Seeds were germinated as described previously (Schnabel et al., 2010). Briefly, seeds were scarified using concentrated sulfuric acid, imbibed in water, vernalized at 4°C for 2 days, and allowed to germinate overnight at room temperature in the dark. One-day-old seedlings were placed in an aeroponic chamber and grown at 21°C–25°C on a 14-hour/10-hour light/dark cycle and inoculated as described in Cai et al. (2023). To determine nodulation phenotypes, no N was included in the media, and plants were inoculated with Sinorhizobium meliloti RM41 (Putnoky et al., 1990) 4 days post-germination as indicated. Nodule count and root length measurements were performed at 10 dpi. For seed collection and genetic crosses, plants were grown in a greenhouse with supplemental light on a 14-hour/10-hour light/dark cycle at 21°C–25°C. For mycorrhizal assays, seed germination and inoculation with Rhizophagus irregularis were performed as previously described (Chaulagain et al., 2023). In brief, 3-day-old seedlings were planted in Cone-tainers SC10R (Stuewe and Sons, Tangent, OR, USA) and placed in a growth chamber on a 16-hour/8-hour light/dark cycle at 22°C–24°C and 40% humidity. The growth mixture was 50% play sand (Quikrete, Atlanta, GA, USA) and 50% fine vermiculite (Ferry-Morse, Norton, MA, USA). All substrate components were washed in distilled water and autoclaved for 55 min before use. For colonization with R. irregularis, 250 spores (Premier Tech, Rivière-du-Loup, QC, Canada) were placed 5 cm below the surface in a layer of fine sand. All plants were treated twice a week with 15 mL of half-strength, low Pi Hoagland’s fertilizer (20 µM phosphate) to aid in symbiosis initiation. Plants were harvested 4.5 or 6 weeks post-planting.

Tnt1 insertion lines created by Tnt1 mutagenesis of the R108 ecotype (Tadege et al., 2008) were screened electronically for insertions in MtTML genes. While no MtTML1 insertion lines were identified, a pool of plants contained a mutation in the MtTML2 gene in M. truncatula (NF0679; Noble Foundation Medicago Mutant Database, now located at https://medicagomutant.dasnr.okstate.edu/mutant/index.php). Seeds from the NF0679 pool were grown in the greenhouse, and a line containing the insertion in the second exon was identified by PCR using primers from Supplementary Table 2 . The line was selfed to yield homozygotes, which were then backcrossed to the R108 wild type and re-isolated. The amplicons generated by primers 2189/1925 and 2690/3227 were sequenced to determine the exact position of the Tnt1 insertion in relation to the gene. This line is called tml2-1.

Multi-target constructs were designed for MtTML1 alone and to target both MtTML genes in the same plant ( Supplementary Figure 1 ). Two target sites per gene were chosen following the target selection criteria (Ma and Liu, 2016). Sites unique to the MtTML genes were selected to avoid off-target mutagenesis by comparison to the M. truncatula genome MtV4.0 (Tang et al., 2014). The binary vector pDIRECT_23C (Čermák et al., 2017) (Addgene) was used as a Cas9-containing vector. The target and gRNA were cloned into the pDIRECT_23C vector, and the finished construct was created using Golden Gate assembly (Engler et al., 2008, 2009) following the protocol in Čermák et al. (2017). The resulting vectors pDIRECT_23C+MtTML1 (containing two targets for MtTML1) and pDIRECT_23C+MtTML1/2 (containing two targets each for MtTML1 and MtTML2) were then verified by sequencing. A verified clone in Escherichia coli (Zymo 10B, Zymo Research, Irvine, CA, USA) was used as a plasmid source to move each construct into Agrobacterium tumefaciens EHA105 (Hood et al., 1993) for use in plant transformation.

To obtain alleles of MtTML1, additional alleles of MtTML2, and plants containing mutations in both genes, CRISPR–Cas9 mutagenesis of embryonic callus tissue followed by regeneration of whole plants was performed in the R108 ecotype. Each construct was introduced with Agrobacterium-mediated transformation into root segments of 4–7-day-old R108 seedlings and taken through callus and tissue culture following the protocol 2A of Wen et al. (2019) with the following modifications for selection and rooting: the selection of phosphinothricin (ppt)-resistant callus used 2 mg/L ppt in the media, and 2 mg/L NAA was used in RCTM6 medium described in Wen et al. (2019) for rooting. All plants originating from a callus were considered one independent line. The three most developed plants at the rooting medium stage for each line per callus were transferred into the soil for acclimatization and grown to collect seeds.

