Local conjugation of auxin by the GH3 amido synthetases is required for normal development of roots and flowers in Arabidopsis

Gretchen Hagen 3 (GH3) amido synthetases conjugate amino acids to a carboxyl group of small molecules including hormones auxin, jasmonate, and salicylic acid. The Arabidopsis genome harbors 19 GH3 genes, whose exact roles in plant development have been difficult to define because of genetic redundancy among the GH3 genes. Here we use CRISPR/Cas9 gene editing technology to delete the Arabidopsis group II GH3 genes, which are able to conjugate indole-3-acetic acid (IAA) to amino acids. We show that plants lacking the eight group II GH3 genes (gh3 octuple mutants) accumulate free IAA and fail to produce IAA-Asp and IAA-Glu conjugates. Consequently, gh3 octuple mutants have extremely short roots, long and dense root hairs, and long hypocotyls and petioles. Our characterization of gh3 septuple mutants, which provide sensitized backgrounds, reveals that GH3.17 and GH3.9 play prominent roles in root elongation and seed production, respectively. We show that GH3 functions correlate with their expression patterns, suggesting that local deactivation of auxin also contributes to maintaining auxin homeostasis and is important for plant development. Moreover, this work provides a method for elucidating functions of individual members of a gene family, whose members have overlapping functions.


Introduction
Auxin is essential for many aspects of plant development (Zhao, 2018). Auxin concentrations in plants need to be precisely controlled to ensure optimal growth and development in response to environmental and developmental signals. Plants have evolved sophisticated mechanisms to maintain auxin homeostasis. Cellular auxin concentrations are affected by auxin biosynthesis, polar transport, degradation, and conjugation to various molecules including amino acids and sugars (Woodward and Bartel, 2005;Zhao, 2010).
Both local auxin biosynthesis and polar auxin transport have been extensively studied and they are required for various plant developmental processes. Disruption of local auxin biosynthesis leads to defects in embryogenesis, seedling growth, vascular development, and flower development (Cheng et al., 2006;Cheng et al., 2007;Stepanova et al., 2008;Tao et al., 2008). Defective polar auxin transport causes characteristic phenotypes such as the formation of pin-like inflorescences (Galweiler et al., 1998). In contrast, much less is understood regarding the roles of auxin degradation and conjugation in plant development.
The DAO gene in rice encodes a 2-oxoglutarate and Fe(II) dependent dioxygenase that is able to oxidize indole-3-acetic acid (IAA), the main natural auxin, into 2-oxindole-3-acetic acid (Zhao et al., 2013). Mutations in DAO in rice affects several reproductive processes such as anther dehiscence, pollen fertility, and seed initiation (Zhao et al., 2013). It has been shown that DAO plays a similar role in auxin oxidation in Arabidopsis, but the dao mutants in Arabidopsis only displayed subtle developmental phenotypes (Mellor et al., 2016;Porco et al., 2016;Zhang et al., 2016). Gretchen Hagen 3 (GH3) genes (Wright et al., 1987), which encode amido synthetases that catalyze the formation of an amide bond between a carboxyl group and the amino group of an amino acid (Staswick et al., 2005), were proposed to compensate for DAO functions in Arabidopsis (Zhang and Peer, 2017). Arabidopsis genome has 19 GH3 genes, which can be divided into three groups on basis of sequence similarities and gene structure (Okrent and Wildermuth, 2011). Group I is composed of GH3.10 and GH3.11 (Okrent and Wildermuth, 2011). GH3.11 (JAR1) catalyzes the conjugation of jasmonate (JA) to isoleucine to form jasmonoyl-isoleucine (JA-Ile) conjugate, which is the active molecule perceived by JA receptors (Staswick and Tiryaki, 2004). The jar1 mutants were insensitive to methyl jasmonate and were more susceptible to pathogens (Staswick et al., 2002). The functions of GH3.10 are not understood, but overexpression of GH3.10 leads to short hypocotyl under red light conditions (Takase et al., 2003). The group II GH3 genes in Arabidopsis consists of eight members (Okrent and Wildermuth, 2011). All of the group II GH3 proteins have been experimentally demonstrated to be capable of conjugating auxin to amino acids (Okrent et al., 2009;Westfall et al., 2016). Moreover, the expression of GH3.1, GH3.3,GH3.5,and GH3.6 is induced by auxin, suggesting that the group II GH3 genes have important functions in maintaining auxin homeostasis (Okrent and Wildermuth, 2011). The group III GH3 has nine members in Arabidopsis and most of the GH3 in this group have not been functionally or biochemically characterized. The expression of GH3.12 is induced by salicylic acid (SA) and GH3.12 can use benzoate, an SA analog, as a substrate (Okrent and Wildermuth, 2011). Moreover, gh3.12 mutants displayed SA-related phenotypes (Okrent et al., 2009). Recently, GH3.12 (PBS3) was discovered to conjugate isochorismate with glutamate to produce isochorismate-glutamate, which is non-enzymatically and spontaneously converted into SA (Rekhter et al., 2019). GH3.15 uses indole 3-butyric acid and glutamine as substrates in vitro, suggesting that GH3.15 may also participate in auxin homeostasis (Sherp et al., 2018).
