The anterior Hox gene ceh-13 and elt-1/GATA activate the posterior Hox genes nob-1 and php-3 to specify posterior lineages in the C. elegans embryo

Hox transcription factors play a conserved role in specifying positional identity during animal development, with posterior Hox genes typically repressing the expression of more anterior Hox genes. Here, we dissect the regulation of the posterior Hox genes nob-1 and php-3 in the nematode C. elegans. We show that nob-1 and php-3 are co-expressed in gastrulation-stage embryos in cells that previously expressed the anterior Hox gene ceh-13. This expression is controlled by several partially redundant transcriptional enhancers. These enhancers act in a ceh-13-dependant manner, providing a striking example of an anterior Hox gene positively regulating a posterior Hox gene. Several other regulators also act positively through nob-1/php-3 enhancers, including elt-1/GATA, ceh-20/ceh-40/Pbx, unc-62/Meis, pop-1/TCF, ceh-36/Otx, and unc-30/Pitx. We identified defects in both cell position and cell division patterns in ceh-13 and nob-1;php-3 mutants, suggesting that these factors regulate lineage identity in addition to positional identity. Together, our results highlight the complexity and flexibility of Hox gene regulation and function and the ability of developmental transcription factors to regulate different targets in different stages of development.


Introduction
Hox genes encode conserved transcription factors famously expressed in specific positions along the anterior-posterior axis during animal development that specify axial position. Mutations in Hox genes cause a wide variety of developmental defects in both model organisms and humans. Hox gene regulation is complex and includes both transcriptional and post-transcriptional control. In most animals, Hox genes are found in genomic clusters, and their expression along the A-P axis is collinear with their genomic position, and often show "posterior dominance," where posterior Hox genes repress the expression of more anterior Hox genes (Reviewed in [1][2][3]).
Intriguingly, despite their apparently opposite roles in anterior vs posterior morphogenesis, both ceh-13 and nob-1 are expressed in overlapping posterior lineages during gastrulation and both require the Wnt pathway for this expression [24,25]. ceh-13 is transiently expressed in the progeny of 7 of the 8 posterior sister cells derived from the (largely ectoderm-producing) AB blastomere at the 24-cell stage [21]. nob-1 is expressed 1-2 cell cycles later, in the posterior daughters or grand-daughters of four of these ceh-13-expressing cells [23]. In addition, both ceh-13 and nob-1 are expressed in the posterior daughter of the intestinal blastomere E ("Ep"). This raises the question of whether ceh-13 regulates nob-1 in these lineages. Expression of both factors at later stages is regulated by feedback mechanisms; early ceh-13 activity is required for later ceh-13 expression [25], and early nob-1 negatively regulates later nob-1 expression through the microRNA mir-57 [23].
Here we analyzed the cis-regulatory control of nob-1/php-3 expression in the early embryo. We find that nob-1/php-3 expression is regulated by several partially redundant distal enhancers, including at least three that drive overlapping patterns during gastrulation. We find that ceh-13 is required for normal timing and levels of nob-1/php-3 expression, by activating at least two of its enhancers. We further identified the GATA family transcription factor gene elt-1, previously known as a specifier of epidermal fate, as one of several additional positive regulators of nob-1/php-3 expression. Detailed analysis of cell positions and cell division timing identifies both position defects and defects in cell division patterns and timing in ceh-13 and nob-1/ php-3 mutants. This role in cell identity combined with the positive regulation of a posterior Hox gene by an anterior Hox gene suggest novel roles for Hox genes in early lineage specification.