All T0 lines generated from tissue culture were screened by PCR for the presence of the transgene using Cas9-specific primers (primers 3081/3082; Supplementary Table 2 ), and lines testing positive were then screened by PCR for deletions using gene-specific primers ( Supplementary Table 2 ). The deletion was confirmed by the size difference of an amplicon after gel electrophoresis. All the amplicons were sequenced using the same primers as for the PCR to determine the exact mutations.

All T0 lines containing mutations were allowed to set seeds. The T1 generation was screened for plants homozygous for a single allele by PCR. All amplicons were sequenced to confirm homozygosity and sequence. To generate the tml1 tml2 double mutant transgenic lines, a T0 line carrying two mutant alleles for each gene was used. All alleles except tml1-4 were backcrossed once to R018, and all alleles including the homozygous double mutants did not have detectable Cas9 when tested by PCR amplification.

RNA was extracted from 10 roots of each of the mutants and R108 [wild type (WT)] was harvested at 10 days post-inoculation with RM41. Root tissue was frozen at −80°C until RNA extraction. Root tissue from six plants pooled together for each genotype was ground in liquid nitrogen and 100 mg of tissue was used for RNA extraction. RNA was extracted using TRIzol^® Reagent (Life Technologies, Carlsbad, CA, USA) followed by DNase digestion using RQ1 RNase-free DNase (Promega, Madison, WI, USA) according to the manufacturer’s instructions. cDNA was synthesized using iScript™ Reverse Transcriptase (Bio-Rad, Hercules, CA, USA) according to the manufacturer’s instructions. Real-time qPCR was performed in 10-µL reactions in an iQ5 instrument (Bio-Rad, CA, USA) using iTaq™ Universal SYBR^® Green Supermix (Bio-Rad, CA, USA) and a 400-nM final concentration of each primer ( Supplementary Table 2 ) Each reaction was run in three technical and three biological replicates. Relative expression levels of genes were assayed using the Pfaffl method (Pfaffl 2001) relative to a previously validated housekeeping reference gene phosphatidylinositol 3- and 4-kinase belonging to the ubiquitin family (PI4K; Medtr3g091400 in MtV4.0, MtrunA17Chr3g0126781 in MtrunA17r5.0-ANR) (Kakar et al., 2008). Data from the three biological replicates were used to estimate the mean and standard error.

DNA was isolated from leaf presses made by pressing a leaflet of each plant to a Plant Card (Whatman™, GE Healthcare UK Limited, Amersham, UK) for long-term storage. A 1.2-mm-diameter piece of Plant Card was excised and washed with Whatman™ FTA Purification Reagent (GE Healthcare UK Limited, UK) followed by TE-1 buffer according to the manufacturer’s instructions and directly used in PCR. DNA from the R108 wild type used multiple times was isolated using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) as per the manufacturer’s instruction.

All amplifications except for use in cloning were performed using GoTaq^® G2 DNA Polymerase (Promega Corporation, Madison, USA) following the manufacturer’s instructions and buffers. Standard PCR conditions used were as follows: 2 min at 95°C, 35 cycles (10 s at 95°C, 15 s at gene-specific annealing temperature, 30 s/Kb at 72°C), 5 min at 95°C. When using a leaf press as a DNA template for amplification, cycles were increased from 35 to 40. The gene-specific annealing temperature was 55°C except as noted for primer pairs 3148/3149 (58°C), 3182/3183 (52°C), and 2493/2494 (56°C). Annealing temperatures were predicted using the “Tm Calculator” tool available at https://www.thermofisher.com/. Amplification of inserts for cloning was performed using iProof™ High-Fidelity DNA Polymerase (Bio-Rad Laboratories, USA) and buffers according to the manufacturer’s instructions.

