Cis-regulatory elements co-opting core circadian clock regulator CCA1 underlie enhanced expression of HMA4 for metal hyperaccumulation in Arabidopsis halleri

Abstract

The naturally selected extreme traits of zinc and cadmium hyperaccumulation and hypertolerance in Arabidopsis halleri depend on strongly elevated HEAVY METAL ATPase 4 (HMA4) transcript levels compared to those in the closely related Arabidopsis thaliana. This difference is regulated in cis; AhHMA4 upstream sequences alone are sufficient to confer increased expression, as previously demonstrated using reporter gene fusions stably introduced into both A. halleri and A. thaliana. However, the underlying cis-regulatory divergence specific to A. halleri remains unknown. Here, we identify cis-regulatory metal hyperaccumulation elements (MHEs) that increase AhHMA4 promoter activities by examining stably transformed reporter lines carrying partial deletions or mutations in AhHMA4 upstream sequences. MHE1 (consensus TGTAAC) functions in the distal regions of AhHMA4 promoters, and all three tandem AhHMA4 gene copies share a proximal upstream pair of MHE2 motifs (consensus AAATATCT), corresponding to the evening element. The evening element is a known target of Arabidopsis CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1), a core circadian clock transcription factor that mediates light-dependent and circadian gene expression. We show that the elevated activity of the AhHMA4-1 promoter depends on MHE2 in cis and CCA1 in trans, and it can be recapitulated by introducing an intact pair of MHE2 motifs into the A. thaliana HMA4 promoter using site-directed mutagenesis. We also found that HMA4 transcript levels show diel rhythmicity in A. halleri but not in A. thaliana. In summary, we identify the causal cis-regulatory elements that co-opt a known regulator of diel and seasonal transcriptional rhythms to mediate enhanced expression of a gene critical for a naturally selected extreme trait syndrome.

Metal hyperaccumulation and hypertolerance in Arabidopsis halleri rely on elevated HMA4 transcript levels. This study reports that, compared with the A. thaliana HMA4 promoter, enhanced AhHMA4 promoter activity requires CCA1, along with the cis-regulatory metal hyperaccumulation elements MHE1 and MHE2, the latter of which resembles the CCA1-binding evening element. These findings reveal causal cis-regulatory divergence and the co-option of a circadian clock transcriptional regulator that underlie naturally selected extreme adaptations in A. halleri.

Introduction

Evolutionary novelties can result from alterations in coding regions which affect the function, abundance, or stability of a gene product, or from changes in non-protein-coding DNA sequences altering the regulation of gene expression (Hill et al., 2021). Indeed, cis-regulatory changes contribute substantially to the genetic variation underlying naturally and anthropogenically selected traits in both plants and animals (Jacob and Monod, 1961; King and Wilson, 1975; Stern, 1998; Crawford et al., 1999; Wang et al., 1999; Clark et al., 2006; Hanikenne et al., 2008; Hufford et al., 2012; Alonge et al., 2020; Liu et al., 2020; Song et al., 2020). However, studies that establish a mechanistic link between specific cis-regulatory mutations and the corresponding gene expression outcomes remain scarce (Hill et al., 2021; Schmitz et al., 2022).

The naturally selected extreme traits of heavy metal hyperaccumulation and hypertolerance are characteristic of Arabidopsis halleri. A. halleri, a diploid, stoloniferous perennial and an obligate outcrosser, is a sister species of the well-studied short-lived diploid selfing species A. thaliana, from which its lineage diverged between 5 and 10 million years ago (Krämer, 2010; Novikova et al., 2018). As the only metal-hyperaccumulating species in the Arabidopsis genus (which is in lineage I of the Brassicaceae), A. halleri can contain more than 3000 (and up to 53 900) μg Zn g⁻¹ in its dry leaf biomass and more than 100 (and up to 3640) μg Cd g⁻¹ , in its natural habitats (Verbruggen et al., 2009; Krämer, 2010; Stein et al., 2017). Thus, A. halleri can accumulate and tolerate more than 10-fold and up to around 10 000-fold higher internal metal levels in aboveground organs than typical plants. In Europe and East Asia, A. halleri is among the natural colonizers of calamine metalliferous soils, which contain high, toxic levels of Zn and Cd from geogenic or anthropogenic sources (Ernst, 2006). Species-wide Zn and Cd hypertolerance has been confirmed on synthetic media under laboratory conditions (Bert et al., 2003; Becher et al., 2004; Meyer et al., 2010). Unlike most of the >700 metal-hyperaccumulating plant species identified to date, A. halleri is a facultative metallophyte; it is also found naturally on unpolluted soils containing only background levels of heavy metals, where it also exhibits metal hyperaccumulation (Reeves et al., 2017). Modest Zn hyperaccumulation is also observed in the allotetraploid A. kamchatica, which arose through hybridization between A. halleri and A. lyrata (Paape et al., 2020).

According to cross-species transcriptomics, the transcript levels of dozens of metal-related genes are elevated in A. halleri, mostly constitutively, by comparison to the closely related non-hyperaccumulating A. thaliana (Becher et al., 2004; Weber et al., 2004; Talke et al., 2006). Subsequently, a reverse-genetic approach identified one of these candidate genes, HEAVY METAL ATPase 4 (HMA4) encoding a plasma membrane–localized Zn²⁺- and Cd²⁺-exporting pump, as a key causal locus that is required for metal hyperaccumulation and that makes the largest known contribution to metal hypertolerance (Hanikenne et al., 2008, 2013). This was supported by the genomic position of HMA4 within quantitative trait loci mapped for the traits of Zn and Cd hypertolerance in a segregating population from a cross between A. halleri and the non-hypertolerant A. lyrata (Courbot et al., 2007; Willems et al., 2007). Relative to AtHMA4, cis-regulatory functional divergence enhances promoter strength for all AhHMA4 gene copies (AhHMA4-1 to AhHMA4-3) in both the A. halleri and A. thaliana genetic backgrounds (Talke et al., 2006; Hanikenne et al., 2008, 2013). In combination with the tandem triplication of AhHMA4, these cis-regulatory mutations result in up to 100-fold higher HMA4 transcript levels in A. halleri than in A. thaliana.

Beyond enhanced HMA4 gene product dosage (Hanikenne et al., 2013), the evidence available to date does not support any predominant role for divergent transcript localization, regulatory responses, or protein functions among the AhHMA4-1 to AhHMA4-3 gene copies relative to AtHMA4 (Krämer, 2010; Nouet et al., 2015). Ectopic gene conversion, also termed interlocus gene conversion, among AhHMA4-1, AhHMA4-2, and AhHMA4-3 is well documented. Concordantly, coding sequence identities of ≥99% between HMA4 gene copies indicate that their protein-coding sequences undergo concerted evolution (Hanikenne et al., 2008, 2013). The patterns of nucleotide polymorphism are consistent with positive selection and a hard selective sweep in the genomic region containing AhHMA4-1 to AhHMA4-3 (Hanikenne et al., 2013).

Here, we identify the divergent cis-regulatory sequence elements required for enhanced activity of the promoters of the A. halleri HMA4 gene copies compared to A. thaliana HMA4. We used sequence comparisons and analyses, promoter deletion series, promoter mutations, and segmental promoter swap constructs. In the three AhHMA4 promoters, but not in the A. thaliana HMA4 promoter, we identified conserved cis-regulatory enhancers that we designated metal hyperaccumulation elements 1 and 2 (MHE1 and MHE2). MHE1 and MHE2 contributed to high reporter gene expression in the A. thaliana genetic background. In contrast, we found that A. thaliana HMA4 is subject to complex repressive cis-regulatory mechanisms conferred by a distal upstream promoter segment and a segment corresponding to an intron in the 5′ UTR of the transcript. A. halleri MHE2 sequences are identical or highly similar to the evening element (EE), a well-known binding site for the MYB family transcription factor CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1), which mediates light-dependent regulation and forms part of the core oscillator of the Arabidopsis circadian clock (Wang et al., 1997; Wang and Tobin, 1998; Alabadí et al., 2002). We found that both the elevated levels and the diel dynamics of reporter gene transcript abundance depend on the combined functions of CCA1 in trans and MHE2 in cis. Introducing both MHE2 copies from the AhHMA4-1 promoter into an AtHMA4 promoter context reproduced the strongly elevated reporter gene transcript levels and their diel rhythms. Endogenous HMA4 transcript levels in A. halleri exhibited similar diel dynamics, supporting the validity of these findings. Taken together, the results of this study illustrate how cis-regulatory divergence contributed to the evolution of a naturally selected extreme trait through the co-option of a core circadian clock transcriptional regulator, and they advance our understanding of the mechanistic basis of physiological adaptations in plants.