The anterior Hox gene ceh-13 and the posterior Hox gene nob-1 are expressed sequentially in gastrulating embryos
To better understand the expression dynamics of the ceh-13, nob-1, and php-3 Hox genes during embryogenesis (Fig 1), we collected 3D time-lapse movies (~1.5 minute temporal resolution) of embryos expressing either ceh-13, nob-1 or php-3 modified at the endogenous locus by CRISPR/Cas9 genome editing to tag the encoded proteins with green fluorescent protein (GFP) at the C-termini. The same embryos also expressed a ubiquitously expressed mCherrytagged histone transgene to allow for cell lineage tracing. We traced cells from soon after fertilization (4-8 cell stage) through the last round of cell divisions for most cells (bean stage) by using StarryNite automated cell tracking software, and quantified reporter expression in each cell across time [26][27][28][29]. Note, each endogenously tagged Hox::GFP fusion line is homozygous viable, fertile, and displays no obvious phenotypes, demonstrating that the fusion proteins largely function like the wild-type proteins.
Consistent with previous studies using shorter reporters [21], endogenously tagged CEH-13::GFP expression is first seen in eight Wnt-signaled cells (that are the more-posterior daughter after cell division) and their progeny at the 26-cell stage, during early gastrulation (Figs 1B, 1C, and S2). These include the AB lineage-derived cells ABalap, ABalpp, ABarpp, ABplap, ABplpp, ABprap and ABprpp, and the posterior endoderm progenitor Ep. For simplicity, we use the term "lineage" to denote all cells derived from a specific blastomere; thus, the "Ep lineage" would refer to all descendants of Ep. The ABalap and ABalpp lineage expression was barely detectable and quickly faded, while the other lineages had more robust and persistent fluorescence, indicating the protein was degraded or maintained in distinct cells. By the 200-350 cell stage, many of the granddaughters of the initially expressing cells have lost CEH-13:: GFP expression, while a few cells fated to become blast cells or neurons sustained or increased their expression. Also at this stage, some cells in the D and MS lineages as well as a few additional AB-derived cells show CEH-13::GFP accumulation (Fig 1D). We compared the tagged protein expression pattern to that of endogenous mRNA as measured in a lineage-resolved single cell RNA-seq dataset, and found that the mRNA and tagged protein expression patterns were consistent [30]. In comparison, a rescuing 35 kb CEH-13::GFP fosmid transgene had brighter expression in the cells expressing the endogenously tagged protein, plus additional weak expression detectable in some epidermal precursors from the C lineage (possibly detectable due to higher copy number). A shorter reporter containing only 8.2 kb of upstream sequence lacked expression in the MS lineage, emphasizing the role of distal regulatory elements in regulating C. elegans Hox gene expression (S1D Fig) [31,32]. elegans embryos carrying two transgenes, a ubiquitous fluorescent histone to mark all nuclei (shown in green) and a reporter of interest (shown in red), which can be a cis-regulatory element driving a fluorescent histone (transcriptional) or a fluorescently tagged transcription factor protein (translational). Image analysis software identifies nuclei, and quantifies reporter intensity within the nuclei, which can be displayed as a lineage tree colored by expression as in (C). B) 3D projections of nuclei expressing endogenously tagged CEH-13::GFP (red) and endogenously tagged NOB-1::GFP and PHP-3::GFP (blue), with overlap shown in magenta, at 120 minutes (50 cell stage), 150 minutes (100 cell stage), and 280 minutes (400 cell stage) post fertilization. Endogenously tagged NOB-1:: GFP and PHP-3::GFP were expressed in indistinguishable patterns (S1 Fig), but PHP-3::GFP was slightly brighter so is shown in the following panels. C) Lineage tree through the 100-cell stage, showing early expression of endogenously tagged CEH-13::GFP and PHP-3::GFP, colored as in (B). D) Expressing lineages showing endogenously tagged CEH-13::GFP and/or PHP-3::GFP expression to 350 minutes of development, colors as in (B). Note that CEH-13::GFP Next, we defined the embryonic expression patterns of the posterior Hox genes php-3 and nob-1 tagged with GFP at the endogenous locus. As expected, nob-1 and php-3, which are adjacent in the genome with php-3 directly downstream of nob-1, are expressed in identical patterns in both our GFP imaging data (S1A Fig) and in the single-cell RNA-seq data [30]. Expression of both NOB-1::GFP and PHP-3::GFP begins during mid-gastrulation in cells derived from the cells ABplapp, ABplppp, ABprapp, ABprppp, Ep, and the four posterior great-granddaughters of the C blastomere. Expression becomes stronger one cell cycle later in the AB-derived lineages with the exception of that of ABp(l/r)appa. Additional late embryonic expression occurs in the ABp(l/r)papppp lineages starting at the onset of morphogenesis ("bean" stage). NOB-1-expressing lineages give rise to fates including neurons, hypodermis, seam cells, epithelial cells, intestine, and death, and have posterior positions clustered in and near the developing tail at comma stage [33]. Like CEH-13, early NOB-1::GFP expression persists through the terminal cell divisions in only a subset of cells ( Fig 1D). A previously described NOB-1::GFP rescuing transgene that includes 9kb of sequence upstream of the nob-1 transcript expresses GFP in a similar pattern, except it is not expressed in the C great-granddaughter lineages, suggesting this expression requires additional regulatory sequences outside of this region (Figs 1D and S1) [23].
Early CEH-13::GFP expression occurs in cells that are distributed broadly along the anterior-posterior axis (Fig 1B), while endogenously tagged NOB-1::GFP and PHP-3::GFP expression is more limited to the posterior of the embryo. In the AB lineage, NOB-1/PHP-3 are expressed exclusively in cells that were the posterior sister after cell division and whose mother expressed CEH-13 (Fig 1D-1H). In addition, these three genes are co-expressed at similar onset times in the posterior intestine lineage derived from Ep. In contrast, in late embryos, high levels of CEH-13::GFP and NOB-1::GFP/PHP-3::GFP are largely mutually exclusive (Figs 1D, S1F, and S1G), consistent with classic models of Hox expression and posterior dominance.