Genetic dissection of GH3 functions has been difficult because of the genetic redundancy among the large number of GH3 genes in Arabidopsis. Gain-of-function studies of GH3 genes have clearly suggested that GH3 genes likely play important roles in plant development. For example, overexpression of GH3.2 in Arabidopsis (ydk1-D mutants) leads to short primary roots, few lateral roots, and dwarf plants (Takase et al., 2004). The dwarf in light 1 (dfl1-D) mutants, which overexpress GH3.6, have short hypocotyls and is extremely dwarf (Nakazawa et al., 2001). Overexpression of GH3.5 in wes1-D mutants leads to very short hypocotyls (Park et al., 2007b). Single loss-of-function gh3 mutants in Arabidopsis only display subtle phenotypes in hypocotyl elongation, primary root development, and lateral root initiation (Khan and Stone, 2007;Porco et al., 2016;Zheng et al., 2016). Another approach for analyzing GH3 functions was to generate knockout mutants in plants that have fewer GH3 genes than Arabidopsis. The moss Physcomitrella patens only has two GH3 genes.
Moss mutants without the two GH3 genes have elevated free IAA concentrations and a decreased level of amide-IAA conjugates (Bierfreund et al., 2004;Ludwig-Muller et al., 2009). Moreover, the double mutants were more sensitive to exogenous IAA. However, the double mutants appeared to develop normally under laboratory growth conditions (Bierfreund et al., 2004;Ludwig-Muller et al., 2009).
In order to assess the roles of GH3 genes in auxin homeostasis and plant development, we used CRISPR/Cas9 gene editing technology to generate true knockout mutants for each of the group II GH3 genes in Arabidopsis in the Columbia background. The single gh3 mutants did not display dramatic developmental defects. However, Arabidopsis plants with all group II GH3 genes inactivated (gh3 octuple) display strong auxin over accumulation phenotypes.
Light grown gh3 octuple mutants have extremely short primary roots, long lateral roots, dense and long root hairs, and much elongated petioles. Dark-grown gh3 octuple seedlings have short hypocotyls. We are able to assess the relative contributions of individual GH3 genes in auxin homeostasis and Arabidopsis development by characterizing several gh3 septuple mutants, which we believe provide sensitized backgrounds for analyzing the roles of auxin conjugation. We show that GH3.17 plays a predominant role in controlling root elongation whereas GH3.9 is the main player in controlling Arabidopsis fertility. Moreover, functions of GH3 genes appear to correlate with the expression patterns of GH3 genes, suggesting that local auxin conjugation plays key roles in maintaining auxin homeostasis and plant development.