Results

Identification of regions governing functional divergence between the A. halleri HMA4-1 and A. thaliana HMA4 promoters

We generated a series of promoter deletion constructs fused to a GUS reporter gene, and introduced these into A. thaliana, considering that the magnitudes and localization of HMA4 promoter activities are similar in both the A. halleri and A. thaliana genetic backgrounds (Hanikenne et al., 2008). Among the promoter regions of the three HMA4 gene copies from A. halleri (accession Lan3.1)—AhHMA4-1, AhHMA4-2, and AhHMA4-3—the AhHMA4-1 promoter has the highest sequence similarity to the AtHMA4 promoter (accession Col-0), which facilitated direct sequence comparisons (Supplemental Figures 1A–1D) (Hanikenne et al., 2008, 2013). The AhHMA4 genomic region is transposon-rich, but previously annotated transposons are located outside the promoter regions examined in this study (Hanikenne et al., 2008; EU382073.1 and EU382072.1).

The full-length (FL) AhHMA4-1_P construct employed here consisted of a 2326-bp genomic fragment, comprising 1602 bp upstream of the transcriptional start site, the 5′ UTR (exon 1, an intron, and the first 3 bp of exon 2), and the initial 30 bp of the AhHMA4-1 coding sequence (Hanikenne et al., 2008). We designed a series of progressive deletions from the 5′ end, with the choice of breakpoints guided by the boundaries of segments exhibiting sequence similarity between the promoters of A. halleri HMA4-1 and A. thaliana HMA4 (Figure 1A and Supplemental Figure 1D). The AhHMA4-1_P Δ698 construct, generated by deleting the initial 904 bp at the 5′ end of AhHMA4-1_P that we refer to as the distal region (DR), conferred substantially decreased GUS transcript levels and GUS specific activity in total protein extracts, with residual GUS transcript levels of 37%, 24%, and 35% in shoots, roots, and whole seedlings, respectively (Figures 1B and 1C; note the log-scaled vertical axes, and compare FL AhHMA4-1_P with Δ698; Supplemental Figures 1E and 1F). We therefore named this region “enhancing region 1” (ER1) (Figure 1A). Further progressive deletions in the downstream 568-bp intermediate region (IR) had no or only minor effects (Figures 1A–1C, IR; constructs are named Δ438 to Δ130). Deletion of the intron had no significant effects in the FL AhHMA4-1_P construct and in a promoter construct lacking the DR and IR (Figures 1B and 1C; compare FL with FLΔi and Δ130 with Δ130Δi; Supplemental Figures 1E and 1F). This suggests that the intron does not influence AhHMA4-1 promoter activity. In a construct lacking the DR, IR, and intron, the additional deletion of the 44-bp proximal region (PR) to generate the AhHMA4-1_P Δ86Δi construct led to a substantial decrease in GUS transcript levels, resulting in 3%, 1%, and 2% of the FL promoter levels in shoots, roots, and seedlings, respectively (Figure 1B, Δ86Δi; Supplemental Figure 1E). In these lines, residual GUS transcript levels and GUS specific activities were very low and indistinguishable from those of control transformants lacking any promoter upstream of the GUS reporter gene (Figures 1B and 1C, Δ86Δi and EV; Supplemental Figures 1E and 1F). We therefore named this 44-bp segment “enhancing region 2” (ER2) (Figure 1A).

Figure 1.

Analysis of deletion series of AtHMA4 and AhHMA4-1 promoter regions in transgenic A. thaliana reporter lines.

(A)GUS reporter constructs for AtHMA4_P and AhHMA4-1_P. Numbers indicate distances in bp from the transcriptional start site (+1) for each breakpoint in the promoter deletion series of reporter constructs introduced into A. thaliana. Positions of putative CAAT and TATA boxes are marked by magenta and cyan vertical lines, respectively. 5′ UTR, 5′ untranslated region; ∗, translational start site (vertically aligned between promoters).

(B–E) Relative GUS transcript levels (B and D) and GUS specific activities (C and E) for full-length AhHMA4-1_P(B and C), AtHMA4_P(D and E), and their respective deletion series in transgenic A. thaliana GUS reporter lines. Bars show mean ± SD (n = 3–14) of independent transgenic lines on log-scaled axes, with each data point representing the mean of three replicate multi-well plates of qPCR reactions (B and D) or enzyme assays (C and E). Data are from one representative experiment out of two independent experiments. Different letters indicate statistically significant differences based on one-way non-parametric ANOVA with Dunn’s multiple comparison test (p < 0.05). FL, full length; Δi, deletion of the intron (see A); EV, transformants with a construct lacking any promoter fragment upstream of the GUS gene (see methods); MU, 4-methylumbelliferone.

The 2799-bp FL promoter fragment of A. thaliana HMA4 includes 2000 bp upstream of the transcriptional start site (Hanikenne et al., 2008). GUS transcript levels driven by AtHMA4_P were 8% in shoots, 54% in roots, and 14% in seedlings relative to those driven by a single FL A. halleri HMA4 promoter (Figures 1B and 1D; see also Figure 2P). In the same AtHMA4_P reporter lines, GUS enzyme activities were far below those observed in the AhHMA4_P lines, at less than 0.5% (Figures 1C and 1E; see also Figure 2Q), consistent with previously published results (Hanikenne et al., 2008). Linear regression analyses confirmed that, relative to GUS transcript levels, GUS specific activities in FL promoter–reporter lines were considerably lower for AtHMA4_P than for AhHMA4-1_P (Supplemental Figures 2A–2D). As a function of relative GUS transcript levels, GUS specific activities in AhHMA4-1_P lines increased, exhibiting slopes of 65 in shoots (Supplemental Figure 2A) and 8.2 in roots (Supplemental Figure 2C). In contrast, AtHMA4_P FL lines exhibited slopes of 0.07 in shoots (Supplemental Figure 2B) and 0.09 in roots (Supplemental Figure 2D). Thus, GUS enzyme activity was in qualitative agreement with GUS transcript levels for AhHMA4-1_P, whereas it was much lower than the transcript levels for AtHMA4_P.

Figure 2.

Identification of two homologous enhancer regions in the promoters of all three A. halleri HMA4 gene copies.

(A)GUS reporter constructs for AtHMA4, AhHMA4-1, AhHMA4-2, and AhHMA4-3 promoters. See also Figure 1A.

(B–O) Histochemical detection of GUS activity in rosettes (left) and the root hair zone of roots (right) of representative transgenic A. thaliana GUS reporter lines. Scale bars: 2 mm for rosettes and 0.2 mm for roots.

(P and Q) Relative GUS transcript levels and GUS specific activities in transgenic A. thaliana reporter lines. Bars show mean ± SD (n = 3–7) of independent transgenic lines on log-scaled axes, with each data point representing the mean of three technical replicates (see also Figure 1). Ah-1, AhHMA4-1; Ah-2, AhHMA4-2; Ah-3, AhHMA4-3; ΔDIR, deletion of both distal and intermediate regions; ΔDIPR, deletion of the distal, intermediate, and proximal regions (see also Figures 1B–1E).

Compared with the FL AtHMA4_P construct, the combined deletion of the DR and the intron caused an approximately 100-fold increase in GUS enzyme activities in total protein extracts from both shoots and roots (Figure 1E; compare Δ597Δi and FL). In these AtHMA4_P Δ597Δi lines, GUS activity expressed as a function of GUS transcript levels yielded regression lines with slopes similar to those observed in FL AhHMA4-1_P lines (Figures 1D and 1E and Supplemental Figures 2B and 2D; AtHMA4_P ΔDRΔi exhibited slopes of 17 in roots and 84 in shoots). Deletion of either the DR or the intron alone had no effect on GUS activity. This suggests that the DR and the intron of AtHMA4_P independently and strongly repress GUS reporter activity at the protein level.

At the GUS transcript level, deletion of only the DR of AtHMA4_P resulted in a seven-fold increase in shoots and no significant change in roots (Figure 1D; compare FL and Δ597). Moreover, deletion of only the intron of FL AtHMA4_P caused an 11-fold increase in GUS transcript levels in shoots and a two-fold increase in roots (Figures 1A and 1D; compare FL and FLΔi). Consequently, the DR and intron of AtHMA4_P each appear to repress GUS transcript levels, primarily in shoots. However, these effects appear to be complex, as simultaneous deletion of the DR and the intron resulted in GUS transcript levels similar to those of the FL AtHMA4_P construct. Furthermore, our results suggest that the region between 234 and 129 bp upstream of the transcriptional start site of AtHMA4_P has a two- to three-fold enhancing effect on GUS transcript levels in the presence of the intron (Figure 1D; compare Δ234 and Δ129), consistent with similar changes in GUS activity (Figure 1E). In the absence of the intron, we detected a similar effect only in the roots. In contrast to AhHMA4-1_P, deletion of the PR in AtHMA4_P did not cause any change in GUS transcript levels or GUS activity (Figures 1D and 1E; compare Δ129Δi and Δ88Δi).