Several overlapping lineage specific enhancers regulate nob-1 embryonic expression
To identify regulators of nob-1 embryonic expression, we tested nearby genomic sequences for embryonic cis-regulatory activity (Fig 2A and 2B). Previous work showed that other C. elegans Hox genes are regulated by distal enhancers located as much as 20kb from a given gene's promoter [25,31,34], but enhancers for nob-1 and php-3 have not been identified. We took advantage of existing reporters of different lengths to narrow the sequence search space for nob-1 embryonic enhancers (Fig 2C). A transcriptional reporter containing just 5.3 kb of upstream sequence driving histone-mCherry reporter expression, and the rescuing 9kb NOB-1::GFP reporter are expressed in most of the same lineages as endogenous NOB-1::GFP, indicating they contain regulatory sequences sufficient for expression in the full set of expressing lineages. However, while the protein fusion reporters are expressed at similar levels in each lineage, the shorter 5.3kb transcriptional reporter drives much lower fluorescence in the ABp(l/r)ppp lineages compared to other lineages (Fig 2C and 2D). This suggests that additional sequences between -5.3kb and -9kb are required for the full endogenous expression levels in the ABp(l/r) ppp lineage. precedes PHP-3::GFP and NOB-1::GFP in all lineages except ABp(l/r)papppp (asterisk-expression is consistent in ABprpapppp and variable in ABplpapppp). (E-H) Quantitative detail for highlighted lineages, showing nuclear fluorescence intensity of CEH-13::GFP (fosmid) and NOB-1::GFP transgene reporters across developmental time for the cells leading to ABp(l/r)appaaa (E), ABp(l/r)appppp (F), ABp(l/r)ppppppp (G), and Ep(l/r)p (H). For cell labels, x = (l/r). Nuclear fluorescence intensity is in arbitrary units. Grey bars mark cell divisions.
https://doi.org/10.1371/journal.pgen.1010187.g001 showing the genes (black), candidate enhancers tested (brown), sequence conservation with other nematodes (grey), and NHR-2 ChIPseq trace (dark blue) from modENCODE [43] and ATAC-seq traces (red:ABa lineage, green: ABp lineage) from Charest et al, 2020 [66]. B) Schematics of the nob-1 reporter constructs examined, shown in alignment with (A). Enhancers were primarily tested in an orientation 5' to the pes-10 minimal promoter, but a downstream orientation was also used for the -4.3 kb enhancer. C) Lineage trees colored to show expression patterns for the various reporters and tested enhancers with relevant reproducible activity, using a rainbow color scale to increase visible dynamic range. Major expressing lineages are underlined: ABp(l/r)ap: pink, ABp(l/r)ppp: cyan, Ep: black, Cpapp: orange, ectopic: purple. Note the changes in expression driven by the -4. Although other Caenorhabditis species have large (12-20kb) intergenic regions upstream of nob-1, there is no detectable sequence conservation in the 5.3kb region upstream of nob-1, and only two short conserved stretches (85 and 275 nucleotides long) between 5.3 and 9kb upstream [35][36][37][38]. This indicates that any conserved regulatory elements in this region have diverged substantially at the sequence level during Caenorhabditis evolution, limiting utility of evolutionary conservation to identify enhancers. Chromatin marks typically used to identify transcriptional enhancers such as H3K27ac, H3K4me, or chromatin accessibility from whole embryos did not show strong peaks near nob-1 or near 20 additional genes expressed in lineage-specific patterns at the same stage as nob-1 [32,[39][40][41].

PLOS GENETICS
We hypothesized that these chromatin signals are diminished in early embryonically expressed genes because the experiments used embryos of mixed stages where later stage nuclei dramatically outnumber nuclei from earlier stages. Therefore, we searched for other factors with published genomic binding patterns that could mark embryonic enhancers. One pattern stood out as preferentially bound near genes expressed with lineage-specific patterns in the early embryo: binding of a NHR-2::GFP fusion protein [42,43]. NHR-2::GFP binds in the intergenic sequences upstream of 19 of these 20 genes, and 16 genes have multiple clustered NHR-2 binding sites (vs. 22% and 8% of randomly chosen genes, respectively, p < 0.001; chisquared test). NHR-2 is a nuclear hormone receptor distantly related to mammalian thyroid and PPAR receptors, and NHR-2::GFP is expressed in most or all somatic cells from thẽ 50-cell to~200-cell stages. The functional importance of NHR-2 binding is unclear; nhr-2 RNAi causes embryonic and larval arrest, but partial deletion alleles are viable. However, regardless of its function, we hypothesized that these clustered NHR-2 binding sites might be useful proxies for accessible chromatin and could predict enhancer activity in the early embryo.
We tested four NHR-2::GFP-bound regions upstream of nob-1 for enhancer activity by generating transgenic worms expressing histone-mCherry under the control of each candidate enhancer placed upstream of a pes-10 minimal promoter, which drives no consistent embryonic expression on its own (Fig 2A and 2B). We also tested four additional sequences for which we observed no NHR-2::GFP binding but which contained putative binding motifs for POP-1/TCF, which is required for expression of the nob-1 transcriptional reporter [24]. We identified embryonic cells expressing each enhancer reporter by confocal time-lapse imaging and StarryNite. Only three of the regions (located at -3.4 kb, -4.3 kb and -8.3 kb) tested showed enhancer activity that was consistent between embryos and also overlapped with the endogenous NOB-1::GFP expression pattern, suggesting they represent functional enhancers (Figs 2C and S2). All three functional enhancers were identified on the basis of NHR-2 binding.
The enhancer located at -3.4 kb (hereafter referred to as the -3.4kb enhancer) recapitulates most of the nob-1 early embryonic expression pattern, with some differences in the C lineage, and drives much weaker expression in ABp(l/r)ppp than in ABp(l/r)app, similar to the nob-1-5.3 kb upstream transcriptional reporter. The enhancer located at -4.3 kb also drives expression in ABp(l/r)app and ABp(l/r)ppp at a very high level ( Fig 2D) as well as variable misexpression in cells that do not normally express NOB-1 (Figs 2C and S2A). When this enhancer was cloned downstream of the reporter, it produced less ectopic expression and stronger expression in ABp(l/r)app relative to ABp(l/r)ppp, similar to the -3.4kb enhancer. This confirms the -4.3 kb region contains a bona fide enhancer capable of acting at a distance and suggests that the placement of this enhancer relative to the promoter influences its activity differently in different lineages and may be important for specificity. Embryos carrying multiple copies of the -4.3 kb reporter transgene occasionally displayed the "no backend" phenotype observed in nob-1/php-3 mutant embryos (S2B Fig) [5]. A third nob-1 enhancer (located at -8.3 kb) drives early embryonic expression only in ABp(l/r)ppp, as well as later expression in ABp(l/r)papppp.
This most-distal enhancer is included in -9 kb NOB-1::GFP transgene but not the shorter -5.3kb transcriptional reporter, and includes the only substantial stretch of conserved sequence in the nob-1 promoter region. This region thus can explain why ABp(l/r)ppp lineage expression is relatively stronger in the translational reporter compared to the shorter transcriptional reporter.
As the -5.3kb nob-1 transcriptional reporter includes both the -3.4kb and -4.3kb enhancers, we deleted each enhancer from this construct to test their necessity for nob-1 expression ( Fig  2C). Deleting both enhancers led to a complete loss of early embryonic expression in both AB lineages, suggesting there are no additional enhancers present sufficient for early embryonic expression in these lineages. The deletion of the -3.4kb enhancer did not disrupt reporter fluorescence in the ABp(l/r)app and ABp(l/r)ppp lineages, and in the absence of the other enhancer located at -4.3 kb, this enhancer was not sufficient to drive reporter fluorescence in these lineages. This indicates that while the -3.4kb enhancer may be sufficient to drive expression when placed directly next to the promoter, it is unable to drive expression in the endogenous context. Conversely, the -4.3kb enhancer is a key driver of expression in these lineages. The reporter fluorescence that remains when both enhancers are absent indicates that additional sequences that drive the expression in the cells of the Ep and Cpapp lineages must exist within the -5.3 kb upstream region.

