The Maize Hairy Sheath Frayed1 (Hsf1) Mutation Alters Leaf Patterning through Increased Cytokinin Signaling[OPEN]

Increased cytokinin signaling in the maize Hairy Sheath Frayed1 mutant modifies leaf development, leading to changes in patterning, growth, and cell identity. Leaf morphogenesis requires growth polarized along three axes—proximal-distal (P-D) axis, medial-lateral axis, and abaxial-adaxial axis. Grass leaves display a prominent P-D polarity consisting of a proximal sheath separated from the distal blade by the auricle and ligule. Although proper specification of the four segments is essential for normal morphology, our knowledge is incomplete regarding the mechanisms that influence P-D specification in monocots such as maize (Zea mays). Here, we report the identification of the gene underlying the semidominant, leaf patterning maize mutant Hairy Sheath Frayed1 (Hsf1). Hsf1 plants produce leaves with outgrowths consisting of proximal segments—sheath, auricle, and ligule—emanating from the distal blade margin. Analysis of three independent Hsf1 alleles revealed gain-of-function missense mutations in the ligand binding domain of the maize cytokinin (CK) receptor Z. mays Histidine Kinase1 (ZmHK1) gene. Biochemical analysis and structural modeling suggest the mutated residues near the CK binding pocket affect CK binding affinity. Treatment of the wild-type seedlings with exogenous CK phenocopied the Hsf1 leaf phenotypes. Results from expression and epistatic analyses indicated the Hsf1 mutant receptor appears to be hypersignaling. Our results demonstrate that hypersignaling of CK in incipient leaf primordia can reprogram developmental patterns in maize.


INTRODUCTION
Proper leaf morphogenesis in vascular plants requires defined patterns of growth polarized along three axes: adaxial-abaxial axis, medial-lateral axis, and proximal-distal (P-D) axis (McConnell and Barton, 1998;Tsukaya, 1998;Bowman et al., 2002;Byrne et al., 2002). Growth along the P-D axis is particularly evident in grass leaves, such as maize (Zea mays), that are composed of four distinct segments: the sheath is proximal, the blade is distal, and the auricle and ligule form the boundary between the two ( Figure 1A; Sylvester et al., 1996). A number of genes have been identified that influence P-D patterning, with BLADE-ON-PETI-OLE (BOP) genes affecting proximal identity in eudicots and monocots (Ha et al., 2003(Ha et al., , 2004Norberg et al., 2005;Tavakol et al., 2015;Toriba et al., 2019). In grasses, ectopic expression of class I knotted1-like homeobox (knox) transcription factor genes in developing leaf primordia alters P-D patterning, primarily disrupting the formation of a defined sheath-blade boundary (Freeling and Hake, 1985;Hake et al., 1989Hake et al., , 1991Smith et al., 1992;Schneeberger et al., 1995;Muehlbauer et al., 1997;Foster et al., 1999a;Tsiantis et al., 1999;Byrne et al., 2001). Class I knox genes typically function in meristem formation and maintenance, and their downregulation is required for normal development of determinant organs, such as leaves with a simple morphology (Kerstetter et al., 1994;Endrizzi et al., 1996;Long et al., 1996).
In meristems, KNOX proteins function to increase cytokinin (CK) accumulation by positive regulation of CK synthesis genes and simultaneously decrease gibberellic acid (GA) accumulation by suppression of GA biosynthesis genes or activation of GA catabolic genes (Ori et al., 2000;Sakamoto et al., 2001Sakamoto et al., , 2006Hay et al., 2002;Jasinski et al., 2005;Yanai et al., 2005;Bolduc and Hake, 2009). Recently, a rice (Oryza sativa) KNOX transcription factor was shown to also affect brassinosteroid accumulation by upregulating brassinosteroid catabolism in the shoot apical meristem (Tsuda et al., 2014). Determinate leaf primordia form when knox expression is downregulated by the action of ROUGH SHEATH2 (RS2) and related proteins, resulting in a decrease in CK and an increase in GA accumulation (Hay et al., 2006;Guo et al., 2008). In addition to the action of CK and GA, auxin is required for proper leaf initiation and positioning. The polar transport of auxin by PINFORMED1 (PIN1) auxin efflux carriers in the Brassicaceae and SISTER OF PIN1 (SoPIN1) in maize, tomato (Solanum lycopersicum), and other species guides the formation of auxin maxima, localized regions of high auxin accumulation, that are required for initiation of leaf primordia (Reinhardt et al., 2000;Pozzi et al., 2001;Scarpella et al., 2006;Benjamins and Scheres, 2008;Zhao, 2008;Guenot et al., 2012;O'Connor et al., 2014O'Connor et al., , 2017Martinez et al., 2016). The emerging model predicts that spatial differences in CK:auxin ratios control final cell fate (Shani et al., 2006;Müller and Sheen, 2008). Ectopic knox expression presumably shifts critical phytohormone ratios in developing leaf primordia, but the exact molecular mechanisms by which phytohormone ratios determine leaf patterning remain incomplete.
As phytohormones play pivotal roles in many developmental programs, the pathways that signal their perception and response have been well characterized. For example, the perception and response to the CK phytohormones rely on a two-component signal transduction system (Hwang and Sheen, 2001;Yonekura-Sakakibara et al., 2004;Hwang and Sakakibara, 2006;Du et al., 2007;To and Kieber, 2008). The perception of CK is mediated via a partially redundant signaling system of histidine kinases (HKs), histidine phosphotransfer proteins (HPTs), and response regulators (RRs). CK signaling begins with the perception of CK by binding to HK receptors at the endoplasmic reticulum, and probably also plasma membrane, which triggers receptor phosphorylation (Lomin et al., 2011). The activated receptors initiate phosphorelay by transferring phosphoryl groups to HPTs, which shuttle between the cytoplasm and nucleus. Once in the nucleus, phosphorylated HPTs transfer their phosphoryl groups to type-B RRs that, in turn, activate expression of type-A RRs and other CKresponsive genes (Rashotte et al., 2006). The type-A RRs and other CK-responsive genes mediate several CK-regulated processes including shoot and root growth, de-etiolation, leaf expansion, root vascular development, senescence, and CK homeostasis (To and Kieber, 2008). In maize, multiple members of each of the CK signaling components have been identified (Yonekura-Sakakibara et al., 2004). Maize has seven HKs (ZmHKs) of which three have been shown to bind and signal various types of CKs in heterologous assays (Lomin et al., 2011;Steklov et al., 2013). Three HPTs, three type-B RRs, and seven type-A RRs (ZmRRs) have also been identified in maize (Asakura et al., 2003). Of these signal transduction components, the function of only ZmRR3, a type-A RR, has been defined by null mutations and shown to underlie the aberrant phyllotaxy1 (abph1) mutation (Jackson and Hake, 1999;Giulini et al., 2004). Our understanding of the functions of other components of the CK signal transduction pathway remains largely incomplete for cereal species such as maize.
To gain a better understanding of the signaling mechanisms that mediate leaf pattern specification, we initiated a study of the semidominant Hairy Sheath Frayed1 (Hsf1) mutation that alters P-D leaf development in maize (Bertrand-Garcia and Freeling, 1991a). Although Hsf1 disrupts the P-D leaf pattern, similar to dominant class I knox mutations, Hsf1 is not itself a knox gene, (A) Adaxial view of midrib-to-left-margin half-leaves from the wild type (WT) and Hsf1-1603/1 sibs showing the P-D organization of the sheath (s), ligule (l), auricle (a), and blade (b) and a prong outgrowth (red arrowhead). Bar 5 5 cm. (B) Close-up of a blade margin (b) from the wild type (WT) and Hsf1-1603/1 plants showing a prong consisting of the proximal leaf segments sheath (s), ligule (l), and auricle (a) juxtaposed to the blade (b). Bar 5 1 cm. (C) Comparison of leaf phenotypes between the three Hsf1 alleles. L4 (top), fourth leaf below tassel; L5 (bottom), fifth leaf below tassel. Bar 5 10 cm. (D) Amino acid alignment of a portion of the CHASE domain from different plant HK CK receptors and the three Hsf1 mutant alleles. Missense residues are marked by black triangles for the Hsf1 alleles and by a white triangle for the Lotus snf2 allele. Amino acid sequences derived from AT2G01830 (AHK4), AM287033 (LHK1 and LHK1-snf2), XM_003570636 (BdHK1), XM_002454271 (SbHK1), BAD01584 (ZmHK2), NP_001104867 (ZmHK3a), NM_001111389 since it does not map to the location of any maize knox genes (Bertrand-Garcia and Freeling, 1991b). In this report, we show that the Hsf1 phenotype results from specific missense mutations in the maize CK receptor Z. mays Histidine Kinase1 (ZmHK1) gene (Yonekura-Sakakibara et al., 2004). Our analyses of mutant receptor function, the effects of exogenous CK treatment on leaf development, and epistatic interaction suggest that the ZmHK1 receptor is hypersignaling in Hsf1 mutants. Overall, our results indicate CK hypersignaling can influence the specification of P-D leaf patterning in maize and underscore the capacity of CK to alter developmental programs.