Generation of new null mutants for GH3 amido synthetase genes
The group II GH3 family in Arabidopsis consists of 8 genes (GH3.1, GH3.2, GH3.3,GH3.4,GH3.5,GH3.6,GH3.9,and GH3.17), which can be divided into three sub-groups ( Figure 1A) based on their genomic structures, sequence homology, and the number of intron/exons. GH3 proteins are also more closely related to their sub-group members than to members in a different sub-group ( Figure 1B). In order to analyze the roles of GH3 genes in auxin metabolism and in plant development, we took advantage of the recently developed CRISPR/Cas9 gene editing technology to generate new null alleles for all of the group II GH3 genes in Arabidopsis ( Figure S1). All of the mutants were generated using two guide RNAs, which led to a deletion of a large fragment of a GH3 gene. Because the coding regions of GH3s are largely deleted, the resulting gh3 mutants are null. Another advantage of using double guide RNAs in editing GH3 genes is that the mutants can be easily genotyped using PCR-based methods (Table S1).
We analyzed the single gh3 mutants (at least two independent alleles of each gene) grown under normal laboratory growth conditions. The single mutants did not display dramatic developmental defects. The lack of strong phenotypes of the gh3 single mutants was likely caused by genetic redundancy among the GH3 genes. We next constructed multiple mutants on basis of sequence homology of the GH3 proteins. We generated gh3(1/2/3/4) quadruple mutants in which GH3.1, GH3.2, GH3.3, and GH3.4 had been deleted. We also generated gh3(5/6) double mutants, gh3 (9/17) double mutants, and gh3(5/6/9/17) quadruple mutants ( Figure 1C, Figure S2). We generated the double and quadruple mutants by crossing the single mutants together.

A complete removal of the group II GH3 genes causes dramatic developmental defects
We generated gh3 octuple mutants by crossing the gh3(5/6/9/17) quadruple mutants to gh3(1/2/3/4) quadruple mutants and genotyped the offspring of the next few generations. The gh3 octuple mutants displayed extremely short primary roots ( Figure 2A, Figure S3). The gh3 octuple mutants also had longer and more densely packed lateral roots ( Figure 2A).
Moreover, the gh3 octuple mutants had more and longer root hairs than WT plants ( Figure   2B). Interestingly, the gh3 octuple mutants had shorter hypocotyl and less-prominent apical hook when grown in the dark. ( Figure 2C & 2D). Dark-grown seedlings of the previously characterized auxin overproduction mutants such as sur1 and yuc1-D also developed shorter hypocotyls and lacked an apical hook (Boerjan et al., 1995;Zhao et al., 2001), suggesting that gh3 octuple mutants might accumulate more auxin than WT. Many of the flowers in gh3 octuple mutants aborted before they became mature ( Fig 2E). The gh3 octuple mutants were almost sterile: gh3 octuple mutants hardly produced any pollen grains ( Figure 2F). The gh3 octuple mutants had much shorter filaments and larger anthers than WT ( Figure 2F). The number of floral organs of the gh3 octuple mutants was similar to that of WT flowers.

Altered auxin homeostasis in gh3 mutants
Because the GH3 proteins conjugate free IAA to amino acids and the IAA-amino acid conjugates are considered inactive, we hypothesized that gh3 octuple mutants would accumulate more active auxin. Moreover, the observed developmental defects of gh3 octuple mutants were consistent with the hypothesis that gh3 octuple mutants accumulated more auxin ( Figure 2). It was known that auxin stimulates rooting in tissue culture and that explants of auxin overproduction mutants are able to develop extensive roots in auxin-free media (Zhao et al., 2001). Explants of gh3 octuple mutants developed roots in auxin freemedia within seven days ( Figure 3A, Figure S4). The roots from gh3 octuple explants had long and dense root hairs ( Figure 3A) whereas WT explants failed to develop roots under the same growth conditions, suggesting that gh3 octuple mutants indeed accumulated more auxin. Furthermore, gh3 octuple mutants initiated adventitious roots out of hypocotyls ( Figure 3B), a characteristic phenotype of known auxin overproduction mutants such as sur1 and sur2 (Barlier et al., 2000;Boerjan et al., 1995).