ceh-13 and Hox cofactor genes activate nob-1 expression
Because CEH-13::GFP expression precedes that of NOB-1::GFP in many lineages, we asked whether ceh-13 is required for nob-1 expression. To evaluate this, we examined nob-1 reporter fluorescence in embryos homozygous for the likely null mutation ceh-13(sw1) [22]. Fluorescence driven by the -5.3kb nob-1 transcriptional reporter is absent in the ABp(l/r)ppp lineage (p < 0.001), 25% lower in the Abp(l/r)ap lineage, and 40% lower in the Ep lineage (p < 0.02) ( Fig  3A-3C). Depletion of ceh-13 by RNAi gave similar results with the transcriptional reporter (p < 0.002). To test the relevance of this to the endogenous locus, we measured fluorescence levels of endogenously tagged NOB-1::GFP knock-in allele in ceh-13(sw1) mutant embryos. This fluorescence was reduced in the Ep, Abp(l/r)appa, and Abp(l/r)ppp lineages by 59%, 49%, and 23% respectively (P < 0.001) (Fig 3A, 3D, and 3E). The initial onset of expression of NOB-1::GFP was delayed relative to control embryos in many of these cells (Figs 3E and S3B). Expression of endogenously tagged PHP-3::GFP in ceh-13(sw1) mutant embryos showed similar changes to endogenously tagged NOB-1::GFP (S3C Fig), indicating both genes are regulated by ceh-13, possibly through the same enhancers. In addition, the fact that the nob-1 promoter reporter completely loses ABp(l/r)pp lineage expression in the absence of ceh-13 while the endogenously tagged alleles show only temporal delay and quantitative reduction indicates the importance of additional cisregulatory sequences outside of the promoter region for nob-1 and php-3 expression.
To determine which enhancers require ceh-13, we measured the activity of individual nob-1 enhancer reporters in worms lacking ceh-13 (Figs 3A and S3A). In ceh-13(sw1) mutants, reporter fluorescence driven by the -3.4 kb enhancer decreased significantly in all expressing cells, excluding the C lineage (p < 0.003) with an average reduction of 90% and complete absence of reporter fluorescence in ABp(l/r)ppp. ceh-13 loss also reduces reporter fluorescence driven by the -4.3 kb enhancer in ABp(l/r)ppp (47% decrease) but fluorescence in Abp(l/r)ap is unchanged. This indicates that other factors besides ceh-13 activate the -4.3kb enhancer in the Abp(l/r)ap lineage and suggests that this enhancer may be responsible for the residual expression of the 5.3 kb transcriptional reporter in this lineage in ceh-13 mutant embryos. Fluorescence from the transgene driven by the distal -8.3kb enhancer was also absent in ABp(l/r)ppp in the ceh-13(sw1) mutant (p < 0.002), indicating its activity also requires ceh-13. Motif analysis identified six putative ceh-13-binding motifs in the -3.4 kb enhancer; mutating these sites  -1 and php-3 depends on ceh-13. A) Fold change in expression level relative to mean wild-type control for nob-1 promoter, endogenously tagged NOB-1::GFP knock-in allele, and enhancer reporters in ceh-13 mutant and RNAi conditions. Number of biological replicates for each condition ranges from 3-16, however, since there are, for example, 36 nob-1 cells in the ABp(l/r)ap lineages, the number of points represented by the box plots is much greater. P values determined by Wilcoxon Ranked Sum Test, p<0.05 significant. B) Lineage view of wild-type nob-1-5.3kb upstream reporter ("promoter") expression in specified lineages. C) -5.3kb nob-1 resulted in a lack of reporter fluorescence in the AB and C lineages (p < 0.02, p < 0.03), but did not affect reporter fluorescence in the Ep lineage indicating ceh-13 regulation of nob-1 may be indirect in the E lineage (Figs 3A and S3B). These results show that multiple ceh-13-dependent enhancers work together to regulate nob-1/php-3 expression in gastrulating embryos.
Expression of classically defined Hox targets often requires Hox cofactors such as those encoded by homothorax or extradenticle. These cofactors form larger TF complexes with Hox factors to increase binding specificity and may also have Hox-independent functions [44][45][46]. To determine when and where Hox cofactors are expressed in early embryos, we used Starry-Nite to trace the expression of translational reporters for unc-62, the C. elegans homothorax/ Meis ortholog, and of ceh-20 and ceh-40, the orthologs of extradenticle/Pbx. A third extradenticle ortholog, ceh-60, is only expressed in later development (>200 minutes) and does not overlap with ceh-13, so was not investigated further [30]. We found that fosmid translational reporter transgenes for each of these genes show specific and dynamic expression patterns that overlap with each other and with ceh-13 and nob-1 expression (Figs 4A-4D and S5). Notably, all three cofactors are co-expressed with CEH-13 in AB-derived cells that will later express NOB-1. Other CEH-13 and NOB-1 expressing cells also express all three cofactors, except only CEH-20 is expressed in the early E lineage, and all three are lost in the NOB-1-expressing ABp (l/r)appp lineage as the embryo approaches morphogenesis. We conclude that Hox cofactors are expressed in cells where ceh-13 activates nob-1 expression.
We tested whether Hox cofactors regulate nob-1 expression by examining nob-1 reporter expression after loss of each gene. We found that depleting unc-62 by RNAi led to significantly less expression of the nob-1-5.3kb reporter in all AB lineages (80% reduction, p = 0.001), with near complete absence of reporter fluorescence in the ABp(l/r)apap and ABp(l/r)ppp lineages ( Fig 4E). The -9kb NOB-1:GFP protein reporter transgene showed similarly limited expression, although this was only significant in the ABp(l/r)appa lineage and variable in other lineages (S4B and S4C Fig). In contrast, CEH-13::GFP reporter expression was unchanged after unc-62 RNAi (S4D Fig). These results show that nob-1 is regulated by unc-62. The fact that unc-62 RNAi has a stronger phenotype in the ABp(l/r)apap lineage than loss of ceh-13 suggests that unc-62 may regulate nob-1 independently of ceh-13 in these cells.
To investigate the role of the extradenticle orthologs, we examined the nob-1 transcriptional reporter in a ceh-40(gk159) null mutant background in embryos of worms treated with control or ceh-20 RNAi. We found that the combination of ceh-20 RNAi with the ceh-40 mutation resulted in an absence of reporter fluorescence from virtually all expressing cells including the Ep and Cpap lineages as compared to control RNAi (97% decrease, p < 0.004). This indicates that the exd orthologs ceh-20 and ceh-40 are required to activate regulatory elements upstream of nob-1 and that they are at least partially independent of ceh-13 in the cells of the ABp(l/r)ap, Ep, and Cpap lineages. This finding is consistent with previous reports that nob-1 RNAi causes increased lethality in a ceh-40 mutant background relative to wild-type animals [47].

