Short-interval traffic lines: versatile tools for genetic analysis in Arabidopsis thaliana

Abstract Traffic lines are transgenic stocks of Arabidopsis thaliana that contain a pair of linked seed-specific eGFP and DsRed markers. These stocks were originally developed for the purpose of studying recombination, but can also be used to follow the inheritance of unmarked chromosomes placed in trans to the marked chromosome. They are particularly useful for this latter purpose if the distance between markers is short, making double recombination within this interval relatively rare. We generated 163 traffic lines that cover the Arabidopsis genome in overlapping intervals of approximately 1.2 Mb (6.9 cM). These stocks make it possible to predict the genotype of a plant based on its seed fluorescence (or lack thereof) and facilitate many experiments in genetic analysis that are difficult, tedious, or expensive to perform using current techniques. Here, we show how these lines enable a phenotypic analysis of alleles with weak or variable phenotypes, genetic mapping of novel mutations, introducing transgenes into a lethal or sterile genetic background, and separating closely linked mutations.


Introduction
Identifying a mutation or polymorphism of interest in the progeny of a genetic cross is central to genetic analysis. If the molecular identity or chromosomal location of the mutation/polymorphism is known, this can be done using molecular methods. However, these methods are destructive and time-consuming and require resources that may not be available in some circumstances. We developed a set of transgenic lines of Arabidopsis thaliana that make it possible to visually determine the genotype of seeds in a segregating population prior to planting (Wu et al. 2015). Each transgenic line-termed a "Traffic Line" (TL) because of its green, red, and yellow phenotypes-possesses a pair of linked pNAP::eGFP and pNAP::DsRed transgenes. The reporters in these transgenes are under the regulation of the seed-specific Napin promoter and thus produce, respectively, green or red fluorescent seeds (Stuitje et al. 2003). Such lines were originally developed for the purpose of studying recombination (Melamed-Bessudo et al. 2005), but also make it possible to follow the transmission of the chromosome segment that is in trans to marked segment (Wu et al. 2015). Indeed, a TL can be used much like a balancer chromosome (Casso et al. 2000;Hentges and Justice 2004) to follow the segregation of genes located in a defined region of the genome.
Recombination between the transgenes in a TL is useful if the goal is to study factors that influence recombination (Melamed-Bessudo and Levy 2012; Ziolkowski et al. 2015Ziolkowski et al. , 2017Saini et al. 2020;Nageswaran et al. 2021;Kbiri et al. 2022). However, recombination is disadvantageous if the TL is being used to follow the segregation of an unmarked chromosome segment in trans to the TL because double recombination within the interval generates an unmarked chromosome that may be genetically different from the original unmarked chromosome. Recombination is minimized if the interval between the markers in a TL is relatively short. Such "short-interval" TLs are not only useful as balancers but also facilitate experiments that rely on the ability to detect recombination in small, defined chromosomal regions. These include studies of the factors that influence recombination frequency, removing second-site mutations/polymorphisms linked to a gene of interest, and fine mapping a QTL or a newly induced mutation known only by its phenotype.
Our initial set of TLs contained transgenes located an average of 4.8 Mb (21 cM) from each other (Wu et al. 2015). Here, we describe a collection of 163 TLs that covers the Arabidopsis genome in much shorter overlapping segments. In addition, we present examples of how TLs facilitate genetic analysis by providing a rapid, visual method for determining the genotype of plants in a segregating population.

Plant material
The recombinant lines described here are in the Columbia (Col) accession and were produced from transgenic lines containing

Construction of FIE-HA and its transformation into fie-11
A 4.1-kb FERTILIZATION-INDEPENDENT ENDOSPERM (FIE) genomic region without the stop codon was fused to 6xHA and cloned into the P1 position of the pAGM4723 vector in the Golden Gate system; Basta resistance was cloned into the P2 position in this vector (Engler et al. 2014). The primers used in cloning are listed in Supplementary Table 2. TL 3.31 was crossed to fie-11 heterozygotes (genotyped using the fie-11 F and fie-11 R1 primers) and the fie-11 heterozygotes from this cross were selected for further analysis. Moderately fluorescent seeds from the fie-11 heterozygote's progeny were planted and the FIE-HA construct was transformed into these plants. Fully developed nonflorescent seeds from the T0 population were planted and single T-DNA insertion lines were selected from the T2 generation by Basta selection. Plants that had 2 copies of fie -11 and FIE-HA were identified by quantitative PCR using a Bio-Rad CFX 96 real-time System.