RESULTS
The Hsf1 Mutation Induces Specific Alterations to Maize Leaf Patterning The original Hsf1 mutation arose via ethyl methanesulfonate (EMS) mutagenesis of the inbred line Mo17 and was designated Hsf1-N1595 (also called Hsf1-O; Bird and Neuffer, 1985). A second mutation, Hsf1-N1603 (hereafter called Hsf1-1603), was shown to be allelic or very closely linked (Bertrand-Garcia and Freeling, 1991a). We isolated three additional alleles in independent EMS mutagenesis screens in different inbred backgrounds: Hsf1-AEWL in A619, Hsf1-2559 in Mo17, and Hsf1-7322 in A632. All Hsf1 alleles have very similar phenotypes compared to the Hsf1-N1595 (hereafter called Hsf1-1595) reference mutation. As was shown previously for Hsf1-1595, plants heterozygous for any of the Hsf1 alleles display a highly penetrant mutant leaf patterning phenotype with outgrowths consisting of an organized auricle/ ligule and sheath emanating from the distal blade margin (Figures 1A to 1C;Bertrand-Garcia and Freeling, 1991a). The outgrowths have proximal identity, since they resemble the leaf pattern at the sheath-blade boundary, and were called prongs, the term we adopted to describe this structure ( Figure 1B). Although Hsf1 mutant plants have proximal tissue growing on the distal blade, they have a normal sheath-blade boundary ( Figure 1A; Bertrand-Garcia and Freeling, 1991a). All of the pleiotropic phenotypes described for Hsf1-1595 in Bertrand- Garcia and Freeling (1991a) are shared by all of the other Hsf1 alleles, including an increase in macrohair size and density on the abaxial sheath, adaxial blade, and blade margin; and an increase in leaf number, a shorter stature, short and narrow leaves, and reduced root growth (Supplemental Figures 1A and 1B;Supplemental Table 1). Bertrand-Garcia and Freeling (1991a) also showed homozygous Hsf1 plants have a stronger mutant phenotype, being extremely stunted, with multiple shoots arising from the coleoptile node at germination, and having adventitious needle-or club-shaped leaves (Supplemental Figures 1A and 1B).
Since plants heterozygous for either Hsf1-1595, Hsf1-1603, or Hsf1-AEWL were phenotypically very similar ( Figure 1C), we chose the Hsf1-1603 allele to characterize the temporal and spatial patterns of prong formation to better understand how the Hsf1 mutation affected leaf patterning. In Hsf1-1603 heterozygotes, prongs first appeared on leaf 5 in a few plants and most commonly appeared on leaf 6 or leaf 7, but not on earlier arising leaves (Supplemental Figure 1C). The earliest sign of P-D leaf polarity specification is the formation of the preligule band (PLB; Sylvester et al., 1990;Johnston et al., 2014) that is first observed in plastochron 5 (P5)-stage or P6-stage leaf primordia, with the initiating ligule becoming visible in P7 or P8 primordia (Johnston et al., 2014). Plastochron describes the stage of leaf primordia development and refers to the position of the primordia relative to the shoot apical meristem (Sylvester et al., 1990). To determine whether the Hsf1-1603 mutation influenced the timing of the acquisition of P-D polarity, we examined leaf primordia in Hsf1-1603/1 and the wild-type sibling (sib) plants for signs of early ligule development (see Methods). The initiating ligule was most commonly first visible on P7 primordia in both the wild type and Hsf1-1603 heterozygotes, indicating no influence on P-D polarity acquisition (Supplemental Figure 1E). To determine whether the appearance of prong primordia on the blade margin coincided with the acquisition of P-D polarity, developing leaf primordia from Hsf1-1603 heterozygotes were dissected and examined for the presence of initiating prongs. Prong initials were most commonly observed on the blade margins of P5 or P6 leaf primordia, but some were noted as early as stage P4 (Supplemental Figures 1D,  1F, and 1G), consistent with prong formation observed in Hsf1-1595 heterozygotes (Bertrand-Garcia and Freeling, 1991b). Thus, prongs typically initiated from blade margins about the same plastochron stage as formation of the PLB.
Prongs were observed to occur in different sizes and at different positions along the leaf blade margin ( Figure 1C; Supplemental Figure 2A). To determine whether prong formation was random or patterned, we measured the size and position of each prong from both margins of mature leaves collected from different positions on the shoot of Hsf1-1595, Hsf1-1603, and Hsf1-AEWL heterozygous plants. Prong formation was more frequent on leaves higher on the shoot (Supplemental Table 2), with prongs occupying more of the blade margin in the topmost leaves compared to leaves lower on the shoot (Supplemental Figures 2B and 2C). Next,   , in the loop domain (red) forming one face of the binding cavity, is predicted to form a salt bridge with Glu-236, the residue altered in Hsf1-1603. Hsf1-1595 is Pro-190 to Leu, Hsf1-1603 is Glu-236 to Lys, and Hsf1-AEWL is Leu-238 to Phe. we determined where prongs formed along the P-D axis of the blade. Analysis indicated prongs only formed in the proximal 70%, and not in the distal 30%, of the blade, with the majority of prongs forming within a region encompassing the proximal 15 to 40% of the blade (Supplemental Figure 2D). Next, we examined the range of prong sizes for each Hsf1 allele within this prong-forming region. For all three alleles, the majority of prongs were about a centimeter in size, but a few were larger, ranging from 3 to 6 cm (Supplemental Figure 2E). With the relative position and size known, we next asked whether prong position was related to its size. In general, the largest prongs often formed in the basal 20% of the blade and smaller prongs formed at any position within the prong-forming region (Supplemental Figure 2F). Thus, our analysis indicated prong formation was not random but occurred in particular regions of the blade and initiated at specific developmental stages.