PLOS GENETICS
Only two of the three enhancers (-3.4kb and -4.3kb) drive expression in the ABp(l/r)ap lineage, so to identify additional potential regulators of nob-1 in this lineage, we used two criteria. First, we searched for TFs expressed specifically in ABp(l/r)ap, but not ABp(l/r)pp. Second, we used a TF binding site scanning approach to identify which of these TFs have predicted binding motifs in the -3.4kb and -4.3kb enhancers [9,49]. From this analysis, the GATA zinc finger TF encoded by elt-1 emerged as the strongest candidate. An ELT-1::GFP reporter is observed prior to gastrulation in the ABp(l/r)ap progenitors ABpla and ABpra, in the other major epidermis-producing lineages ABarp, Caa and Cpa, and in the primarily neuronal lineages ABalap and ABalppp (Fig 5A). ELT-1::GFP fluorescence is later downregulated in cells that do not adopt epidermal fates (Fig 5C). ELT-1::GFP fluorescence in the ABpla/ABpra lineages could be detected at least 20 minutes prior to nob-1 reporter expression in ABp(l/r)ap (Fig 5B and  5C). GATA motifs predicted to bind ELT-1 are enriched in the enhancers that are active in ABp(l/r)ap, -3.4kb (8 sites) and -4.3kb (7 sites), compared to the -8.3kb enhancer (1 site), which does not drive expression in ABp(l/r)ap. While the C. elegans genome encodes several other GATA factors, only elt-1 is expressed prior to nob-1 in the ABp(l/r)a lineage [30,50].
We previously showed that two homeodomain bicoid family TF genes, ceh-36/OTX and unc-30/PITX, are redundantly required for proper development of the ABp(l/r)p lineage [51]. Since ceh-13 and elt-1 mutants still have some remaining NOB-1::GFP expression in the ABp (l/r)p descendants, we hypothesized that ceh-36 and unc-30 might be important for nob-1 expression in these lineages. All three enhancers contain at least two putative CEH-36/UNC-30 binding sites, and the -8.3 kb enhancer, which activates reporter fluorescence only in ABp (l/r)p, has nine (Tables C-E in S1 Table). To test this hypothesis, we examined expression of the NOB-1::GFP transgene (which includes all three enhancers) in eight ceh-36;unc-30 double mutant embryos, and found that two of the eight embryos (25%) lacked reporter fluorescence in this lineage (S7A Fig). Furthermore, the ratio of ABp(l/r)ap to ABp(l/r)ppp reporter and grey bars mark cell divisions. E,F) Fold change values for -5.3 kb nob-1 promoter expression in untreated (n = 5) and unc-62 RNAi treated (n = 7) (E) or ceh-40(gk159) mutant embryos treated with ges-1 (control, n = 5) or ceh-20 RNAi (n = 5) (F) in specified lineages. Significant changes (p<0.05) marked by asterisk; P values determined by Wilcoxon Ranked Sum Test. https://doi.org/10.1371/journal.pgen.1010187.g004