Taken together, our results are consistent with the presence of cis-regulatory elements that enhance transcription, located in the DR (ER1, between −1602 and −698) and near the transcriptional start site in the PR (ER2, between −130 and −86) of the A. halleri HMA4-1 promoter. For AtHMA4, analysis of our reporter constructs suggests the presence of complex repressive functions involving the DR and the intron in the 5′ UTR. The corresponding sequence segments had much stronger effects on reporter protein activity than on transcript abundance. In brief, our results implicate two transcription-enhancing regions in the A. halleri HMA4-1 promoter and two repressive regions upstream of AtHMA4 which act in a complex manner.

All three AhHMA4 gene copies share two homologous cis-regulatory enhancers

Because the promoters of all three paralogous HMA4 gene copies of A. halleri are known to produce high levels of reporter activity (Hanikenne et al., 2008), we hypothesized that ER1 and ER2 of AhHMA4-1_P are functionally conserved in both AhHMA4-2_P and AhHMA4-3_P (Figure 2A). The 2060-bp FL AhHMA4-3_P construct included 1494 bp upstream of the transcriptional start site (Nouet et al., 2015). Deletion of the DR (–1494 to −310), delineated based on its sequence similarity with AhHMA4-1, caused reductions in both GUS transcript levels and GUS specific activity to 30% in shoots and seedlings and 25% in roots, very similar to our observations for AhHMA4-1_P (Figures 2B, 2D, 2E, 2G, 2P, and 2Q; Supplemental Figures 1E and 1F; compare FL and ΔDR for AhHMA4-1_P and AhHMA4-3_P). For AhHMA4-2_P (2293 bp long), which included 1761 bp upstream of the transcriptional start site, deletion of the DR (−1761 to −515) resulted in only 10% promoter activity according to both GUS transcript levels and GUS specific activity, regardless of tissue type (Figures 2C, 2F, 2P, and 2Q; Supplemental Figures 1E and 1F; compare FL with ΔDR for Ah-2_P). Similar to our findings for AhHMA4-1_P, the additional deletion of PR in constructs lacking the DR, IR, and intron caused a severe decrease in activity for both AhHMA4-2_P and -3_P (Figures 2H–2Q; Supplemental Figures 1E and 1F; compare ΔDIRΔi with ΔDIPRΔi for AhHMA4-1_P, -2_P, and -3_P). This suggests that both ER1 and ER2 are functionally conserved in the promoters of all three A. halleri HMA4 gene copies. In addition, a small difference was observed between the promoter of AhHMA4-2 and those of the other two A. halleri HMA4 gene copies. In AhHMA4-2_P lacking the DR, deletion of both the IR and the intron led to a modest increase in GUS transcript levels—up to three-, four-, and two-fold higher in seedlings, shoots, and roots, respectively—which was also partially reflected in GUS activity (Figures 2F, 2I, 2P, and 2Q; Supplemental Figures 1E and 1F; compare ΔDR with ΔDIRΔi for Ah-2_P). Thus, either the IR of the promoter or the intron in the 5′ UTR of AhHMA4-2 may uniquely harbor a moderately repressive element, which was not analyzed further.

Identification of conserved cis-regulatory elements among the A. halleri HMA4 promoters

Further dissection of ER1 in the A. halleri HMA4 promoters using an additional series of AhHMA4-1_P deletion constructs revealed a predominant contribution from the segment between positions −909 and −806, within the DR of AhHMA4-1_P (Figures 3A and 3B; Supplemental Figures 3 and 4A). Deletion of this segment resulted in decreases in both GUS transcript levels and GUS specific activity to residual levels of about 30%, 37%, and 25% in shoots, seedlings, and roots, respectively (Figures 3D–3G, 3P, and 3Q; Supplemental Figures 4C–4J). Within this segment, we identified an approximately 36-bp region with a high degree of sequence similarity among the A. halleri HMA4 promoters and a shared divergence from AtHMA4_P (−895 to −860 in AhHMA4-1_P, designated ER1^+; Figure 3B; Supplemental Figures 1D, 3, and 4B). Multiple sequence alignments and motif elicitation analyses of ER1⁺ across the promoters of AhHMA4 gene copies from diverse A. halleri populations identified a conserved 12-bp putative cis-regulatory element (−895 CTTTGTAACCAT −884 in AhHMA4-1_P) (Figure 3B; Supplemental Figure 5A; Supplemental Dataset 1). This element contains the 7-bp core motif TGTAACC (−892 to −886), designated MHE1, which is absent in A. thaliana (Supplemental Figure 4B; see methods). We designated this instance MHE1a because a second copy of the MHE1 core motif, designated MHE1b, appears to be present upstream (positions −914 to −908) in AhHMA4-1_P (Figure 3B; Supplemental Figures 3 and 4B) but not in AhHMA4-2_P or AhHMA4-3_P. In the Δ909 construct, disruption of MHE1b in AhHMA4-1_P resulted in a subtle but statistically significant decrease in promoter activity at the GUS transcript level in roots only (Supplemental Figure 4I; compare Δ1362 and Δ909); however, we did not detect any effect on GUS activity or in whole seedlings (Figures 3P and 3Q; Supplemental Figure 4J). Therefore, MHE1b of AhHMA4-1_P exerts a minor effect, if any. In AtHMA4_P, the sequence (−684)-TGTAATC-(−678) is highly similar to MHE1b of AhHMA4-1, with a single C-to-T substitution at the penultimate position, whereas MHE1a is absent in AtHMA4_P and AhHMA4-2_P (Figure 3B; Supplemental Figures 3, 4B, and 5A).

Figure 3.

Cis-regulatory enhancers MHE1 and MHE2 are required for AhHMA4-1 promoter activity.

(A–C) Schematic of AhHMA4-1_P constructs (A), alignments of the ER1⁺(B) and ER2 (C) regions of AhHMA4-1_P with homologous regions of AtHMA4_P (left), and the sequence logo of the core MHE2 motif shared by the promoters of all three AhHMA4 gene copies (right; see also Supplemental Figure 6). The sequences of putative cis-regulatory elements in ER1⁺ (the effective sub-region of ER1; red cross symbol in A, red bar in B) are boxed: metal hyperaccumulation element 1a (MHE1a) and a second identical upstream copy (MHE1b, B), as well as MHE2a and MHE2b in ER2 (light red bar, C). Between-species sequence differences are shown on a red background and also indicate the mutations introduced into AhHMA4-1_P (conversion to the corresponding AtHMA4_P sequence) to disrupt MHE1a/b and/or MHE2a/b (B and C). See also Figure 1A.

(D–O) Histochemical detection of GUS activity in rosettes (left) and the root hair zone of roots (right) of representative transgenic A. thaliana GUS reporter lines. Scale bars: 1 mm (rosettes) and 0.2 mm (roots).

(P and Q) Relative GUS transcript levels (P) and GUS specific activities (Q) in whole seedlings of transgenic A. thaliana GUS reporter lines. Bars show mean ± SD (n = 3–15) of independent transgenic lines on log-scaled axes. Each data point represents the mean of three technical replicates. Different letters indicate statistically significant differences based on one-way non-parametric ANOVA followed by Dunn’s multiple comparison test (p < 0.05). ΔER2, deletion of ER2 (see C) in a full-length promoter; mut, mutated (nucleotide substitutions introducing the AtHMA4_P sequence into the respective pair of MHE motifs; shown on a red background in B and C); MHE1/2_mut, mutation of both MHE1 and MHE2 pairs. See also Figure 1.

Figure 4.

Introduction of two cis-regulatory enhancer MHE2 elements is sufficient to increase AtHMA4 promoter activity.

(A)AtHMA4_P constructs used for the introduction of the cis-regulatory elements MHE1 and MHE2 from AhHMA4-1_P. A shortened AtHMA4_P construct (Δ754Δi) that lacks both the intron and 1246 bp at the 5′ end (corresponding to part of the DR) was used to introduce MHE1 or MHE2 pairs at positions equivalent to those in AhHMA4-1_P (marked by flag symbols) via site-directed mutagenesis. See also Figures 3B and 3C. lncRNA denotes a long noncoding RNA annotated in the DR of AtHMA4_P (AT2G07115; see discussion).

(B–K) Histochemical detection of GUS activity in rosettes (left) and the root hair zone of roots (right) of transgenic A. thaliana GUS reporter lines. The image for EV is shown in Figure 3O. Scale bars: 1 mm (E, F, and K), 0.5 mm (all other rosette images), and 0.2 mm (root images).

(L and M) Relative GUS transcript levels (L) and GUS specific activities (M) in seedlings of transgenic A. thaliana GUS reporter lines. Bars show mean ± SD (n = 3–24) of independent transgenic lines on log-scaled axes. Each data point represents the mean of three technical replicates. Different letters indicate statistically significant differences based on one-way non-parametric ANOVA followed by Dunn’s multiple comparison test (p < 0.05). >MHE1, >MHE2, and >MHE1&2 indicate introduction of the MHE1 pair, the MHE2 pair, or both pairs, respectively, from AhHMA4-1_P into the AtHMA4_P Δ754Δi or Δ597Δi construct. See also Figures 1B–1E.