Gain-of-Function Mutations in the Maize CK Receptor Gene ZmHK1 Underlie the Hsf1 Mutation
Previous studies mapped Hsf1-1595 to the long arm of chromosome 5 (Bertrand-Garcia and Freeling, 1991b). To isolate the gene underlying the Hsf1 locus, we screened a backcross mapping population of more than 3000 plants with linked molecular markers derived from the maize reference genome (B73 Re-fGen_v1). The Hsf1 locus was localized to a 21-kb interval with a single gene model (GRMZM2G151223, B73 RefGen_v2). This gene model was well supported with abundant EST evidence and was annotated as encoding ZmHK1, one of seven maize HK CK receptors (Yonekura-Sakakibara et al., 2004;Steklov et al., 2013). To confirm ZmHK1 was the correct gene and to identify the causative lesions, ZmHK1 was sequenced from all five Hsf1 alleles. The entire ZmHK1 genomic region, including ;2 kb upstream and downstream of the transcription start and stop, was sequenced from Hsf1-1595, Hsf1-1603, Hsf1-2559, Hsf1-7322, and Hsf1-AEWL homozygotes and their progenitor inbred lines. As expected for EMS-generated mutations, single nucleotide transitions were identified in the five Hsf1 alleles compared to their progenitor sequences. Although each allele arose independently, Hsf1-1595 and Hsf1-1603 had the exact same transition mutations as Hsf1-7322 and Hsf1-2559, respectively. Thus, hereafter, we refer to the three different Hsf1 alleles as follows: Hsf1-1595, Hsf1-1603, and Hsf1-AEWL. Each transition mutation produced a missense mutation in a highly conserved amino acid located in the cyclases/histidine-kinase-associated sensory (CHASE) domain of the ZmHK1 protein, where CK binding occurs ( Figure 1D; Hothorn et al., 2011;Steklov et al., 2013). The Hsf1-1595 mutation changed Pro-190 to Leu (CCA>CTA), the Hsf1-1603 mutation changed Glu-236 to Lys (GAG>AAG), and the Hsf1-AEWL mutation changed Leu-238 to Phe (CTT>TTT). The missense mutation in Hsf1-AEWL is particularly significant because this is the same type of amino acid substitution, although at a slightly different position in the CHASE domain, that was found in another gain-of-function mutation in a CK receptor, the spontaneous nodule formation2 (snf2) mutation in the lotus Lhk1 receptor ( Figure 1D; Tirichine et al., 2007). The snf2 mutation was shown to cause mutant LHK1 to signal independently of the CK ligand in a heterologous signaling assay, suggesting the snf2 mutation locked LHK1 in an active signaling state (Tirichine et al., 2007).
Based on the location and nature of the amino acid substitutions in the three Hsf1 mutations and the presumed mode of action of the snf2 mutation in Lhk1, we hypothesized that the Hsf1 mutations might also lock the ZmHK1 receptor in an active CK signaling state and signal independently of CK.

Hsf1 Mutant CK Receptors Have Altered HK Signaling and Ligand Binding Activities
To determine whether the Hsf1 mutant receptors signal independently of CK, we used a heterologous HK signaling assay system developed in the yeast Saccharomyces cerevisiae (Suzuki et al., 2001). In the yeast assay, the cognate HK of an endogenous two-component phosphorelay signal transduction system was deleted. Functional replacement of the endogenous HK with the assayed CK receptor, in this case ZmHK1, allowed the activity of the receptor to be determined as the output of the endogenous yeast transduction system, which is the ability to grow on Glc media (Suzuki et al., 2001). We engineered the exact point mutation found in each Hsf1 mutation into the ZmHK1 cDNA in the p415CYC-ZmHK1 plasmid for expression in yeast (Suzuki et al., 2002;Higuchi et al., 2009). We next tested receptor activity in the sln1 deletion yeast strain TM182 carrying each of the Hsf1 missense mutations, the wild-type ZmHK1 cDNA, and the empty p415CYC vector grown on Glc media with and without the CK ligand ( Figure 1E). As expected, the wild-type ZmHK1 strain only grew well on Glc media supplemented with higher concentrations of the three CKs tested ( Figure 1E), and at lower CK concentrations this strain only grew robustly on Glc with the preferred ligand N 6 -(D 2 -isopentenyl)adenine (iP; Supplemental Figures 3A to 3C). In the absence of added CK, strains carrying either ZmHK1-AEWL or ZmHK1-1603 grew robustly on Glc media ( Figure 1E). This result indicated that the ZmHK1-AEWL and ZmHK1-1603 receptors signaled independently of added CK in this assay. To determine whether the mutant receptors were still CK responsive, they were also grown on Glc media supplemented with the three tested CKs ( Figure 1E). Growth on Glc supplemented with different CKs did not reveal any receptor activity differences between ZmHK1-AEWL and ZmHK1-1603. Surprisingly, growth of the ZmHK1-1595 strain was different from the other two mutant receptors and the wild type. The ZmHK1-1595 strain did not grow on Glc media without added CK, similar to the wild-type ZmHK1 ( Figure 1E). Instead, the ZmHK1-1595 strain showed strong growth on Glc media with 10 mM of the preferred CK iP and weaker growth on Glc with 10 mM of two other bioactive CKs, trans-zeatin (tZ) and ciszeatin (cZ), suggesting ZmHK1-1595 had weak receptor activity in this assay ( Figure 1E).
To investigate ligand specificity differences, CK ligand binding affinities were determined for the mutant and the wild-type receptors. Affinities were determined for six different CKs and adenine (Ade) using two binding assays with receptors expressed in bacterial spheroplasts or residing in Nicotiana benthamiana membranes after transient expression in planta (Romanov et al., 2005;Lomin et al., 2011Lomin et al., , 2015. The ZmHK1-1595 receptor was not stable when expressed in Escherichia coli (Supplemental Figure 3D), necessitating an analysis in tobacco microsomes, which was shown to produce comparable results to the bacterial assay (Supplemental Figure 4; Lomin et al., 2015). The ligand preferences for the wild-type ZmHK1 receptor were comparable with those determined previously (Table 1; Lomin et al., 2011Lomin et al., , 2015. The mutant receptors, on the other hand, all showed increased affinities for most of the CKs tested (Table 1). The preference ranking of the mutant receptors for different CKs was mostly similar to the wild type (Supplemental Figure 5), but the affinities were increased between two-and eightfold ( Table 2). The only exception was the affinity for the synthetic CK thidiazuron, which was reduced for all the mutant receptors compared to the wild-type ZmHK1. Thus, the missense mutations in the Hsf1 alleles increased the relative binding affinity of the receptor for all the natural CKs tested, suggesting the mutant receptors might be hypersignaling.

Hsf1 Missense Mutations Localize near the CK Binding Pocket in ZmHK1
To gain insight into how each Hsf1 missense mutation might impact CK binding, we determined the effect these mutations had on the structure of the CHASE domain, which was facilitated by the publication of the crystal structure of the ARABIDOPSIS HISTI-DINE KINASE4 (AHK4) gene CHASE domain (Hothorn et al., 2011). AHK4 is likely co-orthologous to ZmHK1 (69% identical and 83% similar within 245 residues of the CHASE domain) and three other likely paralogous HKs in the maize genome (Steklov et al., 2013). To explore the effects of the Hsf1 mutations on receptor structure, homology modeling was used to model the three-dimensional structure of the CHASE domain of ZmHK1 using the structure of AHK4. This was done with and without CK occupying the binding pocket, which did not change the results. Given the high degree of amino acid identity between ZmHK1 and AHK4, the ZmHK1 CHASE domain structure was resolved with high confidence. Next, each mutant receptor was modeled based on the derived ZmHK1 structure. The models were subjected to dynamics simulation with appropriate solvation (see Methods). The results of homology modeling showed that the amino acids mutated in each Hsf1 allele do not occur within the CK binding pocket ( Figure 1F) and thus do not contribute to direct polar contacts with the ligand. Instead, each altered residue is located near a loop domain that forms one face of the binding cavity. An indication of how the mutated residues at these positions might affect ligand binding was provided by the structure model of the ZmHK1-1603 receptor. The residue altered in ZmHK1-1603 is Glu-236, which is predicted to form an ion-pair interaction with Arg-192 located in the loop domain. This polar interaction may help stabilize the position of the loop domain ( Figure 1F). The Hsf1-1603 mutation converts Glu-236 to Lys, a negative-to-positive residue change, which is expected to break the polar interaction with Arg-192 and possibly destabilize the position of the loop due to the nearness of the two positively charged residues. Altering the position of the loop may change the overall conformation of the ligand binding pocket and thus account for differences in ligand binding affinities. The missense residues in the other two mutant receptors could potentially alter the conformation of the CK binding pocket via a different mechanism, although our modeling results did not reveal an obvious mechanism.