PLOS GENETICS
Regulation and function of Hox genes in the C. elegans embryo

ceh-13 and nob-1 are required for correct cell position and division patterns in the early embryo
Previous work suggested that nob-1/php-3 and ceh-13 are required for cell position but not for other aspects of cell fate specification [5]. However, mutating lineage identity regulators often leads to defects in cell cycle length or division patterns [51,52] that could have been missed in previous studies. To test for such defects, we measured cell positions and cell division timing in six embryos carrying the nob-1(ct223) mutation, which also deletes php-3 (referred to as nob-1 mutant for simplicity), and ceh-13(sw1) mutant embryos by time-lapse microscopy and StarryNite cell tracking and compared them to a database of 17 wild-type embryos (Fig 6) [ 28,46,51].
Consistent with previous work [5], we observed striking global patterns of cell position defects in nob-1 mutant embryos. The most severe and highest penetrance defects were in cells from NOB-1-expressing lineages (Figs 6A, 6C, 6D, 6G, and S8). In total, 48% (62 of 127) of cells descended from NOB-1::GFP expressing lineages had severe position defects (>5 micron deviation, z score > 3.5) in at least two embryos compared to 1.2% (14 of 1094) of cells from non-expressing lineages (p <<10 −20 ). However, a large number of cells from non-expressing lineages were moderately mis-positioned, suggesting that many defects are not cell-autonomous. Globally, dorsal cells that normally express NOB-1 fail to migrate to the posterior at 230 minutes post-fertilization; by 300 minutes non-expressing ventral cells compensate by inappropriately moving to the posterior, resulting in a global counterclockwise rotation of cell positions when the embryo is viewed from the left aspect (Fig 6G). The position defects we observed are consistent with the severe posterior morphology defects observed in rare surviving larvae.
We also identified broad cell position defects in ceh-13 mutant embryos (Fig 6B, 6E, 6F, 6H, and 6I). Intriguingly, the global patterns of cell position defects differed from nob-1 mutants. The number of severe position defects was lower for ceh-13 than for nob-1, with 4% (15/357) of cells descended from CEH-13::GFP expressing lineages severely displaced in at least two embryos, as opposed to 0.5% (5/874) of cells from non-expressing lineages (p << 10 −20 ). At 210 minutes, (200 cell stage), the most strongly mispositioned cells were ceh-13-expressing cells that do not express nob-1 in the posterior ventral region of the embryo. At this stage, cells on the dorsal surface were posteriorly mispositioned, such that cells in the posterior half of the embryo were rotationally mispositioned clockwise, the opposite direction to that seen in nob-1 mutants (Fig 6H). Cells at the anterior ventral surface (both ceh-13-expressing and nonexpressing) were mispositioned towards the posterior, resulting in a counterclockwise rotational defect in the anterior half of the embryo. At later stages, these global defects were largely resolved, and after the onset of morphogenesis the position defects were largely limited to a small number of strongly mispositioned cells (Fig 6I). The most extreme were the DA and SAB motor neurons after 300 minutes (bean stage), which were mispositioned several cell diameters anterior of their wild-type position (S8B Fig). Given the recovery of most cell positions by the beginning of morphogenesis, it is unclear if cell position defects contribute to the severe anterior morphogenesis defects in ceh-13 mutants that hatch [22], although the earlier cell position defects could disrupt normal cell-cell signaling interactions required for fate specification. In contrast to previous studies [5,22], we also identified numerous defects in cell division timing or pattern in both nob-1 and ceh-13 mutants, and these were also heavily restricted to expressing lineages (Fig 6A and 6B). 13% of nob-1 expressing cells and 5% of ceh-13 expressing cells had severe division-time defects in at least two of the corresponding mutant embryos (>5 minute deviation from wild-type, and z-score >5), vs. none in non-expressing cells (p << 10 −20 ). The most common cell division defect in nob-1 mutants was the PLM/ALN neuroblast mother (ABp(l/r)apappp), which failed to divide in 8 of 12 observations (Fig 6J). The neuroblast that normally produces the PHA, PVC, and LUA neurons instead underwent programmed cell death in two of 12 mutant lineages (Fig 6J). Other cells that had multiple cell cycle delays in nob-1 mutants included the PVQ neuroblast, PHB/HSN neuroblast, and the mother of the epidermal cells PHsh, hyp8 and hyp9 (Fig 6J and Table J in S1 Table). All of these cells also were mispositioned (mean 5 micron deviation from expected position, S8C  Fig). The most striking defects in ceh-13 mutants were in the cells ABarppaaa(a/p), which normally differentiate into the ADE sheath (glial) cell and H2 epidermal cell, but instead each divided inappropriately in 4 of 6 mutant embryos (Fig 6K). In addition, ABplppaaap, which normally undergoes programmed cell death, instead divided in two ceh-13 mutant embryos ( Table I in S1 Table). Several neuroblasts including those producing the DA, DD and SAB motorneurons also had cell division delays in ceh-13 mutants ( Table I in S1 Table). We confirmed that the most severe cell division defects have similar penetrance in additional alleles of ceh-13 and nob-1/php-3 (S8D Fig), indicating that these defects are due to loss of these Hox genes rather than other mutations in the genetic background.
We conclude that both ceh-13 and nob-1/php-3 are required not only for proper cell positioning, but also for normal division patterns, suggesting a broader role in fate specification, although it remains unclear whether this is controlled directly by the regulation of key fate specifiers. Alternatively, the control of cell identity could be indirect if cell positions are critical for promoting proper local signaling interactions to specify cell fates. Importantly, the changes in division patterns are not consistent with broad homeotic changes in lineage identity. Furthermore, the incomplete penetrance of both cell cycle and cell position defects in ceh-13 and nob-1 mutants indicate redundancy in developmental programming that likely contributes to the robustness of C. elegans embryonic development [33].