We analyzed the concentrations of free IAA and IAA-Asp and IAA-Glu, two auxin conjugates that are believed inactive, in the gh3 mutants ( Figure 3C & D). Because gh3 octuple mutants set very few seeds, we used a segregating population from plants that had a single copy of GH3.9 [gh3.9 +/-gh3 (1/2/3/4/5/6/17)], which was fertile. As shown in Figure   3C, the gh3 mutants contained almost 2 times more IAA than WT plants. In addition, both IAA-Asp and IAA-Glu in the gh3 mutants were below the detection limit of our assay ( Figure 3D), suggesting that the GH3 genes analyzed here are the main contributor to synthesizing IAA conjugates.

Determination of relative contributions of GH3 genes to Arabidopsis development
The genetic redundancy among the GH3 genes makes it difficult to dissect the unique roles of individual GH3 in auxin homeostasis and plant development. We hypothesized that the gh3 septuple mutants might provide sensitized backgrounds for analyzing the functions of individual GH3 genes. Analyzing plants with only one GH3 may reveal the functions of the GH3 in the absence of the interferences from other GH3 genes. We focus on four gh3 septuple mutants that still had one of the GH3 genes from the subgroups b and c ( Figure 1A) because GH3.5, GH3.6, GH3.9, and GH3.17 appeared to play a more prominent role based on our analysis of gh3(1/2/3/4) and gh3(5/6/9/17) quadruple mutants (Figure 1). We name a septuple-mutant ∆7gh3. The GH3.5∆7gh3 refers to plants that lacked all group II GH3 genes except the GH3.5.
Hypocotyl elongation is very sensitive to changes in auxin concentrations (Collett et al., 2000) . The gh3 octuple mutants had much longer hypocotyls than WT plants ( Figure 4A).
The GH3 genes are important in controlling petiole elongation (Figure 1). A complete deletion of the group II GH3s led to much elongated petioles ( Figure 4C). Petioles in gh3 octuple mutants were more than 2-fold longer than WT ( Figure 4C, Figure S3). Compared to gh3 octuple mutants, the petiole length in GH3.5∆7gh3 and GH3.6∆7gh3 did not change much. In contrast, petioles in GH3.9∆7gh3 and GH3.17∆7gh3 were significantly shorter than those in gh3 octuple mutants (Fig 4C, Figure S5), suggesting that the two GH3 genes play a more important role in petiole development than GH3.5 and GH3.6.

The gh3 mutant phenotypes largely correlate with the expression patterns the GH3s
We analyzed the expression patterns of GH3.5, GH3.6, GH3.9, and GH3.17 using the publically available gene expression profile data (Arabidopsis.org) ( Figure S6). The GH3 genes were hardly expressed in dry seeds, but soaking the seeds in media for one day induced the expression of GH3.5 and GH3.6, but not the other two GH3 genes ( Figure S6). By the third day of socking, GH3.5, 3.6, and 3.17 were highly expressed. Among the four GH3 genes analyzed, GH3.9 is the only GH3 that is not highly expressed in roots during seedling stages. The GH3.9 expression patterns were consistent with our observations that GH3.9∆7gh3 and gh3 octuple mutants had similar root defects ( Figure 4B). The GH3 genes that were expressed in roots were able to partially restore root elongation defects of gh3 octuple mutants ( Figure 4A & 4B). Among the four GH3 genes, GH3.6 was highly expressed in hypocotyls ( Figure S6). Interestingly, GH3.6 had a more pronounced impact on hypocotyl elongation ( Figure 4A). GH3.5, GH3.6, and GH3.9 were expressed in flowers and they were able to partially overcome the fertility defects observed in gh3 octuple mutants ( Figure 5B).
In siliques, the highly expressed GH3 gene was GH3.9 ( Figure S6). Consistent with the expression pattern, GH3.9 alone was sufficient to maintain Arabidopsis fertility in the absence of all other group II GH3 genes ( Figure 5)

Discussion
The roles of GH3 amido synthetases in auxin homeostasis and plant development have been difficult to define because of the genetic redundancy among the GH3 genes. By generating new null alleles of group II GH3 genes and by constructing plants without any group II GH3 genes, we are able to demonstrate that GH3 genes are required for normal development of Arabidopsis. Plants without the group II GH3 accumulate more free-IAA and fail to make any IAA-Asp and IAA-Glu, two presumed inactive auxin conjugates.