Exogenous CK Treatment Recapitulated the Hsf1 Phenotype
The biochemical analyses and heterologous signaling assays suggested the Hsf1 mutant receptor might be hypersignaling the perception of CK, which altered leaf development. To test the idea that increased CK signaling could produce Hsf1-like phenotypes, wild-type, B73 inbred seeds were transiently treated with the CK 6-benzylaminopurine (BA). The embryo in a mature maize seed possesses approximately five leaf primordia, and it is these primordia that experienced the hormone treatment (Kerstetter and Poethig, 1998). Imbibed seeds were treated for 6 d with 10 mM BA, rinsed, and transplanted to soil (see Methods). After growth for 3 weeks, the first four seedling leaves were examined for developmental changes (Figures 2A to 2G). Similar to Hsf1, 100% of the CK-treated B73 seeds produced smaller seedling leaves covered with abundant macrohairs (Figures 2A to 2E). Leaf sheath   length, blade length, and blade width were reduced by 10 to 20% for leaf 3, similar to leaf size reductions in the Hsf1 seedlings ( Figure 2C). In addition, macrohair density increased on the abaxial sheath, near the auricle, on the adaxial blade, and blade margins in 100% of the CK-treated B73 seedlings ( Figures 2D and 2E). This pattern of ectopic macrohair formation was similar to that seen in Hsf1 seedlings (Bertrand-Garcia and Freeling, 1991a). In addition to alterations in leaf size and pubescence (covering of fine hairs), nearly 20% of the CK-treated B73 seeds produced seedlings with prongs on leaf 4 ( Figures 2F and 2G). This was in contrast to Hsf1 seedlings where prongs rarely, if ever, developed on leaf 4 ( Figure 2I; Supplemental Figure 1C). Increasing the concentration of exogenous BA to 100 mM increased the number of B73 seedlings with prongs on leaf 4 to nearly 90% ( Figure 2G). Thus, transient, exogenous CK treatment recapitulated three prominent aspects of the Hsf1 phenotype: reduced leaf size, increased macrohair abundance, and formation of prongs on blade margins, confirming these developmental changes can be induced by CK. If CK hypersignaling in Hsf1 was due to increased ligand affinity, we would expect Hsf1 to be hypersensitive to CK treatment. To test this idea, we performed 6-d treatments on segregating Hsf1-1603/1 seeds using 0.1 mM CK, a concentration that did not elicit leaf size changes in B73 inbred seed (Supplemental Figure 6A). To distinguish segregating heterozygous Hsf1 plants from the wildtype sib plants, PCR genotyping was used to detect a size polymorphism in the Hsf1-1603 allele (Supplemental Table 3). After CK treatment, seedlings were grown for 3 weeks, after which leaf phenotypes were measured. While 0.1 mM CK treatment had no effect on the wild-type sib leaf size (Supplemental Figure 6A), it did reduce the leaf size of Hsf1-1603/1 plants by 10 to 30% (Supplemental Figure 6B). Thus, Hsf1-1603/1 seedlings were responsive to a lower concentration of CK that did not elicit a response in the wild-type sib or B73 inbred seedlings. Treatment with 10 mM BA was also used to assess effects on prong and macrohair formation in Hsf1-1603/1 plants. Similar to earlier results (Supplemental Figure 1C), seedlings from control watertreated Hsf1-1603/1 seeds first formed prongs on leaf 5 (;5%) or leaf 6 (;25%), but not on earlier arising leaves ( Figures 2H to 2J). In fact, ;60% of Hsf1-1603/1 seedlings normally first formed prongs on leaves arising on or after leaf 7 ( Figure 2I). By contrast, of the 10 mM BA-treated Hsf1-1603/1 seeds, nearly 60% produced seedlings where prongs first formed on leaf 4, and only ;30% formed prongs on leaves arising on or after leaf 7 ( Figures 2H to 2J). In addition, macrohair abundance appeared increased for CKtreated Hsf1-1603/1 compared to control Hsf1-1603/1 or the BAtreated wild-type sib seedlings, but this was not measured ( Figure 2K). Thus, CK treatment of Hsf1 resulted in earlier arising and enhanced mutant phenotypes, indicating the mutation was hypersensitive to the CK hormone, consistent with the biochemical analysis of the receptor.

CK-Responsive Genes Are Upregulated in Hsf1
Leaf Primordia Based on the Hsf1 mutant plant phenotypes, we presumed that hypersignaling in developing leaf primordia gave rise to the alterations in P-D leaf patterning and other phenotypes. To test this idea, we determined the expression of ZmHK1 and several CKresponsive genes in Hsf1/1 and the wild-type sib plants. Published qPCR and in silico expression analyses (https://www. maizegdb.org/gene_center/gene/Zm00001d017977#rnaseq) indicated ZmHK1 was expressed broadly across several tissues including leaves, roots, stem, and tassel (Yonekura-Sakakibara et al., 2004). We reverse transcribed cDNA from three tissues: shoot apices (shoot apical meristem plus the three youngest leaf primordia), immature leaf, and mature green leaf from 2-week-old seedlings. Using qPCR, we assessed expression in plants heterozygous for the three Hsf1 alleles compared with their wild-type sibs ( Figure 3A). We did not detect an increase in ZmHK1 transcript accumulation in the Hsf1/1 mutants compared to their wild-type controls. Next, we examined expression of CK-responsive genes: two type-A Z. mays Response Regulator genes, ZmRR3 and ZmRR2, and a Z. mays Cytokinin Oxidase gene, ZmCKO4b (Asakura et al., 2003;Giulini et al., 2004). We observed increased transcript accumulation for all three CK-responsive genes in the Hsf1/1 mutants, although there were some inconsistencies across genotypes and tissues ( Figure 3A).
Using in situ hybridization, we assessed transcript localization of ZmHK1 and ZmRR3 in the wild-type and Hsf1-1603/1 shoot apices ( Figure 3B). The ZmHK1 transcript was distributed broadly within developing leaf primordia and shoot apices in both genotypes ( Figure 3B). As was demonstrated previously, ZmRR3 was expressed in a specific wedge-shaped domain in the apical (C) Boxplots of leaf sizes comparing the wild type (WT) with Hsf1-1603/1 sib seedlings, and B73 water-treated (-CK) with 10 mM BA-treated (1 CK) seedlings. Horizontal bars represent the maximum, third quantile, median, first quantile, and minimum values, respectively; dots outside of the plot are outliers. The asterisk (*) indicates a P-value # 0.0001 calculated from a two-tailed Student's t test. (K) Macrohair production on 2-week-old seedlings due to CK treatment, Hsf1-1603/1 mutation, or both. meristem in both longitudinal and transverse sections of the wild-type apices, but no signal was detected in leaf primordia ( Figure 3B; Giulini et al., 2004). However, the spatial expression of ZmRR3 was expanded in Hsf1-1603/1 apices. Strong ZmRR3 expression was visible in its normal meristem domain, but signal was also detected in leaf primordia and was particularly evident at the margins of primordia P1 to P5 ( Figure 3B). Given the expanded pattern of ZmRR3 expression in Hsf1-1603/1 leaf primordia margins and that ZmRR3 expression is CK responsive, we interpreted this to indicate increased CK signaling in the tissue where prongs will eventually form.
Mutation of ZmRR3, a Negative Regulator of CK Signaling, Enhances the Hsf1 Phenotype To test whether the increased transcript accumulation of the CKresponsive genes was biologically relevant, we made use of a null allele of ZmRR3, also known as abph1. Plants homozygous for the recessive abph1 reference allele have an altered phyllotactic pattern and develop leaves paired ;180°at each node instead of having the normal alternating pattern (Figures 4A and  4B), but they have no P-D patterning defects (Jackson and Hake, 1999). Backcross families were produced that segregated four phenotypes-the wild type, heterozygous Hsf1-1603, homozygous abph1, and heterozygous Hsf1-1603 plus homozygous abph1-in equal frequencies ( Figures 4A and 4B). Double mutant plants, heterozygous for Hsf1 and homozygous for abph1, had a synergistic phenotype with a very stunted stature, a proliferative shoot with no obvious central axis, very slow growth, extremely short and narrow leaves, and new leaf morphology defects such as bi-or trifurcation of leaf blades ( Figure 4B; Supplemental Figure 7). The synergistic interaction of Hsf1 and abph1 was consistent with ZmRR3 functioning as a negative regulator of CK signaling. (A) Relative mRNA accumulation of CK genes in different tissues of 2-week-old seedlings of the three Hsf1 alleles and the wild-type (WT) sibs measured by qPCR. For each genotype, values are the means 6 SE of three biological replicates consisting of tissue pooled from at least three plants. Asterisks indicate significant differences between WT and Hsf1/1 sib (Student's t test, P # 0.05). GL, green leaf; IL, immature leaf; SA, shoot apex. (B) Pattern of ZmHK1 and ZmRR3 transcript accumulation in the wild type (WT) and Hsf1-1603/1 shoot apex. Longitudinal and transverse sections were hybridized with ZmHK1-or ZmRR3-specific antisense probes. The longitudinal section of ZmRR3 hybridized to WT is not medial, so ZmRR3 expression appears to be apically localized, but it is not. Initiating leaf primordia (yellow arrows) and leaf primordia margins (red arrowheads) are marked in the Hsf1/1 sections probed with ZmRR3. Bar 5 30 mm.