Discussion
Classic work in many species defined the concept of Posterior Dominance, in which posterior Hox genes repress the expression and activity of more anterior Hox genes [1]. Indeed, this 1::GFP and PHP-3::GFP expression overlap with plots of frequency of cell cycle defects and cell position defects for each cell plotted below. Cells expressing both NOB-1::GFP and PHP-3::GFP are highlighted in cyan on the plots. C, D) Plot of mean deviation in microns vs. mean neighbor score for each cell in the embryo, colored by level of NOB-1::GFP expression, for nob-1(ct223) mutant which disrupts both nob-1 and php-3 (C) and wild-type (D) embryos. Neighbor score is a measure of whether a cell is close to its normal neighbors; values above 0. 8

PLOS GENETICS
principle is evident in later stages of C. elegans development, where previous work showed that nob-1 represses ceh-13 in the posterior ventral nerve cord, consistent with posterior dominance [46]. In contrast, we show that ceh-13 activates nob-1/php-3 expression through at least two enhancers during C. elegans gastrulation. To our knowledge, this is the first example of an anterior Hox gene positively regulating the expression of a posterior Hox gene during development. ceh-13 and nob-1/php-3 are also expressed at an earlier phase in development (pre-gastrulation for ceh-13 and mid-gastrulation for nob-1/php-3) than in many other organisms. These observations, along with the lineage defects seen in ceh-13 and nob-1 mutants, suggest that the role of ceh-13 and nob-1 is to regulate lineage identity, distinct from their later role in positional identity. The role of Hox cofactor genes such as unc-62, ceh-20 and ceh-40 in early embryonic nob-1 regulation suggests that these novel functions in lineage specification nonetheless share some common mechanisms with the later roles. The fact that C. elegans is largely unsegmented may have allowed the co-option of these Hox genes for this distinct role. This emphasizes the striking amount of evolutionary flexibility in the function of the genes of the C. elegans Hox cluster.
Several studies have highlighted that conserved transcription factors often "moonlight," with the same factor playing apparently distinct roles in early progenitors and in terminal cell types. For example, the OTX homeodomain TF-encoding gene ceh-36 and the PITX homeodomain TF-encoding gene unc-30, which each specify distinct terminal neuron types, also act redundantly 6-7 cell divisions earlier in the ABp(l/r)p lineages to specify broad features of lineage identity [51,[53][54][55], and skn-1, which is required maternally to specify the endomesodermal blastomere EMS at the 4-cell stage is required post-embryonically for oxidative stress resistance and longevity [56,57]. Our work shows that ceh-13 and nob-1/php-3 are another example of this phenomenon. Our identification of cell division defects in each mutant demonstrates they also play a role in lineage specification. In addition, we confirmed that each is also required for correct cell position; it is important to note that position defects could also result from incorrect specification of lineage identity or that mispositioning of cells or their progenitors could disrupt additional signals that specify cell fate. Previous work showed that at least one neuron (the PLM touch receptor) is defective in less severe nob-1 mutants [20]; the failure of PLM neuroblasts to divide in nob-1/php-3 mutants indicates a function in PLM progenitor lineages. Similarly, ceh-13 mutants have cell division defects, and ceh-13 is required for the lineage-specific activity of specific nob-1 enhancers. Finally, while elt-1 has a well-defined known role in epidermal fate specification [58], our work shows that it is also required for nob-1 expression, including in cells that adopt non-epidermal fates. This is consistent with other recent work identifying defects in cell division patterns and the expression of the neurogenic TF-encoding gene lin-32 in elt-1 mutants [59]. The ability of developmental transcription factors to "moonlight' is likely facilitated by complex regulation by multiple enhancers as seen for nob-1/php-3, and ceh-13 [31,32] since enhancer evolution can enable genes to be expressed at different locations and times in development [60,61]. An intriguing question is whether these multiple Hox functions are made possible by the unique nature of the C. elegans lineage, where positional identity is defined at least in part by lineage history [18,62], or if they are conserved in other species.
While early reporter studies in C. elegans found that the expression of many genes expressed in terminal cells is well approximated using just the promoter-proximal region [63][64][65], open chromatin mapping studies have identified many distal regions of open chromatin, many of which can act as enhancers [32,39,40]. Our work reinforces these studies and emphasizes the importance of enhancers in C. elegans gene regulation. Intriguingly, other genes expressed in lineage-specific patterns around the same time as nob-1 have similar tendency to have multiple nearby regions of NHR-2::GFP binding and lineage-specific accessible chromatin as measured by ATAC-seq [66], suggesting they may also be regulated by multiple distal enhancers. This could reflect these genes' need for more complex regulation. The nob-1/ php-3 locus contains at least two or three enhancers responsible for expression in each expression domain during gastrulation, with different enhancers more or less dependent on (or independent of) particular activating TFs. This is reminiscent of so-called "shadow enhancers" identified in other organisms, for which redundancy appears to confer robustness in the face of environmental variability [67][68][69]. This high level of redundancy also appears to extend to the number of transcription factors regulating each enhancer. Current data implicate at least eight TFs as activators of nob-1 enhancers including those encoded by the genes ceh- 13, elt-1, ceh-36, unc-30, the Wnt effector gene pop-1, and the Hox co-factor genes ceh-20, ceh-40 and unc-62 (S9 Fig). In some cases, the same binding motif may be regulated by different factors in different lineages. For example, mutating six GATA motifs in the -3.4 kb enhancer caused the loss of expression in ABp(l/r)ap, potentially due to loss of ELT-1 binding. However, this mutant also lacked reporter fluorescence in Ep, which does not express elt-1, but does express genes for other GATA factors with similar binding specificity including end-1, end-3, and elt-7 [70]. Similarly, the greater reduction in nob-1 enhancer reporter fluorescence when CEH-13 motifs are mutated than observed in ceh-13 mutant animals indicates that other TFs may also bind these sites. The partially penetrant nature of many of the mutant phenotypes further emphasizes the high level of redundancy. In sum, multiple modes of regulatory redundancy could provide a mechanism for the remarkable developmental robustness observed in C. elegans.

