Sudhir et al., 2016). Bootstrap analysis was performed using 1,000 replicates. The protein motifs of GmTic110a-like genes were subsequently profiled by Multiple Expectation maximization for Motif Elicitation (MEME; Bailey et al., 2009).

To determine pigment contents, leaves of 21-day-old Gmpgl3-1 mutants and Williams 82 plants were collected and measured according to a previously reported procedure (Gregor and Marsálek, 2004). The pigment contents were calculated according to the following formulas: chlorophyll a = 13.95^*A665-6.88^*A649; chlorophyll b = 24.96^*A649-7.32^*A665; and carotenoids = (1,000^*A470-2.05^*Ca-114.8^*Cb)/245. The photosynthesis rate (Pn), stomatal conductance (Gs), intercellular CO₂ concentration (Mckenna et al.), and transpiration rate (Tr) of the leaves were measured using an LI-6400 photosynthesis system (LI-COR, Lincoln, NE, United States; Yamori et al., 2011), and the initial fluorescence (F0), maximal fluorescence (Fm), and variable fluorescence (Fv) values were measured using an OS-30p chlorophyll fluorometer (Opti-Sciences, Hudson, NY, United States). The maximum quantum yield of photosystem II (Fv/Fm) and the maximum photochemical yield of PSII (Fv/F0) were calculated as previously described (Genty et al., 1989). The plants were dark-adapted for 30 min before measurement. All the measurements involved the use of ten plants and were performed from 11:00 am to noon during the beginning of the flowering period. The operation of the machine and subsequent analysis were performed according to the manufacturer’s instructions.

Williams 82 and Gmpgl3-1 mutant plants grown for 21 days were selected. The leaves were cut into rectangular pieces that were approximately 2 mm^*1 mm, and the plant materials were vacuum fixed in 2.5% glutaraldehyde solution with 0.2 mol of phosphate buffer. The samples were postfixed for 3 h in 1% osmium tetroxide at 4°C. The samples were then treated according to previously described methods (Kowalewska et al., 2016). Ultrathin sections were obtained using an MT-X (RMC, Tucson, AZ, United States) ultramicrotome and stained with uranyl acetate for 20 min followed by lead citrate for 10 min. Observations of the samples and recording of images were performed using a Hitachi H-7650 electron microscope (Tokyo, Japan).

To obtain GmTic110a-knockout plants, the CRISPR/Cas9 gene editing system for targeted genome modification of plants was used (Shan et al., 2013). Several 20-nt single-guide RNAs (sgRNAs) highly specific for Cas9 target sites were identified using the web-based tool CRISPR-P version 2.0² (Lu et al., 2017). A pair of 24-bp long oligonucleotides (5′-GATTGCGGCGGCTGGATACGGCCT-3′ and 5′-AAACAGGCCGTATCCAGCCGCCGC-3′) specific to the GmTic110a sequence were annealed and cloned into a modified VK005-04-soU6-2-GmUbi3 knockout expression vector (Du et al., 2016). The resulting recombinant plasmid (VK005-GmTic110a) was introduced into Agrobacterium tumefaciens EHA105, which was then used to transform Williams 82 cotyledonary explants (Zhao et al., 2016). Three independent GmTic110a-knockout transgenic plants were obtained for further phenotypic analysis.

New leaves at the VE (emergence) stage; stem tips, stems, and roots at the V1 (first unrolled trifoliate leaf) stage; leaves and flowers at the R1 (beginning bloom) stage; leaves and flowers at the R2 (full bloom) stage; and leaves at the R3 (beginning of pod development) stage were collected. Total RNA was subsequently extracted using TRIzol reagent (Tiangen, Lot 118,721; China) according to the manufacturer’s instructions. The integrity of the RNA was determined through agarose gel electrophoresis, and complementary DNA (cDNA) was synthesized using 5 μg of RNA with oligo(dT)₁₈ primers and Moloney murine leukemia virus (M-MLV) reverse transcriptase (TransGen Lot N31204; China) according to the manufacturer’s protocol. Relative transcript levels of GmTic110a were analyzed through real-time quantitative PCR (qRT–PCR) on an Mx3005P instrument (Stratagene, La Jolla, CA, United States) in conjunction with SYBR Green Master Mix (Genstar Lot 9 BC01; China). The PCR parameters were 95°C for 30 s (1 cycle), 95°C for 5 s, and 60°C for 20 s (40 cycles), which was followed by a melting curve analysis at 95°C for 60 s, 55°C for 30 s, and 95°C for 30 s. The internal control gene GmActin11 (Glyma.18G290800) was used for normalization of the transcript levels of GmTic110a in the samples (Hu et al., 2009). The relative fold differences were calculated via the 2^-ΔΔCt method. Three independent biological replicates were used to confirm the expression profiles. The specific primer pairs used are listed in Supplementary Table S4.