Figure 5.

Elevated levels and diel dynamics of AhHMA4-1_P-driven GUS reporter transcript levels depend on both MHE2 and CCA1 in A. thaliana, and diel dynamics of HMA4 transcript levels in A. halleri.

(A–D) Relative GUS(A) and CCA1(B) transcript levels, measured through a diurnal cycle, in seedlings of transgenic A. thaliana GUS reporter lines for various HMA4 promoter constructs. Relative GUS transcript levels (C) and GUS specific activities (D) are plotted against CCA1 transcript levels for full-length AhHMA4-1_P and AtHMA4_PGUS reporter constructs introduced into either wild-type (Col-0) A. thaliana plants or the cca1-1 loss-of-function mutant, harvested at ZT1. Two independent transgenic lines (1 and 2) are shown per construct (A–D). Plants were cultivated on agar plates and harvested on day 21 (see also Supplemental Figure 7).

(E–H) Diel time course of relative HMA4(E and G) and CCA1(F and H) transcript levels in shoots (E and F) and roots (G and H) of A. halleri wild-type plants and one HMA4 RNAi line, as well as in A. thaliana. Plants were cultivated hydroponically and harvested on day 21 (E–H; see also Supplemental Figure 8). Data points are mean ± SD (n = 3) of technical replicates.

Horizontal bars indicate day (white fill) and night (black fill) (A, B, and E–H). Data points (A–H) are mean ± SD (n = 3) of technical replicates from one representative experiment out of two independent experiments. For statistics, see Supplemental Dataset 12.

Multiple sequence alignment of the PR, which corresponds to ER2 in the three A. halleri HMA4 promoters, revealed only small segments with sequence similarity (Figure 3C; Supplemental Figure 6A). We identified two copies of the 8-bp consensus motif CACTATCT (Supplemental Figure 6A; Supplemental Dataset 2), with the core motif TATCT, which we designated MHE2 (Figure 3C; −124 AAATATCT −117 and −99 TAATATCA −117 in AhHMA4-1_P). These copies, MHE2a and MHE2b, were absent in the PR of AtHMA4_P, and only MHE2a was found in A. lyrata HMA4_P (Supplemental Figure 6A). In conclusion, the cis-regulatory elements MHE2a and MHE2b were identified in the promoters of all three AhHMA4 gene copies, and MHE1a was identified in two of them (Supplemental Figures 3, 5A, and 6A).

Figure 6.

Working model

In the A. halleri lineage, mutation and natural selection resulted in a pair of functional metal hyperaccumulation element 2 (MHE2) cis-regulatory elements (light pink vertical lines; core sequence TATCT), which are located in the proximal promoter region (berry red arrow) of each of the three tandem HEAVY METAL ATPase 4 (HMA4) gene copies (gold arrow). These cis-regulatory elements co-opt the core circadian clock transcriptional regulator CIRCADIAN CLOCK ASSOCIATED 1 (CCA1; round red shapes) to enhance HMA4 expression, resulting in metal hyperaccumulation and hypertolerance in A. halleri. Two MHE1 elements (dark pink dashed vertical lines; core sequence TGTAACC) in the distal region of AhHMA4-1_P (one MHE1 in AhHMA4-3_P and none in AhHMA4-2_P) also contribute through an unidentified transcriptional regulator. In the A. thaliana HMA4 promoter (blue arrow), we identified two regions repressing transcription, both of which independently and strongly repress translation also. The repressive distal promoter region of AtHMA4 contains a long noncoding RNA (white arrow; AT2G07115) with an exact reverse complement in the HMA4 coding sequence (10 nt long, positions 7–16), which are included in our reporter constructs (arctic blue vertical lines). Sequence comparisons across species suggest that these characteristics, which we hypothesize to participate in translational repression, arose in the A. thaliana lineage. Colored oval shapes encircled by gray lines represent transcription–translation feedback loops of the circadian clock network (Haydon et al., 2011). Cis-regulatory elements or regions that enhance (+) or repress (−) transcription are indicated by circled symbols. HMA4 transcripts are represented by gold lines ending in AAA.

We then tested whether the putative cis-regulatory elements MHE1 and MHE2 are necessary for AhHMA4-1_P promoter activity. Using site-directed mutagenesis, we replaced their characteristic nucleotides by converting MHE1 and MHE2 into the corresponding sequences of AtHMA4_P (Figures 3A–3C). Compared with the FL wild-type AhHMA4-1_P, a mutated AhHMA4-1_P (MHE1_mut) carrying five single-nucleotide substitutions that altered both copies of the MHE1 motif resulted in residual GUS activity of only about 25% (Figures 3D, 3H, 3P, and 3Q; compare MHE1_mut with FL for AhHMA4-1_P). The residual GUS activity of MHE1_mut was comparable to that of constructs in which the MHE1-containing region or the entire DR was deleted (Figures 3G–3I, 3P, and 3Q; compare MHE1_mut with Δ805 and Δ698). Similarly, upon conversion of the two MHE2 motifs in FL AhHMA4-1_P into the corresponding sequences from AtHMA4_P through four single-nucleotide substitutions, we observed residual GUS activity of only about 25%, indistinguishable from that of a construct in which only the PR, corresponding to ER2, was deleted in an otherwise FL AhHMA4-1_P (Figures 3C, 3D, 3J, 3K, 3P, and 3Q; compare MHE2_mut with FL and FL ΔER2). Site-directed mutagenesis of both MHE1 and MHE2 in combination led to substantially reduced GUS activity, at about 10% of that observed for the FL AhHMA4-1_P construct (Figures 3B–3D, 3L, 3M, 3P, and 3Q; compare MHE1/2_mut with Δ86 and FL). In summary, these results support the hypothesis that both MHE1 (in the DR) and MHE2 (in the PR) are necessary for the full, strongly elevated promoter activity of AhHMA4-1_P when compared to that of AtHMA4_P.

We also investigated whether the identified cis-regulatory elements of AhHMA4-1_P, when introduced into AtHMA4_P, were sufficient to confer enhanced promoter strength. We used site-directed mutagenesis to introduce MHE1 and MHE2 of AhHMA4-1_P into AtHMA4_P Δ754Δi, which lacks the intron and retains only the 3′-terminal 157 bp of the DR (Figure 4A). AtHMA4_P Δ754Δi comprises the AtHMA4_P sequence segment (−665 to −621), which is homologous to ER1⁺ from AhHMA4-1_P (−895 to −860; Figure 3B), and the PR corresponding to ER2 from AhHMA4-1_P (Figure 3C). Functionally, AtHMA4_P Δ754Δi was equivalent to FL AtHMA4_P in both GUS transcript levels and activity (Figures 4B–4G, 4L, and 4M). In a mutated AtHMA4_P Δ754Δi harboring intact MHE1a and MHE1b from AhHMA4-1_P, we observed no significant change in GUS transcript levels or GUS activity (Figures 4G, 4H, 4L, and 4M; compare >MHE1 with Δ754Δi). The additional introduction of point mutations to insert both of the MHE2 motifs from AhHMA4-1_P into AtHMA4_P Δ754Δi led to a four-fold increase in relative GUS transcript levels (Figures 3C and 4L; compare >MHE1&2 with >MHE1 and Δ754Δi). A similar effect was observed after introducing only the two MHE2 motifs into AtHMA4_P Δ754Δi (Figure 4L; compare >MHE2 with >MHE1&2 and Δ754Δi). However, as expected, GUS enzyme activity remained low unless the entire DR was deleted (Figures 4B–4M; compare Δ597Δi >MHE2 and Δ597Δi with >MHE2 and Δ754Δi), which is consistent with other observations (Figure 1F; note the complex effects of AtHMA4_P on GUS enzyme activity). These data demonstrate that MHE2 increases the transcript levels of the downstream reporter gene in the sequence context of the A. thaliana promoter.

AhHMA4-1 promoter-dependent transcript levels require CCA1

We compared the MHE1 and MHE2 sequences against databases containing plant transcription factor binding sites using Tomtom, an automated motif comparison tool from the Multiple Expectation–Maximization for Motif Elicitation (MEME) suite (Gupta et al., 2007; Bailey et al., 2009). Proteins containing MYB or MYB-related domains were identified as the most likely direct interactors of both MHE1 and MHE2 (InterPro: IPR006447; Supplemental Figures 5B–5D and 6B–6D; Supplemental Datasets 3 and 4). MHE1 retrieved MYB superfamily proteins of the R2R3 and 3R-MYB families, containing two and three MYB repeats, respectively, as well as trihelix family transcription factors, which possess a domain resembling a single MYB repeat (Kaplan-Levy et al., 2012) (Supplemental Figure 5C; Supplemental Dataset 3). MHE2 retrieved MYB-related R1R2-subgroup transcription factors that contain a single R1/R2-type MYB repeat (Dubos et al., 2010) (Supplemental Figure 6C; Supplemental Dataset 4).