CK Influences Specific Developmental Programs in Maize Leaves
In this study, we showed that the Hsf1 mutation conditions a CK hypersignaling phenotype that has multiple effects on plant growth and development, including specific effects on (1) leaf patterning, (2) leaf size, and (3) leaf epidermal cell fate (Bertrand-Garcia and Freeling, 1991a). Supporting this idea, we also show exogenous CK treatment of the wild-type maize seeds produced similar changes in these developmental programs. Prominent among the developmental changes was a specific alteration in P-D leaf patterning where ectopic outgrowths with proximal identity (prongs) formed in the distal blade ( Figures 1A to 1C; Supplemental Figure 2A). Although growth along the P-D axis is fundamental to normal leaf development and morphology, its molecular control has not been fully characterized. In eudicots, the activities of several transcription factor genes, such as BLADE ON PETIOLE1 (BOP1), LEAFY PETIOLE (LEP), and JAGGED (JAG), have been linked to the control of P-D leaf development (van der Graaff et al., 2000Graaff et al., , 2003Ha et al., 2004;Ohno et al., 2004;Norberg et al., 2005). BOP genes have also been shown to influence P-D leaf patterning in several monocots, and recently the activity of three redundant OsBOP genes was shown to be required for sheath identity in two rice species (Tavakol et al., 2015;Toriba et al., 2019Toriba et al., , 2020. In several monocots, the misexpression of several class I knox genes perturb P-D patterning by potentially altering phytohormone ratios in developing leaf primordia (Schneeberger et al., 1995;Foster et al., 1999b;Reiser et al., 2000;Ramirez et al., 2009).
Our analysis of Hsf1, the second characterized mutation of a maize CK signaling gene, has uncovered a connection between CK and the specification of P-D leaf patterning that is consistent with this hypothesis. How CK drives prong formation is not clear, although the interplay of CK and GA is known to control the degree of leaf complexity in eudicots such as Arabidopsis (Arabidopsis thaliana) and tomato, through the specification of marginal lobes or leaflets (Jasinski et al., 2005;Bar and Ori, 2015). Whether there is any overlap between the mechanism(s) of prong formation in Hsf1 and leaflet formation in species such as tomato will require further analysis. Prong formation itself appears developmentally regulated as prong initiation seems to be coordinated with formation of the ligule, suggesting the signals establishing the P-D axis might be transmitted across the entire leaf primordium (Supplemental Figures 1D and 1E). Moreover, prong formation is not random as prongs form only within a certain domain of the blade, with the largest prongs forming more basally (Supplemental Figures 2D  and 2F). Intriguingly, this prong formation region has some overlap with the domain of the leaf blade deleted by mutation of the duplicate wuschel-related homeobox (wox) genes narrow sheath1 and narrow sheath2 (Nardmann et al., 2004). This implies that the marginal domain in the blade specified by these duplicate wox transcription factors may provide a permissive context for prongs to form, which can be tested by analysis of prong formation in the triple mutant.
Leaf sheath and blade length, and blade width, were reduced in Hsf1 heterozygotes compared to the wild-type sib plants at seedling and mature growth stages, consistent with previous reports, and CK treatment recapitulated this phenotype in wildtype inbred seedlings (Figures 2A to 2C; Bertrand-Garcia and Freeling, 1991b). Since CK activity typically promotes cellular proliferation, how CK hypersignaling reduces growth in the shoot is not known, although increased CK signaling is known to reduce root growth (Werner et al., 2001(Werner et al., , 2003. Typically, reducing CK accumulation or signaling results in smaller leaves and other aboveground organs, suggesting increased CK activity might be expected to enhance growth (Werner et al., 2001;Nishimura et al., 2004). Growth of the maize leaf is organized linearly along its longitudinal axis into distinct zones of cell division, cell expansion, and differentiation (Freeling and Lane, 1992). Recent transcriptome, proteome, and hormone profiling studies have enumerated multiple regulatory pathways controlling the size of and transitions between the different growth zones, with GA playing a prominent role (Li et al., 2010;Nelissen et al., 2012;Facette et al., 2013). How increased CK signaling impacts these growth zones to determine final leaf size will require further analysis building upon these previous studies.
In addition to a change in P-D patterning and reduction in leaf size, the Hsf1 mutation and CK treatment of the wild-type seed promoted increased macrohair formation in the leaf epidermis ( Figures 2D, 2E, and 2J). Macrohairs are normally found on adult leaves on the abaxial sheath, at high density near the ligule but declining basipetally, on the adaxial blade, and along the blade margin. Hsf1 increased macrohair production not only on the abaxial sheath, adaxial blade, auricle, and blade margins of adult leaves but also on juvenile and transitional leaves, which are typically glabrous. CK treatment phenocopied the increased pubescence phenotype of Hsf1 (Figures 2D and 2E). The epidermis of the maize leaf has three types of pubescence: bicellular microhairs, macrohairs, and prickle hairs, with macrohairs being the most prominent (Freeling and Lane, 1992). Macrohairs form by differentiation of specialized epidermal cells organized in patterned files beginning in the fifth or sixth leaf (Moose et al., 2004). Little is known regarding the signals specifying macrohair formation, although a recessive mutation affecting macrohair initiation, macrohairless1, has been reported previously (Moose et al., 2004). By contrast, trichome differentiation in the leaves of eudicots, such as Arabidopsis, is known to be controlled by a core network of positive and negative transcriptional regulators (Ishida et al., 2008;Grebe, 2012;Pattanaik et al., 2014). Furthermore, trichome initiation on the inflorescence organs in Arabidopsis is jointly stimulated by the activity of CK and GA and downstream transcription factors (Gan et al., 2007;Zhou et al., 2013). The increase in macrohair formation mediated by CK treatment or the Hsf1 mutant suggests CK can reprogram epidermal cell fate in maize leaves.