We next sought to determine the subcellular localization of GmTic110a, GmTic110a^G114A, Gmtic110a^T805S, GmTic110a^CR1, GmTic110a^CR2, and GmTic110a^CR3 from the knockout transgenic strains. For this analysis, the full-length cDNA sequence and the mutated and knockout sequences of GmTic110a were cloned into pUC19-GFP (Zheng et al., 2017), and the resulting recombinant plasmid was transiently introduced into Arabidopsis (Col-0) protoplasts using 20% polyethylene glycol (Yu et al., 2015). The fluorescence signals were visualized using a Nikon C2 confocal laser scanning microscope (Japan) under a 488 nm excitation wavelength and 495–540 nm emission wavelengths to determine the subcellular localization of GmTic110a, GmTic110a^G114A, Gmtic110a^T805S, GmTic110a^CR1, GmTic110a^CR2, and GmTic110a^CR3. Chloroplast autofluorescence was detected at wavelengths of 488 nm (excitation) and 680–700 nm (emission). Image processing was performed with ImageJ.³ The specific primer pairs used are listed in Supplementary Table S4.

Luciferase complementation assays were performed as described previously (Chen et al., 2008), with minor modifications. The coding DNA sequences (CDSs) of GmTic110a, GmTic20, GmTic40a, and GmTic40b were cloned into either a pCAMBIA1300-NLuc or a pCAMBIA1300-CLuc vector. pCAMBIA1300-GmTic110a-NLuc, pCAMBIA1300-GmTic20-CLuc, pCAMBIA1300-GmTic40a-CLuc, and pCAMBIA1300-GmTic40b-CLuc in various combinations were transferred into Nicotiana benthamiana leaves by A. tumefaciens-mediated transformation. Agrobacterium cells with N-Luc and C-Luc vectors were resuspended in infiltration buffer (pH 5.6; 10 mm 2-(N-morpholino) ethanesulfonic acid, 10 mm MgCl₂, and 150 mm acetosyringone) to reach an optimal optical density at 600 nm in the range of 0.9 to 1. After 3 h of incubation at room temperature, the suspensions were infiltrated into the leaves of 4-week-old N. benthamiana plants, which were then cultivated for 2 days at 23°C. To inject tobacco leaves with 1 mmol of luciferin (115144–35-9, GoldBio) for measurements of luciferase activity, the leaves were maintained in the dark for 5 min. Images were captured using a chemiluminescence image analysis system (4600SF, Tanon). The sequences of the primers used are listed in Supplementary Table S4.

Co-IP assays were performed as described previously (Zhou et al., 2015; Liu et al., 2020; Wang et al., 2020), with minor modifications. The CDSs of GmTic110a, GmTic20, GmTic40a, and GmTic40b were cloned into either a pCAMBIA1300-FLAG or pCAMBIA1300-HA vector, resulting in pCAMBIA1300-GmTic110a-FLAG, pCAMBIA1300-GmTic20-HA, pCAMBIA1300-GmTic40a-HA, and pCAMBIA1300-GmTic40b-HA vectors. To measure protein–protein interactions, A. tumefaciens strain EHA105 containing pairs of these constructs together with pCAMBIA1300-GmTic110a-FLAG, pCAMBIA1300-GmTic20-HA, pCAMBIA1300-GmTic40a-HA, and pCAMBIA1300-GmTic40b-HA were coinfiltrated into the leaves of 4-week-old N. benthamiana plants. Samples (1 g each) were then collected at 3 days after infiltration, ground in liquid nitrogen and homogenized in 1.5 ml of extraction buffer (50 mm Tris–HCl [pH 7.5], 150 mm NaCl, 1 mm EDTA [pH 8.0], 0.2% [v/v] Triton X-100, 20% [v/v] glycerol, and 1× protease inhibitor cocktail [pH 7.5]). The lysates were incubated at 4°C for 30 min and subsequently centrifuged at 15,000 × g for 30 min at 4°C. After instantaneous centrifugation, the supernatants were added to 500-μl suspensions of anti-DDDDK-tag-FLAG magnetic beads (No. M185-11R, Medical and Biological Laboratories), incubated at 4°C for 4 h, and then washed 4 times with extraction buffer. The proteins were eluted from the beads with 30 μl of 1 × Protein Loading Buffer, boiled for 5 min, and then centrifuged at 8,000 × g for 1 min at room temperature. The supernatants were electrophoretically separated via 10% SDS–PAGE and transferred to a nitrocellulose membrane (No. q0600003, GE Healthcare Life Sciences). Immunoblots were performed using an anti-FLAG antibody (1:5000; No. M180-5, Medical and Biological Laboratories) for probing pCAMBIA1300-GmTic110a-FLAG and an anti-HA antibody (1/5000, No. M180-3, Medical and Biological Laboratories) for probing pCAMBIA1300-GmTic20-HA, pCAMBIA1300-GmTic40a-HA, or pCAMBIA1300-GmTic40b-HA. The sequences of the primers used are listed in Supplementary Table S4.