Among the specific transcription factors identified here, the top-scoring R1R2 MYB-related transcription factors predicted to bind MHE2 were members of the REVEILLE subfamily of circadian clock regulators (Rawat et al., 2009). Target genes of these transcription factors typically show circadian and mostly diel oscillations in transcript levels. To test whether activities of the AhHMA4 promoters change in a time-of-day-dependent manner, we quantified GUS transcript levels in our reporter lines over a diel cycle. In the A. thaliana genetic background, relative GUS transcript levels driven by AhHMA4-1_P to -3_P reached a diel maximum around zeitgeber time 9 (ZT9), i.e., 9 h after dawn in the subjective afternoon, at about 1.7- to 2-fold higher than the minimum transcript levels, which were observed at ZT1 or ZT5 (Figure 5A; Supplemental Figure 7A). The transcript levels of CCA1, a member of the REVEILLE subfamily of R1R2 MYB-related transcription factors, were highest in seedlings at approximately ZT1 (Figure 5B). By contrast, GUS transcript levels were much lower and remained constant throughout the day in AtHMA4_P reporter lines, as well as in AhHMA4-1_P reporter lines in which both copies of MHE2 were mutated to the corresponding sequence from AtHMA4_P (Figure 5A; Supplemental Figure 7A; compare AhHMA4-1_P to -3_P with MHE2_mut and AtHMA4_P; see also Figure 3). These results indicate that AhHMA4-1_P to -3_P confer both strongly elevated transcript levels and diel expression dynamics that depend on MHE2 in AhHMA4-1_P, in contrast to AtHMA4_P, which lacks MHE2.

Among the transcription factors predicted to bind MHE2 in silico (Supplemental Figures 6B–6D; Supplemental Dataset 4), we previously observed that CCA1 transcript levels were higher in A. halleri than in A. thaliana (Becher et al., 2004). We therefore tested whether CCA1 is necessary for the elevated promoter strength and diel regulation of the AhHMA4 promoters. We crossed two independent reporter lines of each FL AhHMA4-1_P and AtHMA4_P construct with a cca1 mutant line and a CCA1 overexpression line, both of which were previously characterized (Wang and Tobin, 1998; Green and Tobin, 1999). For AhHMA4-1_P, GUS transcript levels and enzyme activity were reduced to about 55% and 60%, respectively, in cca1 relative to reporter lines in the wild-type genetic background (Figures 5C and 5D). In the cca1 AhHMA4-1_P reporter lines, GUS transcript levels were no longer higher than those in AtHMA4_P reporter lines, in which there was no difference between the Col-0 and cca1 genetic backgrounds (Figure 5C; Supplemental Figures 7B and 7D). Notably, GUS protein levels were repressed in AtHMA4_P reporter lines (Figures 1D and 1E). Diel time courses further support a strong CCA1 dependence of both the elevated magnitude and the diel regulation of transcript levels in both shoots (Supplemental Figures 7B and 7C) and roots (Supplemental Figures 7D and 7E). By contrast, there was no CCA1 dependence of GUS expression in AtHMA4_P reporter lines (Figures 5C and 5D; Supplemental Figures 7B–7E). Diel dynamics in GUS transcript levels under the control of AhHMA4-1_P were also eliminated in a CCA1 overexpression background, with little or no change in magnitude (Supplemental Figures 7F and 7G). Together, these results suggest that the cis-regulatory element MHE2 and CCA1, a transcription factor predicted to bind MHE2 in silico, are necessary for the enhancement and diel regulation of AhHMA4-1_P promoter activity in the A. thaliana genetic background. We also noted that constitutively increased CCA1 expression alone is not sufficient to enhance AhHMA4-1_P promoter activity, which suggests that CCA1-mediated activation may require one or more additional proteins.

Diel regulation of HMA4 transcript levels in A. halleri

Based on our observations in transgenic A. thaliana reporter lines, HMA4 transcript levels are expected to undergo diel cycles in A. halleri. This was confirmed by a diurnal time course of HMA4 transcript levels in shoots and roots (Figures 5E and 5G; Supplemental Figures 8A and 8C). We observed the highest AhHMA4 transcript levels in A. halleri at ZT9 and ZT13 in shoots and roots, respectively. As expected, AtHMA4 transcript levels were low and constant in A. thaliana, whereas AhHMA4 transcript levels in A. halleri depended on the time of day and were between 61-fold and 1200-fold higher in roots and between 43-fold and 77-fold higher in shoots. In shoots, the maximum of CCA1 transcript levels was about three-fold higher in A. halleri than in A. thaliana, which is in qualitative agreement with the microarray hybridization–based observations of Becher et al. (2004) (Figure 5F; Supplemental Figure 8B). In addition, CCA1 transcript levels were unaffected in an A. halleri HMA4 RNAi line, in which HMA4 transcript levels were strongly suppressed in both shoots and roots relative to the wild type around ZT9–ZT13, consistent with Hanikenne et al. (2008). These data demonstrate that endogenous HMA4 transcript levels are under diel regulation in A. halleri but not in A. thaliana, consistent with our promoter analyses in the A. thaliana genetic background. Given our finding that high AhHMA4-1_P activity is dependent on CCA1 in A. thaliana, elevated endogenous transcript levels of AhCCA1 in shoots during the subjective late night and morning (ZT21, ZT25, ZT1, and ZT5) may contribute to the enhanced AhHMA4 transcript levels and their diel dynamics in A. halleri (Figure 5F; Supplemental Figure 8B).

Discussion

AtHMA4 expression is affected by complex repressive mechanisms that predominantly influence the protein level

The DR of the AtHMA4 promoter and the intron upstream of the HMA4 coding sequence very strongly repressed protein accumulation, as shown by the analysis of our reporter constructs (Figure 6; see also Figures 1D and 1E and Supplemental Figure 2). The autonomous repressive effect of the intron on protein levels may involve the annotated short transcript (210593) beginning in the intron that contains a translated upstream open reading frame (µORF) (At2G19110.2_239_350; JBrowse, TAIR), which could be influenced by alternative splicing. A single-nucleotide substitution present in AhHMA4-1 to -3 and AlHMA4 eliminates the translational start codon of this µORF (ATG to ATC; Supplemental Dataset 11, position 2518). In addition, we detected an apparently independent role of the DR in translational repression, which could occur either directly or indirectly by affecting splicing of the AtHMA4 pre-mRNA. Within the transcribed region of an annotated long noncoding RNA (AT2G07115) in the DR of AtHMA4_P (Figure 4A), an upstream sequence (positions −689 to −680, TGTTTTGTAA) is the reverse complement of a short sequence at the 5′ end of the AtHMA4 coding sequence (positions 7–16, TTACAAAACA), which is present in our reporter constructs (Figure 1A; Supplemental Figure 3; Supplemental Dataset 11). In combination, these complementary sequences may contribute to direct translational repression, similar to the mechanistically poorly understood repression of target mRNA translation by the long intergenic noncoding RNA p21 in human cancer cells (Yoon et al., 2012). The high reporter protein activity observed in lines carrying deletions of both the DR and the intron of AtHMA4 suggests that both elements act in cis, as the A. thaliana genetic background contains an intact endogenous AtHMA4 gene and promoter region (alternatively, dosage effects are possible). The 10-bp upstream sequence in the DR is conserved in AhHMA4-1 and AlHMA4 but the formation of the long noncoding RNA is probably not conserved, based on sequence divergence (Supplemental Figure 3). The coding sequences of all three AhHMA4 gene copies, as well as AlHMA4, carry a mismatch C substitution at position 9. Compared with the pronounced effects of the DR and the intron on protein levels, the mildly repressive effects of the same AtHMA4_P regions, i.e., the DR and the intron, on reporter gene transcript levels were different and complex. Overall, these observations and the sequence similarities across species suggest that the complex repressive functions observed in AtHMA4_P are unique to the A. thaliana lineage. Future work will be required to determine the underlying mechanisms and potential causal selective forces.