Missense Mutations in the Maize CK Receptor ZmHK1
Underlie the Hsf1 Phenotype Our data indicate gain-of-function mutations of the maize CK receptor ZmHK1 underlie the semidominant Hsf1 mutations. CK signaling, which is well described (To and Kieber, 2008;Hwang et al., 2012), regulates several developmental and physiological processes, although influences on leaf patterning are not among them. For example, combinations of loss-of-function mutations of the three Arabidopsis CK receptors demonstrate this gene family has partially overlapping and redundant functions in the control of shoot and root growth, seed size, germination, and leaf senescence (Higuchi et al., 2004;Nishimura et al., 2004;Riefler et al., 2006). CK receptors were shown to also possess phosphatase activity by analysis of a specific mutation of AHK4/CRE1, the recessive wooden leg (wol) allele [CRE1(T278I); Mähönen et al., 2006]. Plants homozygous for the wol allele have abnormal root vascular development due to the dose-dependent constitutive phosphatase activity of this allele. A gain-of-function mutation in the CHASE domain of AHK3 (ore12-1) revealed this receptor plays a major role in CK-mediated leaf senescence, although how this mutation affected receptor activity was not explored (Kim et al., 2006). The study of gain-of-function mutations has revealed additional information on CK receptor function. Novel, dominant, missense mutations in AHK2 and AHK3, the repressor of cytokinin deficiency alleles (rock2 and rock3), enhanced CK signaling, increased CK hypersensitivity, and increased transcript accumulation of CK-responsive genes, similar to the Hsf1 mutations ( Figure 3; Bartrina et al., 2017). By contrast, the rock mutations had the opposite effect on phenotype compared to Hsf1, producing early flowering, enlarged rosette leaves and shoots, and longer roots. The contrasting phenotypic effects might be due to differences in signaling strength between the rock and Hsf1 mutations or reflect differences in the downstream circuitry between the two species.

Mutations near the CK Binding Pocket Alter Ligand Affinity and Receptor Signaling
To clarify how the function of ZmHK1 was altered in the Hsf1 mutants, we analyzed their activity in heterologous HK signaling and ligand binding assays. Our results indicate two of the Hsf1 mutant receptors signal independently of added CK in yeast, and all three have increased binding affinities for the natural CKs tested ( Figure 1E; Table 1). The mutant receptors may be in a locked-on state, similar to what was hypothesized for the snf2 mutation or the increased ligand affinities of the Hsf1 receptors may explain their ability to signal independently of CK action. We favor the second idea and think the increased CK affinity explains the ability of the mutant receptors to signal in heterologous hosts. Many microbes, including E. coli and yeast, contain low concentrations of iP as a normal constituent of tRNA, which can become free due to tRNA decay (Skoog and Armstrong, 1970;Hall, 1973;Romanov, 1990;Mok and Mok, 2001). The three mutant receptors all have increased affinity for iP (Table 2). This stronger affinity may be due to stronger complex formation, or longer receptor occupancy, and, as a consequence, stronger signaling even in the presence of low iP concentration. Thus, the ability of the ZmHK1-AEWL and ZmHK1-1603 receptors to signal in yeast without added CKs may be due to their increased affinity for iP already present at low concentration in yeast cells ( Figure 1E). In fact, it has been shown that expressing other HK receptors in the sln1 deletion yeast strain TM182 permits this strain to grow on Glc without added CKs, albeit at a much slower rate than with CKs present, and recombinant HKs synthesized in E. coli cannot be crystalized without iP complexed in the binding pocket (Higuchi et al., 2009;Hothorn et al., 2011). Since both the ZmHK1-AEWL and ZmHK1-1603 receptors show signaling activity in yeast without ligand ( Figure 1E), have increased ligand affinities (Table 1), have nearly identical mutant plant phenotypes in several different genetic backgrounds (Figure 1; Supplemental Table 1), and show similar misexpression patterns of CK-responsive genes (Figures 3A), we conclude these two Hsf1 mutant receptors function similarly in planta. How the ZmHK1-1595 receptor confers the Hsf1 phenotype is less clear, since it has increased ligand affinities (Table 1), but does not show ligand-independent signaling activity, and has weak ligand-dependent signaling activity in yeast ( Figure 1E). As an apparently weak signaling receptor, ZmHK1-1595 may trigger compensatory upregulation of other ZmHKs, or decreased signaling output may result in lowered production of a negative regulator that might enhance the CK response. These and other hypotheses need to be tested.
Our structural analysis localized each residue mutated in Hsf1 to the ligand binding Per-Arnt-Sim-like subdomain of the CHASE domain in ZmHK1 ( Figure 1F; Hothorn et al., 2011;Steklov et al., 2013). Notably, none are within the CK binding pocket or predicted to make contact with the ligand; rather, all are located near a loop domain comprising one face of the pocket, suggesting interactions with this loop may affect the binding pocket, resulting in increased ligand affinity. Interestingly, amino acid substitutions that rendered AHK4 constitutively active in a heterologous E. coli HK assay were located downstream of the CHASE domain in the second transmembrane domain and near the kinase domain (Miwa et al., 2007). In addition, none of the rock mutations are located in the ligand binding Per-Arnt-Sim-like domain (Bartrina et al., 2011). Rather two are in the N-terminal a helices and one is in the C-terminal transmembrane domain. Therefore, further structure-function studies are needed to define which residues are crucial for activity and to resolve the precise mechanism(s) by which individual missense mutations alter ligand binding and receptor signaling.

Hsf1 Affects Downstream Components of CK Signaling
More ZmHK1 signaling in developing Hsf1 leaf primordia resulted in increased transcript accumulation of several early CKresponsive genes in all three Hsf1 mutant alleles ( Figure 3A). Although not all CK reporters responded the same within an allele or tissue, overall our data are consistent with Hsf1 upregulating CKresponsive genes. The most consistent effect was upregulation of ZmRR3 where its normally meristem-confined expression was expanded in Hsf1-1603 to include expression near newly arising leaf primordia and in primordia margins ( Figure 3B). Notably, the increased CK signaling reported by ZmRR3 marks the margins of early-stage leaf primordia ( Figure 3B), which is where prongs will form later in development (Supplemental Figures 1F and 1G). Although we found ectopic ZmRR3 signal along the entire margin, outgrowths do not emanate from the entire blade margin but rather occur sporadically, with outgrowths interspersed with regions of normal blade margin (Figures 1B and 1C; Supplemental Figure 2A). This observation suggests even though CK hypersignaling can promote proximalization of blade margin cells, not all cells at the margin are competent to respond to this signal. Double mutants heterozygous for Hsf1-1603 and homozygous for abph1, a null allele of ZmRR3, show a synergistic interaction (Figure 4; Supplemental Figure 7). Several type-A RRs function to negatively regulate CK signal transduction as well as regulate circadian rhythms, phytochrome function, and meristem size . The increased severity of growth defects in Hsf1 heterozygotes that lack abph1 activity suggests upregulation of ZmRR3 (abph1) partially ameliorates CK hypersignaling. This also suggests that ZmRR3 normally functions to attenuate CK signal transduction in maize shoot apices, in addition to specifying leaf phyllotaxy.
The identification of the CK receptor ZmHK1 as the gene underlying the leaf patterning Hsf1 mutation adds to our understanding of the role CK plays in basic developmental programs. Future studies to determine the molecular determinants functioning downstream of CK signaling that promote prong formation should illuminate mechanisms important for developmental reprogramming and cell fate acquisition.