Compared with the wild-type Williams 82, the Gmpgl3-1 mutant showed a pale green leaf phenotype from the seedling stage to the mature stage (Figure 1A). Moreover, on the basis of their phenotypes and genotypes, we identified two allelic mutants named Gmpgl3-2 and Gmpgl3-3 from within the mutant library. The pale green leaf phenotype was observed for the Gmpgl3-2 mutant (Figure 1A), while the pale green leaf phenotype was not observed for the Gmpgl3-3 mutant (Supplementary Figure S1). We analyzed mature Williams 82, Gmpgl3-1, and Gmpgl3-2 plants separately. The results showed that the overall heights of the Gmpgl3-1 and Gmpgl3-2 mutants were reduced by 7 and 2.5%, respectively, compared with Williams 82, while the number of nodes was reduced by 17.3%. Moreover, compared with Williams 82, the number of pods per plant for the Gmpgl3-1 and Gmpgl3-2 mutants was reduced by 30.8 and 39.5%, respectively; the number of grains per plant for the Gmpgl3-1 and Gmpgl3-2 mutants was reduced by 39.5 and 39.2%, respectively; the grain weight per plant for the Gmpgl3-1 and Gmpgl3-2 mutants was reduced by 50.8 and 48.2%, respectively; and the 100-seed weight for the Gmpgl3-1 and Gmpgl3-2 mutants was reduced by 18.1 and 14.8%, respectively (Supplementary Table S1).

Because the Gmpgl3-1 and Gmpgl3-2 mutants exhibited a pale green leaf phenotype throughout the growth period, the chlorophyll and carotenoid contents of Williams 82, Gmpgl3-1, and Gmpgl3-2 leaves were measured spectrophotometrically, and the results showed that both chlorophyll a and chlorophyll b contents in the Gmpgl3-1 and Gmpgl3-2 mutants were significantly lower than those in the wild type (Figure 1B). This finding indicates that the GmTic110a gene mutation may affect the stability of Chl a/b. The chlorophyll content of the Gmtic110a mutant was reduced by 44.2%. The chloroplast fluorescence Fv/Fm values of Gmpgl3-1 and Gmpgl3-2 were significantly lower than those of Williams 82, which showed that the photosynthetic efficiency of Gmpgl3-1 and Gmpgl3-2 was lower than that of Williams 82 (Figure 1C). The photosynthetic rate (Pn) of the Gmpgl3-1 and Gmpgl3-2 mutants was reduced by 37.7 and 50.9%, respectively, compared with that of Williams 82. In addition, compared with those of the wild type, the stomatal conductance (Gs) of the Gmpgl3-1 and Gmpgl3-2 mutants decreased by 22 and 33.7%, respectively, the transpiration rate (Tr) of the Gmpgl3-1 and Gmpgl3-2 mutants decreased by 17.9 and 26.9%, respectively, and the intercellular CO₂ concentration of the Gmpgl3-1 and Gmpgl3-2 mutants increased by 17.3 and 25.6%, respectively (Figures 1D–G). Taken together, these results showed that the decrease in the chlorophyll content in the Gmpgl3-1 and Gmpgl3-2 mutants significantly affected the photosynthetic ability of the leaves of those plants.