Nodule images were taken using the Leica Microsystems THUNDER Imager Model Organism with the attached DMC4500 color camera. For visualization of AM fungal structures, colonized roots were cleared in 20% KOH (w/v) at 65°C for 2 days. Then, they were rinsed three times with ddH₂O for at least 1 day until the roots appeared white, and the pH was neutralized overnight using 1× phosphate-buffered saline (PBS). Cleared roots were stained in Alexa Fluor^® 488 wheat germ agglutinin (WGA)–PBS solution, so the final concentration of WGA was 0.2 μg/mL. AM fungal root colonization was quantified using the grid-line intersections method using a stereomicroscope and a gridded square petri dish (Sparling and Tinker, 1978; Giovannetti and Mosse, 1980). Images were captured using a Leica M205 stereoscope.

The phylogenetic in tree was generated with (MEGA11) (Tamura et al., 2021) using the maximum likelihood method and tested using bootstrap for 1,000 replicates. The normality of data was tested by fitting a normal curve before performing the statistical tests. Tukey’s comparison for all pairs was used for comparison among more than two groups, and Student’s t-test was used for comparing two groups for normal data. The non-parametric Steel–Dwass comparison for all pairs was used for the analysis of non-normal data. Statistical test was performed using JMP Pro 17.1.

Since only one TML gene has been identified in L. japonicus (Magori et al., 2009), which forms determinate nodules, we wondered if having two copies of TML, like M. truncatula (Gautrat et al., 2019), was correlated with indeterminate nodulation and whether multiple copies appear in non-nodulating plants. Figure 1 displays a maximum likelihood tree for amino acid sequence homologs of LjTML identified by BLASTP in selected determinate and indeterminate nodule-forming legumes, non-legume dicots, and monocots. TML proteins from legumes clustered in a separate branch from non-legumes, while single TMLs from the monocots rice, maize, and sorghum clustered separately forming the tree root; these species have only ~40% amino acid sequence similarity with LjTML. In cases where non-legumes had two TML proteins, with similar low similarity scores to both TML1 and TML2, the labels are based on previous descriptions or ploidy. Arabidopsis does not form either symbiotic association with rhizobia or mycorrhizae, and the Arabidopsis protein AtHOLT (HOMOLOG OF LEGUME TML) has only 51% amino acid sequence identity to LjTML, yet the AtHOLT transcript is a target of miR2111 (Hsieh et al., 2009; Pant et al., 2009) and regulates lateral root formation in response to nitrate in the same manner as LjTML. There is no correlation among legumes between TML gene number and nodule meristem type. Only one gene encoding a TML protein was identified by BLAST search in the determinate nodulating species L. japonicus and Phaseolus vulgaris (common bean), whereas determinate nodulating soybean has genes encoding three TML proteins. GmTML1a and GmTML1b are gene duplicates sharing 93.32% sequence identity, and the encoded proteins cluster together in the phylogenetic tree, likely the result of the partially diploidized tetraploid soybean genome (Shultz et al., 2006). The same a, b designation used for the soybean duplicate proteins was applied to Nicotiana tabacum, a recent allotetraploid of Nicotiana sylvestris and Nicotiana tomentosiformis (Renny-Byfield et al., 2011) and Populus trichocarpa, which underwent a whole genome duplication approximately 60 million years ago (Dai et al., 2014). The TML2 protein sequences in legumes (GmTML2, MtTML2, PsTML2, and TpTML2) cluster together and branch separately from TML1 sequences, suggesting that the two genes encoding TML proteins evolved from a common ancestral duplication in legumes, compared to the independent duplications in other dicots that form clades by species.

To explore the function of the two TML proteins in M. truncatula, mutant lines were created and isolated. The tml2-1 allele was identified in the NF0679 pool from the Tnt1 insertional mutagenesis library and isolated as a homozygous line (see Materials and Methods). Additional mutant alleles of MtTML1 and MtTML2, as well as plants containing mutations in both genes, were generated by CRISPR–Cas9 mutagenesis of embryonic callus tissue followed by regeneration of whole plants in the R108 ecotype (see Materials and Methods). Multiple combinations of heterozygous and biallelic mutations were identified, and two individual alleles of each gene, as well as two double mutant combinations, were used in this study ( Figure 2 ; Supplementary Figure 1 ). The tml1 mutants created by CRISPR include a deletion in tml1-1 resulting in a premature stop codon, which was recovered in the creation of a double mutant and only characterized in the tml1-1 tml2-2 double mutant. The two other alleles of tml1 were isolated and characterized as single mutants: tml1-2, which removes 111 aa from the N terminus of the protein, and tml1-4, a deletion eliminating the start codon. The tml2 single mutants include a CRISPR-generated deletion and the above-described Tnt1 insertion, both resulting in stops early in the protein. As a result, the tml1-1 tml2-2 double mutant is the combination of two early stops, and the tml1-2 tml2-2 double mutant is the combination of an early deletion and a stop. All stops occur prior to or early in the F-box coding region, so that any resulting truncated protein is not expected to contain the F-box, NLS, and kelch repeats ( Figure 2 ). The binding sites for miR2111 in MtTML1 and MtTML2 start at 143 bp and 134 bp into the CDS, respectively, and the binding site is lost only in the tml1-4 mutant ( Supplementary Figure 1 ).