Molecular biology
Candidate enhancers were amplified with Phusion HF polymerase (New England Biosciences) from either pJIM20::nob-1 or N2 genomic DNA, with overhangs for stitching, which were then either gel or PCR purified (Qiagen). Putative enhancers were attached to a pes-10 minimal promoter::HIS-24::mCherry::let-858 3'UTR fragment amplified from POPTOP plasmid [72] (Addgene #34848) by using PCR stitching to create an enhancer reporter which was sequence verified and purified with a PureLink PCR purification kit (ThermoFisher) for injection. The pes-10 minimal promoter drives no consistent early embryonic expression alone, but has previously been shown to facilitate expression driven by different enhancers [72][73][74]. For the downstream enhancer experiment, the -4.3kb enhancer was cloned immediately downstream of the reporter stop codon. Tested enhancer sequences can be found in Table B in S1 Table. Regions deleted from the -5.3 kb transcriptional reporter match the tested enhancer sequences exactly; deletions were performed using the QuikChange II mutagenesis kit (Agilent). Putative transcription factor binding sites were identified using CIS-BP (cisbp. ccbr.utoronto.ca) [9,49] (Table C-E in S1 Table). DNA fragments with desired mutations were synthesized (Integrated DNA Technologies) and PCR stitched as before.

Imaging/Lineaging
We acquired confocal images with a Leica TCS SP5, Stellaris or Nikon A1RSi resonance scanning confocal microscope (67 z planes at 0.5 μm spacing and 1.5 minute time spacing, with laser power increasing by 4-fold through the embryo depth to account for attenuation of signal with depth). Embryos from self-fertilized hermaphrodites were mounted in egg buffer/methyl cellulose with 20μm beads as spacers [75] and imaged at 22˚C using a stage temperature controller (Brook Industries, Lake Villa, IL). We used StarryNite software to automatically annotate nuclei and trace lineages, and AceTree software to identify and fix any errors from the automated analysis, and quantified reporter expression in each nucleus relative to local background (using the "blot" background correction technique) as previously described [26,27,29,76,77]. In brief, the locally calculated background fluorescence is subtracted from the value measured within the defined nucleus of each cell to reduce false positives from out-offocus light from nearby expressing cells. Therefore, expressing cells will have positive fluorescence while non-expressing cells can have negative values due to slightly higher autofluorescence in the cytoplasm than in the nucleus. All Cell Data files, which include reporter levels and position information for individual cells at all curated timepoints, are deposited in the Dryad repository: http://doi.org/10.5061/dryad.zs7h44j86 [78].

Quantitative comparisons
Cell averages of nuclear fluorescence were computed for each cell based on all (typically >20) measurements across its lifetime (Table F in S1 Table). For each control condition, the mean value of each named expressing cell across control embryos was computed and used to determine the fold change for each control and mutant cell with the same name, which were displayed as boxplots using R 3.5.1 (The R Foundation for Statistical Computing). Backgroundcorrected expression values were rounded up to zero if values fell below zero. To determine if changes were significant across the lineage, the sum of expression for each cell in the lineage of interest was calculated for each control and mutant embryo and the two sets of values were compared using a Wilcoxon Ranked Sum test using R. For each condition, at least four control and three mutant embryos were analyzed (exact numbers in Table G in S1 Table). Each embryo analyzed represents an independent biological replicate. To evaluate the enrichment of NHR-2 binding sites, a chi-squared test was used.

Mutant cell position analysis
Cell position defects were identified as previously [51]. Briefly, we corrected for differences in global division rates (which did not differ dramatically from wild-type) and considered divisions as defective/outliers if they deviated from the wild-type cell cycle length by at least five minutes and had a z-score greater than three (Table H-J in S1 Table). Cell positions (S2-S4 Tables) were corrected for differences in embryo size and rotation and considered defective if they deviated from the expected wild-type position by at least five microns, had a z-score greater than five, and a nearest neighbor score greater than 0.8 (defined empirically based on the distribution of wild-type scores).

Dryad DOI
http://doi.org/10.5061/dryad.zs7h44j86 Supporting information S1 Fig. Hox gene are expressed in overlapping patterns. A) Overlap in the expression of endogenously tagged NOB-1::GFP (blue) and PHP-3::GFP (green). Expression is nearly identical (cyan overlap) except php-3 is expressed more consistently in ABplpapppp. B, C) Early (B) and late (C) overlap of a CEH-13::GFP fosmid translational reporter transgene with a -9kb NOB-1::GFP rescuing translational reporter transgene. The expression patterns are nearly identical to the endogenously tagged alleles except the 9kb NOB-1::GFP reporter has reduced expression in the C lineage (see Fig 2C). D) Expression pattern of a CEH-13::GFP transgene that contains 8.