Cis-regulatory mutations in HMA4 during the evolution of metal hyperaccumulation in A. halleri

AhHMA4 makes the largest contribution to metal hyperaccumulation and hypertolerance in A. halleri (Talke et al., 2006; Courbot et al., 2007; Hanikenne et al., 2008). Compared with AtHMA4, the promoter activities of the three tandem AhHMA4 gene copies in A. halleri are enhanced, resulting in strongly elevated AhHMA4 transcript levels. Here, we identify the causal cis-regulatory sequence differences contributing to AhHMA4 promoter activity—namely, the enhancers MHE1 and MHE2—which are predicted to bind R2R3 MYB and R1R2 MYB-related transcription factors, respectively (Figure 6; see also Figures 1–4; Supplemental Figures 1–6). The effects of deleting or mutating MHE1 and MHE2 in AhHMA4-1_P, either alone or in combination, were consistent with their requirement for, and additive effects on, the transcription of the downstream gene (Figures 1 and 3). The MHE1a and MHE2a/2b elements were present in microsyntenic promoter regions of two and all three HMA4 gene copies, respectively, across all A. halleri accessions for which data are currently available (Hanikenne et al., 2013) (Supplemental Figures 5 and 6). Site-directed mutagenesis experiments introducing these cis-regulatory elements from AhHMA4-1_P of A. halleri (accession Lan3.1) into an AtHMA4 promoter suggested that MHE2 alone is sufficient for enhancing the transcript levels of the downstream reporter gene (Figure 4). It is possible that we were unable to detect the enhancing effect of MHE1 after introduction into AtHMA4_P because of a locally repressive sequence context. We detected repressive functions in the DR of the AtHMA4 promoter, and the combination of both MHE1 and MHE2 within AtHMA4_P did not result in reporter gene transcript levels as high as those of FL AhHMA4-1_P (Figure 4). Alternatively, MHE1 may have no functional relevance for the differences in activity between the A. halleri and A. thaliana HMA4 promoters, despite functioning as an enhancing cis-regulatory element in AhHMA4-1_P.

In all A. halleri individuals, we identified two copies of MHE2 (8 bp) in the PR of each of the three HMA4 gene copies, spaced by 17 (AhHMA4-1_P), 18 (AhHMA4-3_P), or 27–30 (AhHMA4-2_P) bp (Supplemental Figure 6). The conservation of both MHE2 elements and the identical diel rhythmicity of AhHMA4-2_P- and AhHMA4-3_P-driven promoter activity suggest that our findings are applicable to the promoters of all three HMA4 gene copies in A. halleri (Figure 5; Supplemental Figures 7 and 8). Comparison between A. halleri and A. lyrata suggests that a single MHE2 element is not sufficient (Supplemental Figure 6). This is supported by the observation that the loss of one MHE2 element in AhHMA4-3_P ΔDIPRΔi resulted in a substantial decrease in promoter activity (Figures 2A, 2P, and 2Q; compare Supplemental Figure 6). Some sequence variation was observed between the two copies of MHE2 in the PR of each HMA4 gene copy and among the three HMA4 gene copies within each genotype. By contrast, there were almost no polymorphisms between A. halleri individuals originating from European populations, consistent with earlier analyses of broader promoter regions (Hanikenne et al., 2013).

Three of four alleles of an AhHMA4-3–like promoter sequence from A. halleri collected at the Tada mine (Japan) carried point mutations in MHE2b, and two of two carried point mutations in the MHE2a of AhHMA4-1_P. Similarly, 13 of 38 AhHMA4-2_P sequences carried a point mutation in MHE2a (Supplemental Figure 6A), all from individuals collected at highly heavy-metal-contaminated sites in the Harz Mountains and in France (Hanikenne et al., 2013). Reduced HMA4 expression in these plants may contribute to the attenuated Zn and Cd accumulation efficiency observed in A. halleri genotypes from such sites (Stein et al., 2017). This could reflect local selection pressure for nutrient balancing and deserves further investigation (Hanikenne et al., 2013; Stein et al., 2017; Krämer, 2024).

MHE2 directs elevated, rhythmic AhHMA4 promoter activity in a CCA1-dependent manner

The consensus MHE2 motif (CACTATCT) resembles the EE (consensus AAATATCT), and in one case is a perfect match (MHE2a of AhHMA4-1_P; Figure 3C; Supplemental Figure 6A). EE are generally known to mediate circadian-clock-regulated gene expression and serve as target sites for CCA1 binding (Wang et al., 1997; Nagel et al., 2015; Kamioka et al., 2016). The amino acid sequence of the DNA-binding domain of the predicted AhCCA1 protein is 100% identical to that of AtCCA1 (Supplemental Figure 9). EEs can also interact with other clock oscillator proteins, such as LHY and RVE8, all of which belong to the same REVEILLE family of R1R2 MYB-related proteins in A. thaliana (Rawat et al., 2009).

In a cca1 mutant, the transcript levels of AhHMA4-1_P–driven reporters were no higher than those driven by AtHMA4_P (Figure 5), and their diel rhythmicity was lost (Supplemental Figure 7), closely resembling the effects of mutating the MHE2 elements in AhHMA4-1_P. This supports the CCA1 dependence of the difference in promoter strength between A. halleri HMA4-1 and A. thaliana HMA4, providing circumstantial evidence for the functioning of MHE2 as a CCA1-dependent cis-regulatory element (Figure 6). Ample experimental evidence supports the DNA-binding preferences of AtCCA1 (Wang et al., 1997; Wang and Tobin, 1998; Alabadí et al., 2001; Farré et al., 2005; Dong et al., 2011; Nagel et al., 2015). The similarity of MHE2 to known CCA1-binding sequences is consistent with a model in which CCA1 directly binds MHE2 to activate transcription of the downstream reporter gene in our transgenic A. thaliana lines or the endogenous AhHMA4 gene copies in A. halleri. In the roots of AhHMA4-1_P lines, reporter transcript levels were higher in the wild type than in the cca1 genetic background, as expected. In cca1 roots, transcript levels remained higher than those observed in AtHMA4_P lines during the daytime (Supplemental Figure 7D), unlike in shoots. This may indicate an additional contribution in roots from another transcription factor with a function similar to CCA1, potentially a member of the same transcription factor family.

According to the literature, CCA1 can function as both an activator and a repressor of transcription, with repressive activity more commonly reported, particularly for genes encoding components of the evening loop in the core circadian oscillator of A. thaliana (Wang et al., 1997; Wang and Tobin, 1998; Alabadí et al., 2001; Farré et al., 2005; Andronis et al., 2008; Lu et al., 2009; Dong et al., 2011; Nagel et al., 2015; Kamioka et al., 2016). Consequently, CCA1, which lacks characterized activation and repression domains, is considered to activate or repress transcription depending on the regulatory context, such as the presence of interacting proteins, yet the mechanistic basis of this remains poorly understood (Nagel et al., 2015; Kamioka et al., 2016). Although CCA1 and LHY often act as a heterodimeric protein complex (Lu et al., 2009), this complex is apparently not required for normal clock functioning (Kamioka et al., 2016).

CCA1 overexpression is known to disrupt the rhythmicity of circadian-regulated processes. CCA1 expression remained rhythmic in overexpressors, but it was about four-fold elevated at ZT1, and more than ten-fold at other times, compared with the wild type (Supplemental Figure 7G), consistent with the literature (Wang and Tobin, 1998; Lu et al., 2012). By contrast, the cca1 mutant exhibits only mild phenotypes, including a shortened circadian period (as indicated by the transcript levels of core circadian oscillator genes) and early flowering (Green and Tobin, 1999; Mizoguchi et al., 2002). Thus, the strong reduction in AhHMA4-1_P–driven reporter gene expression in the cca1 mutant observed in this study (Figure 5C; Supplemental Figure 7B) is unlikely to reflect a general circadian phenotype but rather supports a more direct role for CCA1 in AhHMA4-1_P–regulated gene expression. Similarly, Green and Tobin (1999) interpreted a decrease in light-induced, phytochrome-dependent responses in the cca1 mutant as evidence for a non-redundant function of CCA1 that is distinct from its role in the circadian clock.

In this study, CCA1 transcript levels peaked around ZT1, whereas both AhHMA4_P-directed reporter gene transcript levels in A. thaliana and HMA4 transcript levels in A. halleri peaked considerably later, at ZT9, in shoots (Figures 5E and 5F). The transcript levels of previously proposed direct target genes of CCA1-dependent transcriptional activation were reported to peak in the late morning to afternoon under diel light–dark cycling conditions (Wang and Tobin, 1998; Farré et al., 2005). Accordingly, the transcript levels of LHCB1.1/CAB2 were maximal around ZT4 and remained high until ZT8, although CCA1 protein was only detectable between ZT22 and ZT4 and not at ZT8 (Wang and Tobin, 1998). Similarly, the transcript levels of the proposed direct CCA1 targets PSEUDO-RESPONSE REGULATOR 7 (PRR7) and PRR9, which encode circadian clock transcription factors, peaked around ZT4 (PRR9) and ZT8 (PRR7) under diel cycling conditions (Farré et al., 2005). CCA1 binds to the PRR5 promoter, and CCA1–FLAG protein levels were detectable between ZT0 and ZT6 whereas PRR5 transcript levels peaked around ZT9 under diel cycling conditions (Kamioka et al., 2016). The authors noted that this delay may reflect the time required for the PRR5 transcript to accumulate or may suggest that CCA1 represses PRR5 transcription in the morning. Because PRR5, PRR7, and PRR9 participate in the complex interacting regulatory loops of the circadian clock, changes in their transcript levels in circadian clock mutants are generally difficult to interpret. Transient assay systems suggested that CCA1 alone has an immediate repressive effect on PRR7 and PRR9 transcript levels (Kamioka et al., 2016). Finally, it was proposed that CCA1-mediated indirect repression of target gene transcription could operate through effects on chromatin structure (Perales and Más, 2007).