The nodule number phenotypes of plants carrying mutant tml1, tml2, and tml1 tml2 double mutant alleles were determined at 10 dpi with S. meliloti RM41 in an aeroponic system and compared to wild-type R108 plants and hypernodulating sunn-5 plants (described in Crook et al., 2016; Nowak et al., 2019). Plants carrying a mutation in a single MtTML gene had more nodules than wild-type plants, while plants with mutations in both MtTML genes had many more nodules and shorter roots than wild-type plants ( Figure 3A ). All single mutant lines formed pink nodules similar in size to wild-type plants ( Figures 3B–F ), whereas the double mutant plants formed small, fused nodules in a wider area of the root ( Figures 3G, H ). Most nodules in the double mutants were white at 10 dpi, except for a few slightly pink nodules, indicating leghemoglobin production. No obvious shoot phenotype was observed in any line.

All mutants in a single MtTML gene formed a significantly higher number of nodules than the wild type but less than the known hypernodulating sunn-5 mutant (p ≤ 0.01, Steel–Dwass method of pairwise comparison; Figure 4A ). The evaluation of nodules per cm of root length followed the same pattern as the total number of nodules ( Figure 4B ). However, the double mutants formed a significantly higher number of total nodules and nodules per cm of root length than both the wild type and sunn-5 mutant ( Figures 4C, D ). The single mutants on average formed twofold to threefold more nodules, while the double mutants formed greater than 20-fold more nodules compared to the wild type ( Figure 4 ).

Using the ePlant resource for early nodulation (Schnabel et al., 2023) and past transcriptomics work in our lab (Schnabel et al., 2023), we examined the expression of the two MtTML genes in root segments during the first 72 hours responding to rhizobia ( Figure 5 ). The genes were expressed in wild-type plants at low levels at the 0-hour timepoint, which were 4-day-old N-starved plants in our experimental system, but the expression increased in the vascular tissue for both genes at 12 hpi and 24 hpi ( Figure 5A ). By 48 hpi, when the nodules began to form, MtTML1 was expressed at a similar level in the vascular, cortical, and epidermal laser-captured cells as well as the developing nodule cells, while MtTML2 increased to a higher level in the vascular and inner cortical cells at the xylem poles and did not increase in the nodule. By the time the nodule organized and began to emerge at 72 hpi, this difference in pattern was reinforced, with MtTML1 highly expressed in the nodule and MtTML2 expressed in the nodule at the same low level as in tissues of uninoculated plants. Because of the way the laser capture data were collected and displayed, color indicates expression somewhere within the cell types, rather than throughout the area colorized. Examined at the level of single-cell RNA-Seq over a time course of combined experiments harvested at timepoints 0 hpi, 48 hpi, and 96 hpi, the combined transcriptomics data of individual cells responding to rhizobia showed that both MtTML genes were expressed in the pericycle cells and vascular bundles, but MtTML1 was expressed in different cell lineages than MtTML2 in response to rhizobia (Pereira et al., 2024). Observation of expression levels in undissected root segments over time revealed that both MtTML genes were induced twofold to threefold in wild type responding to rhizobia over 72 hours, as well as in the hypernodulation mutant sunn-4 responding to rhizobia over 72 hours (Schnabel et al., 2023). However, MtTML1 expression did not begin to decrease over this time frame in the sunn-4 mutant, while MtTML2 did. Taken together, the data imply that MtTML1 and MtTML2 expression is dynamic and changes between tissues and over time in nodulation; MtTML1 has higher expression in cortical cells and nodules, whereas MtTML2 is expressed mainly in the pericycle and vasculature.