Results
TLs were generated from a collection of sequenced pNAP::eGFP and pNAP::DsRed T-DNA insertions, using the approach described in Wu et al. (2015) and summarized the Materials and Methods. The chromosomal position of these TLs is illustrated schematically in Fig. 1, along with the locations of the TLs described previously. The insertion sites of the transgenes used in the construction of the new lines, and the physical and recombination distance between them are presented in Table 1 and Supplementary Table 1. Although these recombination distances are sometimes based on relatively few individuals, they are of interest because they represent the frequency of recombination across the entire genome within an inbred background. On average, the physical distance between transgenes was 1.2 Mb, and the recombination distance was 6.9 cM. This yields an average recombination rate of 5.6 cM/ Mb (Table 1), which is higher than has been observed in crosses between different accessions (Singer et al. 2006;Salomé et al. 2012). A few lines (e.g. TL1.55, TL2.20) displayed an unusually high rate of recombination whereas others displayed an unusually low rate of recombination compared to other lines in the same region of the genome. Although TLs with unusually low rates of recombination could reflect the actual rate of recombination in this region, this phenomenon could also be due to small chromosomal rearrangements associated with one or both of the transgenes in the TL. Transgenes associated with reciprocal translocations or large inversions were eliminated by screening T1 plants for semisterility, but this approach cannot detect small rearrangements.
Using a TL to genotype seeds It is often important to be able to distinguish different genotypes in a segregating population. This is particularly important if one wants to compare the phenotype of mutants with weak or poorly penetrant phenotypes with their wild-type siblings. In Arabidopsis, this is typically done by performing allele-specific PCR on plants after germination. To demonstrate that TLs can be used to perform this task, we took advantage of a semidominant allele of CHLORINA 42 (ch42-4sd) that has a yellow-green phenotype when heterozygous and is albino when homozygous. Plants heterozygous for ch42-4sd were crossed to TL4.22, a line that spans the CH42 locus and contains transgenes located 1.8 Mb from each other. Yellow-green progeny from this cross were allowed to selffertilize, and nonfluorescent, moderately fluorescent red and green, and strongly fluorescent red and green seed were selected (Fig. 2). One hundred percent of the nonfluorescent seeds produced albino seedlings, 97% of the moderately fluorescent seeds produced yellow-green seeds, and 97% of the strongly fluorescent seeds produced dark-green seedlings. Thus, TLs are a reliable tool for determining the genotype of a seed prior to planting.

Using TLs for phenotypic analysis
Determining the phenotype of a mutation in a known gene can be problematic if the mutation has a weak phenotype or is present in a genetic background different from the one that is commonly used in a laboratory. The latter problem can occur if the mutation was generated in a different accession (e.g. Ler instead of Col), or is in a mutagenized background, which may contain other linked and unlinked mutations. One way to get around this problem is to compare the phenotype of mutant and wild-type siblings in the same family. On average, these individuals will share the same background, even if this background is not completely uniform.
To test the usefulness of TLs for comparing the phenotype of genetically different siblings, we examined the phenotype of myb75-1 in a moderately heterogeneous genetic background. This mutation was of particular interest because it has been reported to delay the transition from juvenile to adult growth (vegetative phase change) under short days, but not long days (Meng et al. 2021). myb75-1 is a Ds-induced allele originally isolated in the Nö ssen accession (Teng et al. 2005). We crossed myb75-1 to Col 5 times and then crossed these myb75-1 plants to TL1.55, a TL that spans a 1-Mb interval containing the MYB75 locus. Nonfluorescent and moderately fluorescent seeds produced by TL1.55/myb75-1 plants were grown separately in the same flat under long-day conditions. As shown in Fig. 3, nonfluorescent seeds had slightly, but significantly, later abaxial trichomes than their fluorescent siblings. This result demonstrates that myb75-1 delays vegetative phase change under both long-day and shortday conditions.