Compared with that in the wild-type (Williams 82) chloroplasts, the number of starch grains in the chloroplasts of the Gmpgl3-1 and Gmpgl3-2 mutants increased, and the basal thylakoids became thinner (Figures 1H–J). Because there are photosynthetic pigment components on the thylakoid membrane and because photosynthesis mainly occurs within thylakoids, both the chlorophyll a and chlorophyll b contents in the mutants were significantly reduced, which led to thinning of the basal thylakoid membrane and a reduction in the photosynthesis rate.

The Gmpgl3-1 mutant was crossed with Hedou 12 to generate a segregating population for mapping the Gmtic110a gene. The F₂ segregating population comprised 537 plants: 423 wild-type plants and 114 mutant plants. A 3:1 segregation ratio was observed for the three F₂ segregating populations (χ² = 1.89, df = 1, p = 0.17; Supplementary Table S2), indicating that the Gmpgl3-1 mutant is the result of a single recessive gene. The F₂ population was used to identify the Gmpgl3-1 locus. A total of 60 INDEL markers covering all 20 chromosomes were used for mapping, and the mapping results showed that Gmpgl3-1 was restricted to a 2-Mb region (41–43 Mb) on chromosome 02 (Figures 2A,B). To finely map the Gmpgl3-1 locus, we developed 7 INDEL markers, that is, MOL3067, MOL4032, MOL3069, MOL3071, MOL3073, MOL2733, and MOL0699; the Gmpgl3-1 locus was further narrowed down to a 0.44-Mb region between 41.79 Mb and 42.23 Mb on chromosome 02, which harbors 18 annotated genes (Figure 2C). To identify the causal mutation, the DNA from 40 F₂ individuals carrying the Gmpgl3-1 mutation under homozygous conditions and the DNA of 50 F₂ individuals exhibiting the wild-type phenotype were pooled into a Gmpgl3-1 bulk and a Williams 82 bulk for further BSA. The Gmpgl3-1 mutant was resequenced to a depth of approximately 30× using an Illumina HiSeq 2000 device. We identified a G₋₃₄₁ to A₋₃₄₁ transition in the first exon of Glyma.02G233700 (Figure 2D), which caused a non-synonymous substitution of Gly₋₁₁₄ to Asp₋₁₁₄ in the predicted protein. No other mutations were discovered among the 18 genes in the candidate Gmpgl3-1 genomic region (Figure 2C; Supplementary Table S3). However, a single-base mutation (A–2413 to T–2413) was identified in the 14th exon of the GmTic110a gene of the Gmpgl3-2 mutant (Figure 2D). The expression level of Glyma.02G233700 decreased in the Gmpgl3-1 and Gmpgl3-2 mutants (Supplementary Figure S2), suggesting that Glyma.02G233700 is the GmTic110a gene.

BLAST searches revealed that GmTic110a, which encodes a chloroplast inner membrane protein, is highly homologous to the soybean GmTic110b protein (Glyma.14G201500; 98.4% amino acid similarity) and the Arabidopsis Tic110 protein (At1G06950; 84.3% amino acid similarity). In Medicago, the gene with the highest homology to GmTic110a is Medtr5g074690, whose sequence is 90.5% identical to that of GmTic110a and 86.4% identical to that of Medtr3g466170. Both of their proteins are 994 and 985 amino acids in length. Amino acid sequence analysis resulted in the generation of a phylogenetic tree composed of the GmTic110a homologous gene and its homologs from dicotyledonous plant species (G. max, A. thaliana, and Medicago sativa), monocotyledonous plant species (Oryza sativa, Zea mays, and Sorghum bicolor), Selaginella tamariscina, and Physcomitrella patens (Figure 3C). The results suggest that two GmTic110 homologs are evolutionarily conserved among plant species and share a common genomic structure in the observed plant species. The results of MEME analysis showed that GmTic110a contains 12 conserved motifs (Figure 3D). The mutation site of the Gmpgl3-1 mutant is located within the conserved TM2 domain, the mutation site of the Gmpgl3-2 mutant is located within the conserved chaperone-binding (co) domain, and the Gmpgl3-3 mutant site is located within the terminal TM1 domain (Figures 3A,B).