Genetic compensation can occur through several mechanisms, but transcriptional adaptation, in which degradation of the mutant mRNA transcript triggers the increased expression of a paralog or family member (El-Brolosy et al., 2019), is straightforward to test. Using real-time quantitative PCR, we measured the expression of one wild-type MtTML gene when the other MtTML was mutated. In the roots of wild-type R108 and plants containing either a tml2 or tml1 mutant transcript, there was no statistical difference in the expression of the wild-type MtTML message in the absence of rhizobia (0 hpi), and all plants showed statistical increases in the expression of the examined transcript from the uninoculated level (0 hpi) compared to 72 h after inoculation with rhizobia (72 hpi) (p < 0.05) ( Figures 5B, C ). In the tml2 mutant, the expression level of MtTML1 at 72 hpi was statistically equivalent to that of the wild type at 72 hpi, showing no evidence of genetic compensation ( Figure 5B ). However, in the tml1 mutant, the expression level of MtTML2 is increased compared with that of the wild type at 72 hpi (p < 0.05) ( Figure 5C ), suggesting that the loss of a functional copy of MtTML1 increases the expression of MtTML2, but not vice versa. The change in the level of MtTML1 expression at 72 hpi versus 0 hpi in wild-type plants (2.5-fold) is half that of MtTML2 at 72 hpi versus 0 hpi in wild-type plants (fivefold), which could make statistically significant differences in MtTML1 expression harder to detect using this method.

Following the observation that the expression of the wild-type MtTML2 gene increased in a plant containing a mutation in MtTML1 but did not fully rescue the nodule number phenotype of the tml1 mutant, we wondered if there is an effect on the nodulation due to heterozygosity of mutant allele. While an intermediate nodule number for plants heterozygous for a tml mutation may be difficult to detect given the small size of the effect on nodule number of a single mutation (twofold to threefold change increase in single mutants; Figure 4A ), the magnitude of the phenotype of double mutants could allow a nodule number difference to be detectable when plants are homozygous for one mutant allele and heterozygous for the other.

We examined two populations of progeny obtained from selfing plants homozygous for one tml mutation and heterozygous for the other, tml1-2/TML1 tml2-2 and tml1-2 tml2-2/TML2. A discrete recessive phenotype should segregate nodule number phenotypes in a 3:1 Mendelian ratio. While nodule number is a qualitative trait, both populations segregated a broader distribution of nodule number phenotypes with some bi-modal characteristics ( Supplementary Figure 2 ). To test our hypothesis that this distribution of phenotypes is due to changes in the inheritance of the segregating wild-type copy, we used the progeny of plants homozygous for tml2-2 and segregating tml1-2. Out of 141 progenies, a random set of 84 (~60%) of the population was tested by PCR. The PCR-tested progeny were verified for the expected 1:2:1 genotypic segregation using Pearson’s chi-square test (χ² = 7.13, with 2 degrees of freedom) ( Figure 6A ). The analysis revealed that plants wild type for the segregating allele had a mean of 22.5 nodules, plants heterozygous for the segregating allele had a mean of 43.8 nodules, and plants homozygous for both mutant alleles had a mean of 100.3 nodules. Each of the three groups was significantly different from each other in Tukey’s all-pair comparison (p < 0.001) ( Figures 6B, C ), supporting a phenotypic effect of heterozygosity on the nodule number phenotype.

Many AON mutants also have hyper-mycorrhizal phenotypes, and miRNA2111 is induced upon phosphate starvation even in Arabidopsis (Hsieh et al., 2009), which does not form symbiotic associations with rhizobia or AM fungi. Expecting an increase in AM fungal root colonization, the tml1-1, tml2-2, and tml1-2, tml2-2 double mutants were tested for a mycorrhizal phenotype with the AM fungus R. irregularis. No overall root length colonization difference between double mutants and wild types was observed under our growth conditions at 4.5 weeks post-inoculation ( Figure 7A ). However, in the same experiment at 6 weeks post-inoculation, a small but statistically significant increase in AM fungal root length colonization was observed for the tml1-2 tml2-2 double mutant relative to R108 wild-type controls, but not the tml1-1 2-2 allele ( Figure 7A ). Arbuscule morphology appeared normal in the mutants ( Figures 7B–D ). Together, these data suggest that MtTML1 and MtTML2 may play a minor role in AM symbiosis, although their impact appears to depend on the timepoint measured and the mutant allele.