Plant Material, Genetics, Phenotypic Measurements, and Analysis
The Hsf1-1595, Hsf1-1603, and Hsf1-2559 mutants arose via EMS mutagenesis of the inbred Mo17, and seed was obtained from the Maize Genetic Cooperation Stock Center (http://maizecoop.cropsci.uiuc.edu/). Hsf1-AEWL arose via EMS mutagenesis of the inbred A619 and Hsf1-7322 via EMS mutagenesis of the inbred A632 in independent screens. Homozygous Hsf1 mutants of all five alleles were identified for sequence analysis from progeny of self-pollinated heterozygous B73 introgressed plants by phenotype and also by PCR screening of linked sequence polymorphisms unique to the progenitor inbred lines and the backcross inbred B73. Since Hsf1-1595 and Hsf1-1603 were the same transition mutations as Hsf1-7322 and Hsf1-2559, respectively, further analysis was only performed on three mutants: Hsf1-1595, Hsf1-1603, and Hsf1-AEWL. All phenotypic, molecular, and epistatic analyses were performed on the three alleles that had been backcrossed a minimum of six times to the inbred B73. The Hsf1 phenotype of the three alleles was fully penetrant as a heterozygote in all backcross generations. Progeny from self-or sibpollinated Hsf1 heterozygotes of the three alleles segregated 25% severely stunted, very slow growing, multi-shoot plants that only survived when grown in the greenhouse but were sterile. The abph1 mutant seed was backcrossed a minimum of three times to the inbred B73 before making the double mutant family segregating with Hsf1-1603. Hsf1-1603 heterozygotes were crossed by abph1 homozygotes and double heterozygous progeny plants were backcrossed by abph1 homozygotes, creating double mutant families segregating 25% 1/1 abph1/1 (wild type), 25% 1/1 abph1/abph1 (single abph1 mutant), 25% Hsf1/1 1/abph1 (single Hsf1 mutant), and 25% Hsf1/1 abph1/abph1 (double Hsf1 abph1 mutant).
Allele-specific PCR genotyping was done to confirm phenotypes of Hsf1 heterozygotes and abph1 heterozygotes and homozygotes (Supplemental Table 3). PCR amplifications were performed using Econotaq PLUS GREEN 2x Master Mix (Lucigen) following the manufacturer's recommendations on an S1000 thermal cycler (Bio-Rad). PCR conditions were 98°C for 5 min, followed by 35 cycles at 98°C for 30 s, 58°C for 45 s, 72°C for 30 s, and a final extension of 72°C for 5 min.
Measurement of adult plant traits of the three Hsf1 mutant alleles was performed on field-grown families segregating 50% wild type:50% Hsf1 heterozygotes. Measurements were taken on 7 to 11 plants of each genotype in one-row plots with two biological replicates. For analysis of prong position, prong size, and percent prong margin, the third leaf above the ear of adult Hsf1 heterozygous plants was collected from one-row plots of field-grown plants in three replicates in summer 2013. Approximately 6 to 10 leaves were collected per plot for each allele. For each leaf, measurements were made for (1) total blade length, (2) prong position by measuring the distance from the base of the blade to the midpoint of each prong on each blade margin, and (3) prong size by measuring from the basal to the distal position along the margin where proximal tissue emerged from the blade for each prong ( Figure 1B). Percent prong margin was defined as the proportion of leaf blade margin that is occupied by tissue having proximal (sheath, auricle, and/or ligule) identity and was calculated by summing the size of all prongs from both sides of the leaf blade divided by twice the length of the leaf blade.
Analysis of prong position, prong size, and the relationship between prong position and size was estimated with kernel smoothing methods (Silverman, 1986;Wand and Jones, 1995). For all cases, a Gaussian kernel was used and the data reflection method was applied for boundary correction since both position and size are positive variables. The bandwidth was selected using least-squares cross-validation (Bowman, 1984). All computations were performed using R software, kernel density estimation was performed using the ks package (Duong, 2007), and figures were created with the ggplot2 package (Wickham, 2009).

Map-Based Cloning of Hsf1
Hsf1-1595 was introgressed into B73 and crossed to PRE84 which was then used to generate a backcross one (BC1) mapping population. Genetic mapping with 96 BC1 individuals defined Hsf1 between two singlenucleotide polymorphism markers on chromosome 5: PHA12918-F (204590502 bp, B73 RefGen_v2) and PHA5244-F (206614542 bp, B73 RefGen_v2). The two flanking markers were used to screen a BC1 population of 1500 individuals from B73_Hsf1 3 A632 and 1600 individual from B73_Hsf1 3 PRE84. 224 recombinants were identified, and these individuals were used for further fine mapping. Additional markers derived from the Hsf1 interval were developed and used to fine map the Hsf1 mutation with the recombinants, as described in Jiang et al. 2012). The gene underlying the Hsf1 mutation was finally delimited to a 21-kb interval, between Indel marker 410984 (205538463 bp, B73 RefGen_v2, with one recombinant between this marker and Hsf1) and single-nucleotide polymorphism marker 391087 (205559234 bp, B73 RefGen_v2, with three recombinants between this marker and Hsf1). There is only one annotated gene model (B73 RefGen_v3 GRMZM2G151223, B73 RefGen_v4 Zm00001d017977) in this interval that was also annotated in National Center for Biotechnology Information (NCBI) as LOC541634 HK1 a putative CK receptor.

Heterologous HK Assays
Signaling of the wild-type and Hsf1 mutant ZmHK1 receptors in yeast was performed as described previously (Inoue et al., 2001). The exact point mutations for each of the three Hsf1 missense mutations were engineered into the cDNA of ZmHK1 in the plasmid P415-CYC1-ZmHK1 plasmid with the QuikChange II site-directed mutagenesis kit (Agilent Technologies) using the manufacturer's specifications.

Expression of ZmHK1 Protein in Bacteria
Wild-type ZmHK1 and the three Hsf1 missense mutants were expressed in an Escherichia coli strain as described previously (Lomin et al., 2011). Protein samples isolated from overnight cultures by sonication were mixed with SDS-PAGE loading buffer and incubated at 37°C for 1 h. Proteins were separated by running equal volumes on 10% (v/v) SDS-PAGE (Laemmli, 1970) and then transferred to Immuno-Blot polyvinylidene fluoride transfer membranes (Bio-Rad). Membranes with transferred proteins were preincubated in PBS with 5% (w/v) skim milk for 1 h. For receptor detection, monoclonal mouse-derived primary antibodies against the C terminus of ZmHK1 (gift of Hitoshi Sakakibara's laboratory) were diluted 1:3000 as described previously (Lomin et al., 2011). Incubation with primary antibodies was performed in PBS with 5% (w/v) skim milk for 2 h and then the membranes were washed three times in PBS with 0.05% (v/v) Tween 20. Goat anti-mouse IgG secondary antibodies conjugated with peroxidase purchased from Merck (Calbiochem) were diluted 1:10,000. Incubation with secondary antibodies was performed for 1 h, followed by washing four times in PBS with 0.05% (v/v) Tween 20. The peroxidase reaction was performed at room temperature for 3 min using 0.1 M Tris-HCl, pH 8.5, 1.25 mM luminol, 0.68 mM p-coumaric acid, and 0.01% (v/v) hydrogen peroxide solution as substrate, and products were detected by CL-XPosure Clear Blue X-Ray film (Thermo Fisher Scientific).

CK Binding Affinity Determination
CK binding assays were performed with recombinant maize CK receptors expressed in E. coli cells. Spheroplasts were prepared from cell lines expressing the wild-type ZmHK1, and mutant ZmHK1-AEWL and ZmHK1-1603 receptors. Competitive CK binding assays were performed in duplicate or triplicate as described previously (Lomin et al., 2011;Donaldson et al., 2019). Tritium-labeled [2-3 H]trans-zeatin ( 3 H-tZ; 592 GBq mmol 21 , radiochemical purity >99%) was obtained from the Institute of Experimental Botany (Prague, Czech Republic). Transient expression of ZmHK1 and ZmHK1-1595 receptors for the homologous binding assay was done by transformation of Nicotiana benthamiana as described previously (Sparkes et al., 2006;Lomin et al., 2015). Agrobacterium tumefaciens carrying CK receptor genes fused to GFP were grown in parallel with a helper agrobacterial strain p19 (Voinnet et al., 2003). Five-to 6-week-old tobacco plants were infiltrated with the mixture of two agrobacterial strains, and the expression level of receptor genes was checked after 4 d using a confocal microscope. For those cases with sufficient expression, leaves were processed further for plant membrane isolation. For plant membrane isolation, all manipulations were done at 4°C. N. benthamiana leaves were homogenized in buffer containing 300 mM Suc, 100 mM Tris-HCl, pH 8.0, 10 mM Na 2 -EDTA, 0.6% (w/v) polyvinylpyrrolidone K30, 5 mM K 2 S 2 O 5 , 5 mM DTT, and 1 mM phenylmethylsulfonyl fluoride. The homogenate was filtered through Miracloth (Calbiochem), and the filtrate was first centrifuged for 10 min at 10,000g at 4°C and then for 30 min at 100,000g at 4°C. The microsome pellet was resuspended in PBS, pH 7.4, and frozen and stored at -70°C before using.