Using TLs to map a novel mutation
In addition to being useful for tracking the segregation of known genes, TLs can also be used to map novel mutations or natural variants. As an example, we used a set of long-interval TLs (Wu et al. 2015) to map a novel juvenilized mutant present in a transgenic stock of Col obtained from the Arabidopsis Biological Resource Center. This mutation was unlinked to the known T-DNA insertion in this line, and a subsequent "mapping by bulked genome sequencing" approach failed to identify a polymorphism that segregated with this mutation. As an alternative approach, we crossed the mutant to a set of long-interval TLs spanning the Arabidopsis genome, and determined if the mutant phenotype was linked to markers in these lines by planting 12-15 nonfluorescent and a similar number of strongly fluorescent seeds from the F2 progeny of these crosses. In most cases, wild-type and mutant seedlings were present in both the fluorescent and nonfluorescent seed classes in approximately equal ratios, with wild-type seedlings typically being more abundant than mutant seedlings (the expected ratio is 3/ 4 wild type:1/4 mutant for an unlinked TL; Table 2). In contrast, all of the fluorescent seeds from the cross to TL4.6 had a wild-type phenotype, and almost all of the nonfluorescent seeds had a mutant phenotype. This result suggests that the mutation is located within, or close to, the segment marked by TL4.6.
Using TLs to rescue a lethal or sterile mutation with a transgene It is often desirable to generate a transgenic line carrying an epitope-tagged protein (e.g. a reporter protein, or an affinitytagged protein). Ideally, the epitope-tagged construct should be transformed into a line carrying a null mutation of the corresponding gene to demonstrate that the construct is able to rescue the mutant phenotype and ensure that the only functional copy of the protein is the version produced by the transgene. This is straightforward when the homozygous mutant is viable and fully fertile, and can therefore be directly transformed with the construct. It is more difficult when plants homozygous for the mutation are lethal or sterile. In this case, the simplest approach is to transform the construct into wild-type plants and then introduce the construct into a homozygous mutant background by several generations of crossing. This is a time-and space-consuming process and requires a considerable amount of molecular genotyping.
TLs facilitate this process by providing a convenient method for identifying plants homozygous and heterozygous for lethal/ sterile mutations. As an example, we used a TL to generate a line carrying an HA-tagged version of FIE, an essential component of the PRC2 complex (Ohad et al. 1999). Plants homozygous for the null allele, fie-11, are embryonic lethal, making it necessary to maintain this mutation in heterozygous condition (Guitton et al. 2004). We generated a balanced stock by crossing fie-11/þ plants to TL3.31 and identified the fie-11/TL3.31 plants from this cross using PCR primers specific for fie-11. We then selected moderately fluorescent green and red seed from these plants (which were expected to be fie-11/TL3.31), and used floral dipping to transform these plants with a T-DNA construct containing pFIE::FIE-HA and Basta-resistance as a selectable marker (Fig. 4a). Approximately 75% of the T1 seeds resulting from this experiment were fluorescent and were expected to either be heterozygous for fie-11 or homozygous wild type (Fig. 4). Most of the remaining seeds were shriveled and dead, as expected because fie-11 is lethal in homozygous condition. We also identified 14 nonfluorescent, viable seeds (Fig. 4a). These T1 seeds were planted, and their progeny were screened for Basta-resistance. Three of these families segregated 1 =4 dead seed and were sensitive to Basta indicating that the parental T1 seeds were    heterozygous for fie-11 and not transgenic. These seeds were likely the result of double recombination within the marked segment which transferred the wild-type allele of FIE to an unmarked chromosome. The remaining 11 families segregated 1 =4 or less dead seeds, and all the viable seeds were resistant to Basta. Six of the families had very few dead seed, suggesting that they contained multiple, independent, pFIE::FIE-HA insertions. Five families had 1 =4 dead seed, implying that the parental T1 seed was homozygous for fie-11 and heterozygous for a single pFIE::FIE-HA insertion. Seedlings from at least 6 plants in each of these single-insertion families were treated with Basta to identify families homozygous for the transgene. Subsequent quantitative PCR analysis of the dosage of fie-11 and FIE-HA (primer sequences in Supplementary Table 2) demonstrated that these lines were indeed fie-11/fie-11 pFIE::FIE-HA/pFIE::HA. An alternative approach is useful for lethal or sterile mutations whose seeds are phenotypically normal (Fig. 4b). In this case, it is impossible to identify which nonfluorescent seeds contain the transgene of interest because transgenic seeds are visually indistinguishable from nontrangenic seeds. Transgenic plants can be identified after germination using a selectable marker (e.g. Basta resistance) present on the transgene, but if the transformation efficiency is relatively low this will require selecting and planting a large number of nonfluorescent seeds, which is tedious. Alternatively, one can screen a large number of T1 plants for the presence of the selectable marker and then screen the T2 seeds produced by these plants for seed fluorescence. Fig. 2. Self-progeny of a ch42-4sd/TL4.22 plant, planted according to the seed fluorescence phenotype. Nonfluorescent seeds are expected to be predominantly ch42-4sd/ch42-4sd and albino, moderately fluorescent seeds are expected to be predominantly ch42-4sd/TL4.22 and yellowgreen, and strongly fluorescent seeds are expected to be predominantly TL4.22/TL4.22 and dark green. Seedlings that do not conform to the expected phenotype are circled.
Approximately 1/4 of the plants containing the selectable marker should have nonfluorescent seeds and be homozygous for the lethal/sterile mutation.