Consequently, gh3 octuple mutants have extremely short roots, more and dense root hairs, and long petioles. Moreover, we identify the prominent GH3 genes responsible for root development and flower development by analyzing gh3 septuple mutants. This work not only defines the physiological roles of the group II GH3 genes, but also provides a method for elucidating the unique functions of members of a gene family, which are often masked by other family members due to genetic redundancy.
Auxin homeostasis is severely disrupted when the group II GH3 genes are deleted from Arabidopsis plants. The gh3 octuple mutants displayed phenotypes that have been observed in plants treated with exogenous auxin and/or auxin overproduction mutants ( Figure 2). All auxin overproduction mutants and transgenic lines including yuc1-D (Zhao et al., 2001), iaaM overexpression lines (Romano et al., 1995), CYP79B2 overexpression lines (Zhao et al., 2002), sur1 (Boerjan et al., 1995), and sur2 (Barlier et al., 2000) have longer hypocotyls than WT plants when grown in light. In darkness, sur1 and yuc1-D have shorter hypocotyls and lacked apical hook (Boerjan et al., 1995;Zhao et al., 2001). Light-grown gh3 octuple mutants had much elongated hypocotyls, indicative of high auxin phenotypes (Figure 4). Too much auxin caused by either auxin treatments or auxin overproduction inhibits primary root elongation and stimulates root hair and lateral root development. The gh3 octuple mutants had very short primary roots and dense and long root hairs, a phenotype that is likely caused by auxin over accumulation (Figure 2). Our previous studies on yuc1-D demonstrated that auxin over accumulation greatly decreases Arabidopsis fertility (Zhao et al., 2001). We observed that the gh3 octuple mutants displayed defects in fertility similar to that of yuc1-D ( Figure 5). All of the developmental defects observed in gh3 octuple are consistent with the hypothesis that gh3 octuple mutants accumulate more auxin than WT plants. Our analysis of free IAA and IAA-conjugates further confirmed that gh3 mutants fail to conjugate IAA to Asp and Glu. Consequently, the free IAA concentrations in gh3 mutants were almost doubled. Our results are consistent with findings from a previous characterization of the sextuple mutants gh3 (1,2,3,4,5,6), which had elevated free IAA, no detectable IAA-Asp, and elevated IAA-Glu (Porco et al., 2016). The difference in IAA-Glu concentrations between our gh3 mutants and the gh3 (1,2,3,4,5,6) suggests that GH3.9 and Gh3.17 may prefer to using Glu as a substrate, or that GH3.9 and GH3.17 are more activated in the sextuple mutants or both. Nevertheless, our analysis clearly demonstrated that GH3s play a key role in auxin homeostasis and Arabidopsis development. Some of the group II GH3 have previously shown to use SA and JA as substrates (Park et al., 2007a;Westfall et al., 2016;Zhang et al., 2007). Whether some of the phenotypes such as sterility observed in gh3 octuple is partially caused by alteration of the homeostasis of other hormones remain to be investigated. A promoter-swap experiment would help determine whether the different developmental defects are caused by the different expression patterns of GH3 genes or by differences in biochemical properties or both. The genetic materials generated in this work should be very useful for studying hormonal cross-talks.
It is well documented that plant development is very sensitive to changes in auxin concentrations (Collett et al., 2000). Auxin can inhibit or stimulate plant growth depending on the concentrations. Hypocotyl elongation and primary root elongation are two main phenotypic readouts of auxin concentration changes. It is known that auxin overproduction stimulates hypocotyl elongation (Zhao et al., 2001). The gh3 octuple mutants had longer hypocotyls than WT (Figure 4). Interestingly, both GH3.5∆7gh3 and GH3.6∆7gh3 septuple mutants had longer hypocotyls than the gh3 octuple mutants ( Figure 4A). Although we did not analyze the actual IAA concentrations in GH3.5∆7gh3 and GH3.6∆7gh3 septuple mutants, we believe that the two septuple mutants likely have accumulated less auxin than gh3 octuple mutants. Our results suggest that IAA concentrations in gh3 octuple may actually be too high for optimal hypocotyl growth. We also observed that the primary root length of gh3(1/2/3/4) and GH3.17∆7gh3 septuple mutants was longer than WT (Fig1 &4), suggesting that a slight increase of auxin in Arabidopsis stimulates primary root elongation, though further increases of auxin concentrations inhibit root elongation as shown in gh3 octuple mutants. Our results are consistent with previous findings that auxin-dependent plant growth displays a bell-shaped curve.