ZmHK1 Structure Modeling
The amino acid sequence of the ZmHK1 CHASE domain (86 to 270) was obtained from the protein sequence database of NCBI (accession ID: NP_001104859). It shares 71% sequence identity with the Arabidopsis HK4 sensor domain. The homology model for ZmHK1 was generated using Swiss model server (http://swissmodel.expasy.org) with the published crystal structure of AHK4 (pdb code: 3T4J) as a template. Subsequently, the model was solvated and subjected to energy minimization using the steepest descent followed by conjugate gradient algorithm to remove clashes. The stereochemical quality of the ZmHK1 model was assessed using the PROCHECK program. None of the residues were in the disallowed regions of the Ramachandran map.

Exogenous CK Treatment
Exogenous CK treatments were performed with BA (Sigma-Aldrich) dissolved in 10 drops 1 N NaOH and brought to 1 mM concentration with distilled water. All water control treatments were done using a similar stock of 10 drops of 1 N NaOH and diluted in parallel to the CK stock. Further dilutions to the desired CK concentration were done with distilled water. Maize kernels were surface sterilized with two 5-min washes of 80% (v/v) ethanol followed by two 15-min washes of 50% (v/v) bleach and rinsed five times in sterile distilled water. Kernels were imbibed overnight with sterile distilled water prior to the start of the hormone treatment. For hormone treatments, 20 imbibed kernels per replicate were placed embryo-side down on two paper towels in a Petri dish, covered with two more layers of paper towel, and filled with 15 mL of CK treatment or the water control solution. Petri dishes were sealed with parafilm and placed in a lab drawer in the dark at room temperature for 6 d. After treatment, germinating kernels were rinsed with sterile distilled water and planted in 4-cm square pots in soilless potting medium (Metro-Mix 900, SunGro Horticulture) and grown in the greenhouse (day: 16 h/28°C; night: 8 h/21°C) with supplemental lighting (high pressure sodium and metal halide lights) and standard light intensity (230 mE m 22 s 21 at height of 3.5 feet [107 cm]). Growth was monitored, and leaf measurements were taken after the fourth leaf collar (auricle and ligule) had fully emerged from the whorl after 3 to 4 weeks. For measurements, individual leaves were removed from the plant, and each component was measured. Leaf sheath length was defined as the site of insertion of the leaf base to the culm (stem) to the farthest point of sheath adjoining the ligule. Leaf blade length was defined as the most proximal point of blade adjoining the ligule to the distal blade tip. Leaf blade width was measured margin to margin at half of the leaf blade length. All leaf measurements were analyzed using JMP PRO 12 software using a Student's t test to determine significance with two comparisons and Tukey's honestly significant difference test to determine significance with more than two comparisons (Supplemental Data Set). To examine macrohair abundance, epidermal impressions were made using Krazy Glue Maximum Bond cyanoacrylate glue applied to a Fisherbrand Superfrost Plus microscope slide. The adaxial blade of leaf one was pressed firmly into the glue for ;30 s, followed by immediate removal of the leaf. Slides were imaged on an Olympus BX60 light microscope.

In Situ Hybridization
For in situ hybridization, we slightly modified an online protocol from Jeff Long (University of California Los Angeles). For complete details, refer to www.its.caltech.edu/;plantlab/protocols/insitu.pdf. In situ probes were made using T7/SP6 promoter-based in vitro transcription in the cloning vector pGEMT (Promega). Formaldehyde acetic acid-fixed and paraffinembedded maize shoot apices were sectioned at 10-mm thickness and laid on Probe-On-Plus slides (Thermo Fisher Scientific) and placed on a warmer at 42°C. After overnight incubation, the slides were deparrafinized using Histo-Clear (National Diagnostics), treated with Proteinase K, and dehydrated. Probes were applied on the slides, and pairs of slides were sandwiched carefully and incubated at 55°C overnight. The following day, the slides were rinsed and washed. Diluted (1:1250) anti-digoxigenin antibody (SKU 11093274910; Roche) was applied to the slides and incubated for 2 h. After thoroughly washing the slides, sandwiched slides were placed in 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium solution (Roche; 200 mL in 10 mL of buffer C [100 mM Tris, pH 9.5, 100 mM NaCl, and 50 mM MgCl 2 ]) in dark for 2 to 3 d for color development. Color development reaction was stopped using 13 Tris EDTA. The slides were mounted using Immu-Mount (Thermo Fisher Scientific) and observed and imaged under a bright-field microscope.

RT-qPCR
Seedling tissue was collected from 2-week-old, vegetative stage V3-to-V4 Hsf1/1 and the wild-type sib seedlings for each allele and included (1) ;2 cm of mature green leaf blade from the distal half of leaf 4; (2) approximately a 2-cm cylinder of immature leaf tissue, cut ;1 cm above the insertion point of leaf 5 after removing leaf 4; and (3) the remaining 1-cm cylinder of tissue above the insertion point of leaf 5, consisting of the shoot apical meristem, young leaf primordia, and the apical part of the stem. Tissue was bulked from three different plants for each biological replicate, and three replicates were collected. Total RNA was extracted from these tissues using TRIzol reagent, adhering to the manufacturer's protocol (http://tools.lifetechnologies.com/content/sfs/manuals/trizol_reagent. pdf). cDNA was synthesized from total RNA using the SuperScript III First-Strand synthesis system (Invitrogen) for RT-PCR and oligo(dT) primers. Quantitative real-time PCR was performed on the cDNA using an LC480 (Roche) and the SYBR Green assay. The primers were designed near the 39 end of the gene with an amplicon size of between 120 and 250 bp of folylpolyglutamate synthase was used as an endogenous control as it was shown to have very stable expression across a variety of maize tissues and range of experimental conditions (Manoli et al., 2012). Two technical replicates were included for each gene. The comparative DDCt method was used to calculate fold change compared to the endogenous control. DCt of mutant (Hsf1) and DCt of the wild type was expressed as the difference in cycle threshold (Ct) value between target gene and the endogenous control. DDCt was then calculated as the difference of DCt (Hsf1) and DCt (wild type). Finally, fold change in target gene expression between Hsf1 and the wild type was determined as 2 -DDCt .

Supplemental
Supplemental Table 2. Frequency of prongs for the three Hsf1 alleles by leaf position in the upper shoot.
Supplemental Table 3. Primers used for positional cloing or genotyping.
Supplemental Table 4. Primers used for expression analysis.
Supplemental Data Set. Statistical analysis.

ACKNOWLEDGMENTS
We thank Dave Jackson (Cold Spring Harbor Laboratory) for the abph1 mutant seed, Aimee N. Uyehara (University of Hawaiʻi at M anoa) for help with figure production, Erica Unger-Wallace (Iowa State University) for technical assistance, and Erik Vollbrecht (Iowa State University) for support during the early phase of this research. We also thank Thomas Schmülling (Free University Berlin) for providing the opportunity to perform some of experiments in his laboratory. This work was supported by the National Science Foundation (grant 1022452)