Using TLs to separate tightly linked mutations
Next-generation sequencing has revealed that many induced mutations are associated with linked second-site mutations (Schneeberger et al. 2009;Enders et al. 2015;Thole and Strader 2015;Lup et al. 2021). Determining whether the phenotype of these lines is attributable to one or more of these linked mutations can be problematic if the mutations are so tightly linked that they do not readily segregate from each other. We identified an EMSinduced late vegetative phase change mutant that contained missense mutations in 2 members of the PRC1 complex, VAL1 and AtBMI1A (Molitor and Shen 2013;Merini and Calonje 2015). Null alleles of each gene have a much weaker effect on vegetative phase change than this double mutant, so we assumed that the relatively strong phenotype of the double mutant was attributable to the combined effects of these missense mutations. One way to test this hypothesis is to separate these mutations by recombination and examine their individual phenotypes. However, VAL1 and AtBMI1A are located only 45.5 kb from each other on chromosome 2, making this an extremely time-and resource-intensive experiment to perform by molecular genotyping. We selected 12 strong red, moderate green, and 12 strong green, moderate red seeds from the F2 progeny of plants heterozygous for TL2.31 and val1 atbmi1a (Fig. 5). Because these genes are located ca. 600 Mb from both the pNAP::eGFP and pNAP::DsRed transgenes, we expected approximately equal numbers of val1 atbmi1a/þþ and þþ/þþ recombinants and a much smaller number of val1 þ/þþ and þ atbmi1a/þþ recombinants. Consistent with this expectation, we obtained only 2 val1 þ/þþ recombinants and no þ atbmi1a/þþ recombinants. Unexpectedly, we found that the number of recombinants between pNAP::eGFP and val1 was 3 times greater than the number of recombinants between pNAP::DsRed and val1 (18 vs 6). This was surprising because val1 atbmi1a are equidistant from these transgenes. This suggests that the pNAP::DsRed insertion in TL2.31 (CR1124) is associated with a small rearrangement that suppresses recombination. This hypothesis is supported by the observation that the cM/Mb ratio in TL2.31 is significantly less than the ratio in a different line (TL2.32) that spans the same region (Table 1).
Subsequent analysis of the phenotype of the val1 single mutant (val1-5sd) revealed that it is semidominant (Fouracre et al. 2021), and that its phenotype is identical to the phenotype of the original val1-5sd atbmi1a double mutant. This result suggests that-in contrast to our initial hypothesis-the missense mutation in AtBMI1A makes no contribution to the phenotype of the double mutant and may be phenotypically silent. It also suggests that val1-5sd interferes with the function of VAL2 or other members of the PRC1 complex, as null alleles of VAL1 have a weaker phenotype than val1-5sd (Fouracre et al. 2021).