It is relatively straightforward to demonstrate the overlapping functions among members of a gene family by determining whether the higher order mutants of gene family members have enhanced phenotypes than the lower order mutant combinations. In contrast, determining the unique functions of individual members of a gene family is more complicated because the interference of other family members. We hypothesized that a removal of all gene family members except one may provide sensitized backgrounds for determining functions of the particular member. In contrast to analyzing multiple gene knockouts for phenotypic enhancements, here we determine whether a GH3 gene is sufficient to suppress a developmental defect displayed in the gh3 octuple mutants. We showed that GH3.17 alone was sufficient for restoring root elongation defects observed in the gh3 octuple mutants ( Figure 4A&B). But GH3.17 had no impact on restoring the fertility of gh3 octuple mutants ( Figure 5). Interestingly, GH3.9 did not affect root elongation as GH3.9∆7gh3 septuple mutants and gh3 octuple were almost identical in terms of root elongation ( Figure 4B). However, GH3.9 alone was sufficient to reverse the sterile phenotypes observed in gh3 octuple mutants. The unique functions of GH3.17 in root elongation and GH3.9 in fertility only became evident in the septuple mutants (Figure4 & 5), whereas both gh3.17 and gh3.9 single mutants only had subtle phenotypes. Interestingly, the phenotypes of GH3.17∆7gh3 and GH3.9∆7gh3 septuple mutants correlate with the expression patterns of the two GH3 genes ( Figure S6). It is well known that both local auxin biosynthesis and polar auxin transport are required for generating and maintaining auxin gradients. Local auxin conjugation by the GH3.17/VAS2 was previously shown to play an important role in shade avoidance (Zheng et al., 2016). Our results suggest that local auxin conjugation by GH3.17 is important for root elongation and that auxin conjugation by GH3.9 is critical for Arabidopsis fertility. This work also demonstrated the effectiveness of our using sensitized backgrounds for defining unique functions of individual genes of a gene family, whose members have overlapping functions.

Plant materials and Growth Conditions
Arabidopsis thaliana plants used in this study were in the Columbia-0 genetic background.
The gh3 mutants were generated using CRISPR/Cas9 gene editing technology, which was described previously in detail (Gao et al., 2016;Gao et al., 2015;Gao and Zhao, 2014). We used two guide RNA molecules for each GH3 gene to generate deletions of large fragments.
The gRNA target sequences and the locations of the gRNAs are shown in Figure S1. The sizes and the locations of the deletion in each gh3 mutants are shown in Figure S1 as well.
The sequences flanking the deletions are shown in the Figure S1. For example, gh3-c1 had a 1940 bp deletion and gh3.1-c2 harbored a 1943 bp deletion ( Figure S1).
We used a PCR based method to genotype the gh3 mutants (Gao et al., 2016). A pair of primers that are located outside of the two gRNA target sequences can amplify out a large fragment from WT genomic DNA and a much smaller fragment from a deletion mutant ( Figure S1). For example, the GH3.1-GT1 and GH3.1-GT2 primer pair generates a 2.8 kb fragment and a 0.8 kb fragment from WT and gh3.1 mutant DNA, respectively. To further clarify the zygosity of the gh3 mutants, we designed a third primer that was located between the two gRNA target sequences ( Figure S1). The GT3 primer paired with GT1 was used to amplify a fragment from WT DNA, but not from the mutant DNA. The locations and directions of all of the genotyping primers were shown in Figure S1. The primer sequences were listed in Table S1. To generate higher order of gh3 mutants, we used the gh3-c1 alleles of each GH3 gene.