In shoots, we observed higher maximal CCA1 transcript levels in A. halleri than in A. thaliana (Figure 5F; Becher et al., 2004), but this was not the case in roots (Figure 5H). Between-species differences in CCA1 protein levels may also contribute in trans to elevated HMA4 gene expression in A. halleri relative to A. thaliana. However, the cis-regulatory effects are known to predominate (Hanikenne et al., 2008), consistent with the results of the present study (Figure 1 and Supplemental Figures 1E and 1F).

Comparison with the HMA4 promoters of other Zn/Cd hyperaccumulator and non-accumulator Brassicaceae species

Our results suggest that pairs of the EE-related cis-regulatory element MHE2 in each of the promoters of the three A. halleri HMA4 gene copies confer CCA1-dependent regulation as part of the pre-existing circadian clock transcriptional network, unlike the AtHMA4 promoter which lacks MHE2 elements. A single MHE2 copy resembling the EE is present in the PR of the HMA4 promoter of the non-hyperaccumulator species A. lyrata, in which leaves contain two-fold higher HMA4 transcript levels than in A. thaliana (Hanikenne et al., 2013) (Supplemental Figure 6). Similar to the presence of MHE2 elements in the PRs of AhHMA4 promoters, we also identified in silico–predicted MHE2 elements in the promoters of the four HMA4 gene copies in the Zn/Cd hyperaccumulator Noccaea caerulescens, designated NcHMA4-1 to -4 (Supplemental Datasets 5 and 6). N. caerulescens belongs to the phylogenetically distant lineage II of the Brassicaceae family, which diverged from lineage I more than 20 million years ago (Clauss and Koch, 2006). Thus, in addition to convergent HMA4 gene copy number expansion, convergent promoter mutations may have occurred in this species, thus amplifying HMA4 transcript levels (Lochlainn et al., 2011). The HMA4 promoter region of A. lyrata also harbors a single MHE1 element in a microsyntenic position (Supplemental Figure 5; Supplemental Dataset 5). We also observed MHE1 elements in the distal promoter regions of all four HMA4 gene copies of N. caerulescens (Supplemental Figure 5) (Lochlainn et al., 2011). By contrast, all A. thaliana accessions and Capsella rubella in lineage I, as well as Brassica rapa and Arabis alpina in lineage II, lack a perfectly identical MHE1, as well as any MHE2 in their proximal promoter regions (Supplemental Datasets 5 and 6).

Among all the candidate genes exhibiting substantial increases in transcript levels in metal hyperaccumulator plants (Becher et al., 2004; Hammond et al., 2006; van de Mortel et al., 2006), only METAL TRANSPORT/TOLERANCE PROTEIN 1 (MTP1) promoters have been functionally dissected to date (Fasani et al., 2017). The promoters of three AhMTP1 gene copies (α, β, and γ), which originated from a highly metal-contaminated site, mediated much higher reporter gene activity than the promoter of AtMTP1. This difference was attributed to the segment from 810 to 362 bp upstream of the translational start site, which lacked any between-species differences in the occurrence of known cis-regulatory elements. Site-directed mutagenesis of either two predicted telo boxes (−303 and −400) or two predicted MYB-binding sites (−125 and −155), all located in the transcribed region of AhMTP1, indicated mildly repressive cis-regulatory functions. Beyond this, the authors reported that the predicted MYB-binding sites in the AhMTP1γ promoter, which are absent from the A. thaliana ortholog, were required for reporter activity in A. thaliana trichomes and conferred increased Zn tolerance upon introduction into an A. thaliana mtp1 mutant when positioned upstream of the AtMTP1 coding sequence (Fasani et al., 2017).

Co-option of a gene regulatory network following cis-regulatory mutations in HMA4

Gene regulatory network co-option is often conceptualized as involving trans-regulatory changes, such as a novel expression domain of an existing transcription factor, following the rationale that one such mutation can alter the expression of multiple downstream target genes in the regulatory hierarchy (True and Carroll, 2002; McQueen and Rebeiz, 2020). By comparison, cis-regulatory mutations at multiple loci in the genome would be required to achieve an equivalent alteration in overall gene expression, which is far less likely to occur through evolutionary processes. Here, cis-regulatory change results in the co-option of a regulatory network that increases transcript abundance of HMA4, which encodes a pump that exports heavy metal cations, primarily Zn²⁺ and Cd²⁺, from specific cell types (Figure 6). This has the benefits of increasing metal accumulation in leaves, thereby providing an elemental defense against herbivory (Kazemi-Dinan et al., 2014), and of increasing heavy metal tolerance to allow survival in soils containing toxic levels of Zn and Cd (Hanikenne et al., 2008; Kazemi-Dinan et al., 2014).

It remains to be determined whether the diel rhythmicity of HMA4 transcript abundance is mirrored at the HMA4 protein level. This depends on the unknown stability of the HMA4 protein, for instance. Here, we report that both MHE2 elements and a functional CCA1 gene are required not only for diel cycles in HMA4 transcript levels but, importantly, also for the generally elevated HMA4 transcript abundance (Figures 5 and 6; Supplemental Figure 7). The amplitude of diel changes in the transcript levels of genes downstream of AhHMA4 promoters was generally smaller than the differences in expression observed between wild-type AhHMA4 promoters and the AhHMA4-1 promoter carrying MHE2 mutations or expressed in the cca1 mutant background, as well as between the AhHMA4 and AtHMA4 promoters. Therefore, selection may have favored higher, localized HMA4 gene expression in A. halleri, whereby the rhythmicity of HMA4 transcript levels occurred as a side effect with little or no biological consequence.

Metal hyperaccumulation functions as an elemental defense against biotic stress in A. halleri (Kazemi-Dinan et al., 2014), and the defenses of other plants are also regulated by the circadian clock (e.g., Goodspeed et al., 2012). Importantly, CCA1 and HMA4 transcript levels also exhibit seasonal rhythms in the field (Nagano et al., 2019). Strong diel and seasonal dynamics in herbivore feeding activity occur across the geographic range of A. halleri. We expect that biologically relevant fluctuations in total leaf metal concentrations are more likely to result from longer-term seasonal rhythms than from short-term diel oscillations in HMA4 transcript levels.

Alternatively, HMA4 may act in concert with the metal homeostasis network, which is rhythmically organized on both diel and seasonal timescales, consistent with the rhythmicity of transcript levels observed for several other metal homeostasis genes in A. halleri (Nagano et al., 2019) and the circadian transcriptional regulation of genes involved in metal homeostasis in A. thaliana (e.g., Haydon et al., 2011; Zhang and Krämer, 2018). Accordingly, we expect substantial diel dynamics among localized metal fluxes in plants. Future research will address these hypotheses.

Methods

Plant material and growth conditions

Seeds of A. thaliana wild type (Col-0) were obtained from Lehle Seeds (Round Rock, TX, USA). Seeds of A. thaliana cca1-1 (N67781; derived from the cca1-1 mutant N67780 in the Ws background via six backcrosses with Col-0; note that none of the other available and previously characterized lines are in the Col-0 genetic background) carrying a T-DNA insertion in the fourth intron of CCA1 (Krysan et al., 1996; Green and Tobin, 1999; Yakir et al., 2009) and a transgenic line expressing CCA1 (CCA1-OX o38, N67794 in the Col-0 background) under the control of the cauliflower mosaic virus 35S promoter (Wang and Tobin, 1998) were obtained from the Nottingham Arabidopsis Stock Centre. The cca1 mutant lacks both the FL CCA1 transcript and detectable levels of the CCA1 protein (Green and Tobin, 1999), despite the apparent presence of low levels of residual transcripts detected in our RT–qPCR (Figures 5C and 5D and Supplemental Figures 7C and 7D), which likely reflects either functionally distinct variants or antisense transcripts overlapping our amplicon. Homozygous lines were confirmed through late flowering (CCA1-OX o38) (Wang and Tobin, 1998) and by PCR following DNA extraction (Edwards et al., 1991) (cca1-1) (Supplemental Table 1). Transgenic A. thaliana GUS reporter lines (L. Heynhold, accession Col-0) with the FL AhHMA4-1 and AtHMA4 promoters were described by Hanikenne et al. (2008) and those carrying AhHMA4-2 and -3 promoters by Nouet et al. (2015).

The AtHMA4_P construct comprised a 2799-bp genomic fragment that included the initial 204 bp of the AtHMA4 coding sequence, corresponding to 68 amino acids predicted to be cytosolic; in the full-length protein, these residues are followed by eight conserved transmembrane helices (TMHs), with the first TMH predicted to include amino acids 95–115 (TmConsens, https://aramemnon.botanik.uni-koeln.de/). A. halleri (L.) O'Kane and Al-Shehbaz ssp. halleri, population Langelsheim, accession Lan3.1, was collected in the field (Becher et al., 2004). Propagation of A. thaliana homozygous lines, genotyping, and sterile cultivation were performed as described by Sinclair et al. (2018) (Supplemental Methods 1). For RNA and protein isolation, whole seedlings cultivated on agar-solidified sterile media or, alternatively, roots and shoots separated with a scalpel were harvested at ZT1 on day 21 unless otherwise specified, frozen in liquid nitrogen as pooled samples from four replicate plates (80 plants in total) per line, and stored in 50-ml screw-cap polypropylene tubes at −80°C until further processing.