Discussion
The value of linked fluorescent transgenes for studies of recombination has been recognized for some time (Melamed-Bessudo et al. 2005;Berchowitz and Copenhaver 2008). The utility of such transgenes for following the segregation of chromosome segments in trans to the marked chromosome is less well appreciated, but is an equally important feature of these lines. This property relies on the fact that a nonfluorescent chromosome segregating from a doubly marked chromosome will either represent a nonrecombinant chromosome, or a chromosome that has undergone double recombination within the marked interval (Wu et al. 2015). The frequency of double recombination decreases with the distance between markers, so it is beneficial to use TLs that define relatively small intervals if the goal is to use the TL as a balancer.
For an interval of 6.9 cM-the average for the TLs described here-the maximum frequency of double recombination is 0.035 Â 0.035, or approximately 1/1000 chromosomes, although this number is likely to be significantly lower due to crossover interference (Berchowitz and Copenhaver 2010;Salomé et al. 2012). This means that the nonfluorescent chromosomes produced by a plant heterozygous for a short-interval TL are very likely to have the genotype of the non-TL chromosome. Furthermore, for any single nucleotide within this interval (e.g. a mutation or polymorphism), the probability that a nonfluorescent chromosome will carry the parental form of this nucleotide increases the farther the nucleotide is from the midpoint of the marked interval. This is because the probability that double recombination will replace the allele on the nonfluorescent chromosome with the allele   Fig. 3. The first leaf with abaxial trichomes in the self-progeny of a myb75-1/TL1.55 plant, planted according to seed fluorescence phenotype. Nonfluorescent seeds are expected to be predominantly myb75-1/myb75-1 and moderately fluorescent seeds are expected to be predominantly myb75-1/TL1.55. Error bars ¼ SEM; ** ¼ P < 0.01, 2-tailed Student's t-test, n > 45.
from the TL decreases as one moves away from the center of the interval; i.e. the closer a gene is to one of the transgenes in a TL, the less susceptible it is to double recombination. Thus, even relatively large TLs can be useful for gene mapping, as we found in the case of a novel phase change mutation that we mapped to the bottom of chromosome 4. Single fluorescent transgenes are also useful for following polymorphisms/mutations in trans to the transgene if the polymorphisms/mutations is sufficiently close to the transgene (Wu et al. 2015).
Although recombination within a TL is disadvantageous if the TL is being used to follow the segregation of a chromosome in trans to the TL, this feature makes it possible to dissect relatively small chromosomal regions. As described here, we were able to recombinationally separate mutations located only 45 kb (ca. 0.03 cM) from each other by visual screening seeds for fluorescence intensity. This process only took a few hours. In contrast, screening the same F2 population for recombinants by PCR would have required hundreds or possibly thousands of quantitative PCR reactions, at considerable time and expense. The ability to rapidly identify and then rescue recombinants in a small chromosomal region is not only useful for dissociating second-site mutations but also facilitates fine mapping of mutations or natural variants that have been localized to specific regions of the genome. It is particularly useful for mapping variants with weak phenotypes because the dosage sensitivity of the fluorescent markers makes it possible to identify plants homozygous for recombinant chromosomes in F2 populations. This facilitates determining which allele of the variant the chromosome possesses because the phenotype of this allele is likely to be strongest when it is homozygous.
An ideal TL is one in which the transgenes are not subject to silencing, are expressed strongly and uniformly in dry seeds, display clear differences in fluorescence intensity in 1 and 2 doses, and are inserted in functionally silent positions in the genome. Unfortunately, this is difficult to achieve. It is easy to produce T-DNA-transformed lines in Arabidopsis, but a large number of these lines contain more than one insertion, and many contain a Fully formed (i.e. viable) seeds will either be homozygous for fie-11 and hemizygous for pFIE-HA BaR, or be heterozygous for fie-11 as a result of double recombination within the interval marked by TL3.31. Given the relatively low frequency of transformation, this latter class of seeds will probably lack pFIE-HA BaR. b) Using a TL to rescue a seedling lethal or sterile mutant with a transgene. This approach is useful if the mutation does not affect seed morphology or viability, making it impossible to distinguish transgenic from nonransgenic seeds prior to planting.  Fig. 5. Using a TL to separate closely linked mutations. F2 seeds from a val1 bmi1a/TL2.31 plant were screened for seeds that displayed strong red and moderate green fluorescence, or strong green and moderate red fluorescence. The genotype of these seeds was determined by examining the molecular genotype of their F2 progeny, using allele-specific PCR primers.
translocation or other type of large rearrangement at the site of the T-DNA insertion (Clark and Krysan 2010;Wu et al. 2015). Consequently, only some of the primary transgenic lines are useful for the production of TLs. The existence of small inversions that do not lead to semisterility but which eliminate or reduce recombination is another problem. These are useful if the TL is being used as a balancer, but are problematic if the TL is being used to identify recombination in a defined interval, as was the case in our experiment with the val1 atbmi1a double mutant. Finally, there is the problem of transgene silencing. Some of the transgenes we identified silenced after several rounds of propagation, and the frequency of this phenomenon increased for transgenes located close to the centromere. Indeed, it was sometimes necessary to abandon a TL late in the process of generating or amplifying the line because one or both of the markers underwent silencing. Unfortunately, this behavior is typical of transgenes in plants, and only repeated propagation will demonstrate if the transgenes in a TL are sufficiently stable for prolonged use. An expanded collection of pNAP::eGFP and pNAP::DsRed insertions will enable us to minimize these problems in future collections of TLs. Despite these limitations, the collection of TLs we have generated enable many experiments that are difficult or sometimes impossible to perform, and therefore represent a useful addition to the resources available for genetic analysis in A. thaliana.

Data availability
Short-interval TLs are available as homozygous stocks from the Arabidopsis Biological Resource Center (ABRC), under the stock numbers provided in Table 1. The ABRC stock numbers for the long-interval lines described previously are provided in Wu et al. (2015). The transgenes used to construct these lines are available from the ABRC, under the stock numbers provided in Supplementary Table 1. Other genetic stocks described in this article are also available from the ABRC or from the corresponding author. Supplemental material is available at G3 online.