Using qRT–PCR, we examined the GmTic110a expression patterns in new leaves at the VE stage; tips, stems, and roots at the V1 stage; leaves and flowers at the R1 stage; leaves and flowers at the R2 stage; and leaves at the R3 stage. GmTic110a was slightly expressed in the tips and stems and had low expression in the roots and flowers. The highest expression levels were detected in the leaves at various stages (new leaves at the VE stage, leaves at the R1 stage, leaves at the R2 stage, and leaves at the R3 stage; Figure 3E), indicating that GmTic110a may play an important role in leaf development and regulatory processes. These results also explained why the GmTic110a mutation severely affected leaf growth at various stages.

To confirm whether Gmpgl3-1 was the GmTic110a gene, loss-of-function transgenic lines were generated by inducing mutations in the GmTic110a gene using the CRISPR/Cas9 system. The resulting CRISPR/Cas9-induced mutations in three separate mutants led to the development of a GmTic110-specific mutant phenotype (Figure 4A). The results showed that the phenotypes of GmTic110a^CR1, GmTic110a^CR2, and GmTic110a^CR3 were the same as those of Gmpgl3-1 and Gmpgl3-2. The GmTic110a^CR1 mutant contains a 7-bp substitution corresponding to the CDS of the GmTic110a gene from the 261st bp to the 273rd bp. The GmTic110a^CR2 mutant had a 2-bp deletion from the 271st to the 272nd bp, and the GmTic110a^CR3 mutant had a 2-bp deletion at 269th bp of the CDS of GmTic110a and a 1-bp substitution (Gly to Ala) at the 272nd bp (Figure 4B). These data suggested that the GmTic110a^CR1, GmTic110a^CR2 and GmTic110a^CR3 mutants have strong alleles and that complete loss of Gmpgl3 function strongly influences soybean development (Supplementary Table S1). Moreover, the expression levels of Glyma.02G233700 decreased in GmTic110a^CR1, GmTic110a^CR2, and GmTic110a^CR3 (Figure 4C). These results further suggested that Glyma.02G233700 is the GmTic110a gene.

To confirm the subcellular localization of GmTic110a, the colocalization of green fluorescent proteins (GmTic110a-GFP) and the AtPIC1-mCherry marker protein (localization to the inner envelope of chloroplasts; Duy et al., 2007) was analyzed. As shown in Supplementary Figure S3, the GmTic110a protein targeted the inner chloroplast membrane (Figure 5A). At the same time, a single transformation of an empty vector that contained GFP was used as a positive control, and the results revealed that the fluorescent signal of the empty vector was detected throughout the entire cell protoplast (Figure 5G). The results were consistent with the bioinformatics predictions of the subcellular localization of GmTic110a (Supplementary Table S5). To determine the effects of Gmpgl3 mutations on protein localization, we transiently transformed PUC19-GFP-GmTic110a^G114A, PUC19-GFP-Gmtic110a^T805S, PUC19-GFP-GmTic110a^CR1, PUC19-GFP-GmTic110a^CR2, and PUC19-GFP-GmTic110a^CR3 into Arabidopsis protoplast cells. The results showed that proteins resulting from a mutated or knocked out GmTic110a gene were located in the inner chloroplast membrane, but the GFP fluorescence signal remained diffuse (Figures 5B–F), indicating that GmTic110a mutation or knockout altered the structure of the GmTic110a protein, thereby affecting the subcellular localization of the protein.

Previous studies have shown that AtTic110 mediates transport across the inner membrane via interactions with AtTic40 and other proteins (Kovacheva et al., 2005; Chiu and Li, 2008; Yuan et al., 2021). Because the GmTic40 protein contains a highly conserved transmembrane (TM) motif, we investigated whether GmTic40 functions in conjunction with AtTic40. We used split luciferase complementation assays to confirm whether GmTic110a proteins could interact with GmTic20, GmTic40a, and GmTic40b (Figures 6A,C,E). The interactions of GmTic110a with GmTic20, GmTic40a, and GmTic40b were verified by co-IP analyses. As shown in Figure 6, we transiently coexpressed GmTic110a with GmTic20, GmTic40a, and GmTic40b in N. benthamiana leaves. Total proteins were isolated, after which they and anti-FLAG magnetic beads were incubated together to immunoprecipitate anti-FLAG. The results showed that GmTic110a, GmTic20, GmTic40a, and GmTic40b were present in the immunoprecipitate (Figures 6B–F), indicating that GmTic110a could interact with GmTic20, GmTic40a, and GmTic40b in vivo.