For diel time-course experiments including A. halleri, vegetative cuttings from Lan3.1 (wild type) and the AhHMA4 RNAi line 4.2.1 (Hanikenne et al., 2008) were transferred into 50-ml polypropylene tubes containing modified Hoagland solution for rooting and pre-cultivation, and A. thaliana Col-0 seedlings were pre-cultivated under sterile conditions (Supplemental Methods 1). For the experiment, 21-day-old A. thaliana seedlings and 17-day-old A. halleri clones were transferred into 400-ml vessels containing modified Hoagland solution. All hydroponic solutions were exchanged weekly. Roots and shoots were harvested every 4 h over a 24-h cycle beginning at ZT1 on day 21 of cultivation.

DNA cloning and plant transformation

Promoter deletion constructs were generated in groups according to procedures 1–3 (Supplemental Tables 1 and 2). Cloning and plant transformation procedures are described in detail in Supplemental Methods 1.

Crossing and selection of homozygous lines

A. thaliana homozygous cca1-1 mutant and CCA1-OX o38 transgenic plants were used as pollen acceptors in crosses with two independent GUS reporter lines for each AtHMA4_P and AhHMA4-1_P (pollen donors), as well as for AhHMA4-2_P and AhHMA4-3_P (for CCA1-OX o38 only) (Weigel and Glazebrook, 2006). Five to ten inflorescences were pollinated and marked with colored threads, and siliques were harvested individually 21 days later. F1 seeds were germinated on 50 μg ml⁻¹ kanamycin, and resistant plants were transferred to 30 μg/ml hygromycin after 14 days, with wild-type Col-0 plants of the same age as controls. Plants resistant to both antibiotics were transferred to soil, F2 seeds were harvested and sown, and segregation ratios were quantified separately on either 50 μg ml⁻¹ kanamycin or 30 μg ml⁻¹ hygromycin to confirm single-locus T-DNA insertions, until lines homozygous for both markers were obtained (see above). The cca1-1 mutant allele and the GUS transgene were verified by PCR (Supplemental Table 1).

RNA extraction, cDNA synthesis, and reverse transcription quantitative PCR

RNA extraction, cDNA synthesis, and RT–qPCR were performed as described by Zhang et al. (2022). Details are given in Supplemental Table 1, Supplemental Datasets 7–10, and Supplemental Methods 1.

Histochemical staining and imaging

Ten freshly harvested 10-day-old seedlings, cultivated on agar plates containing modified Hoagland solution, were taken from each of the seven independent homozygous lines per construct and immersed in 1 ml of GUS staining buffer in a 2-ml polypropylene reaction vial (Zhang et al., 2022). The tubes were vacuum infiltrated for 1 min and then incubated in the dark at 37°C for 4 h. The GUS staining buffer was removed with a pipette, and the samples were cleared in 1 ml of 75% (v/v) ethanol on a rocking shaker at 20 cycles per minute and room temperature, repeated four times for 2 h each and once overnight. Seedlings were kept in 1 ml of 75% (v/v) ethanol and either directly mounted on microscope slides for imaging with a light microscope (Olympus, Hamburg, Germany; catalog no. VS120) or stored in the dark. Images shown are representative of all stained seedlings and independently transformed lines.

Protein extraction, quantification, and β-glucuronidase activity assays

Samples of 100 mg frozen, homogenized plant tissue were added to 200 μl of ice-cold protein extraction buffer containing 50 mM sodium phosphate buffer (pH 7.4), 10 mM β-mercaptoethanol, 10 mM EDTA, 0.1% (v/v) SDS, and 0.1% (v/v) Triton X-100, followed by alternating intervals of vortexing at 3200–3500 rpm for 10 s and resting on ice for 30 s until fully thawed. After centrifugation in a microcentrifuge at 4°C and 18 500 × g for 20 min, the supernatant was divided into aliquots and stored at −80°C until use. Total protein quantification was performed using the Bradford assay with bovine serum albumin as a standard. Twenty microliters of a 1:10 dilution of protein extract in dilution buffer (50 mM sodium phosphate [pH 7.4] and 10 mM EDTA) was mixed with 180 μl of Bradford solution in the wells of a 96-well microtiter plate. After 10 min at room temperature, absorbance was measured at 595 nm using a BioTek Synergy HTX multimode plate reader (BioTek Instruments, Winooski, VT, USA), and protein concentrations were calculated from the standard curve. Quantification of β-glucuronidase activity was performed as described previously (Jefferson et al., 1987; Gallagher, 1992).

Sequence comparisons and motif analyses

Homologous segments among HMA4 promoter regions were identified based on percentages of nucleotide identity from multiple sequence alignments of FL promoter regions (Supplemental Dataset 11; Supplemental Methods 1) using Clustal Omega (Sievers et al., 2011). The five statistically most supported motifs of 5–12 nt in length, occurring any number of times within the ER1⁺ or ER2 regions of the AhHMA4-1, -2, and -3 promoters (Supplemental Figures 5 and 6; Supplemental Table 3), were identified using default settings in MEME (Bailey et al., 2009; Hanikenne et al., 2013) (Supplemental Datasets 1 and 2). We then performed multiple sequence alignments using the microsyntenic ER1⁺ and ER2 promoter segments described above. Based on MEME analyses and the multiple sequence alignments, sequence motifs corresponding to MHE1 and MHE2 were identified as conserved among the promoters of AhHMA4 gene copies and across A. halleri accessions (Supplemental Table 3). Motif logos were generated automatically using the download option in the MEME suite (Bailey et al., 2009). Subsequently, we used the automated Motif Comparison Tool Tomtom within the MEME suite to identify transcription factor binding sites resembling MHE1 or MHE2 motifs (Gupta et al., 2007). We compared the MHE1 and MHE2 motifs against three plant transcription factor binding site databases in the MEME suite—JASPAR2018_CORE_plants_non-redundant_v2.meme (Castro-Mondragon et al., 2022), ArabidopsisDAPv1.meme (O’Malley et al., 2016), and ArabidopsisPBM_20140210.meme (Franco-Zorrilla et al., 2014)—using default parameters and the three comparison functions available in the Tomtom tool (Pearson correlation coefficient, Euclidean distance, and Sandelin–Wasserman similarity), and recorded each protein hit (Supplemental Datasets 3 and 4). We calculated percentages as the sum of all protein hits for each transcription factor family divided by the total number of transcription factor protein hits. Similarly, we calculated the sum of the hits for each individual transcription factor protein divided by the total number of protein hits for its respective protein family. To assess individual occurrences of MHE1, MHE2, and the EE (Harmer et al., 2000) motifs across HMA4 promoter regions, we used the Find Individual Motif Occurrences tool in the MEME suite with a p-value cutoff of <10⁻⁴ (Grant et al., 2011) (Supplemental Datasets 5 and 6).

Data and code availability

The data underlying this article are either publicly available or provided in this article and its online supplemental information. Raw data and materials will be shared upon reasonable request to the corresponding author.

Funding

This work was funded by the National Council of Humanities, Sciences and Technologies (Consejo Nacional de Humanidades, Ciencias y Tecnologías [CONACYT]) scholarship no. 438349 and a scholarship for the completion of dissertations at Ruhr University Bochum from the Wilhelm and Günter Esser Foundation (L.C.), by the German Research Foundation (Deutsche Forschungsgemeinschaft [DFG]) project no. 5453631 and European Research Council (ERC)-AdG 788380 “LEAP EXTREME” (U.K.), with additional contributions from the Fonds de la Recherche Scientifique - FNRS (FRFC-2.4583.08 and PDR-T.0206.13), the University of Liège (SFRD-12/03), and the Belgian Program on Interuniversity Attraction Poles (IAP no. P7/44) (M.H.).

Acknowledgments

We are grateful to Andreas Aufermann, Martin Pullack, and Dr. Hassan Ahmadi, Ruhr University Bochum, for advice and technical assistance. The authors declare no conflicts of interest.

Author contributions

L.C., M.H., and U.K. conceived the project; L.C. and U.K. designed the research; all shown experiments were conducted by L.C.; M.H., J.C., C.N., and J.S. generated constructs for the initial deletion series; L.C. and J.C. generated the follow-up constructs based on AhHMA4-1_P and AtHMA4_P; L.C., N.J., and U.K. analyzed the data; and L.C. and U.K. wrote the manuscript. All authors read and edited the manuscript.

Published: October 3, 2025