Soybean production in low-latitude regions is more than 50 per cent of the total worldwide production (United States Department of Agriculture, 2023). Therefore, it is very important to increase soybean yield in low-latitude regions. The branching number and flowering time are the major factors affecting soybean grain yield (Fang et al., 2024). Delaying flowering and maturity, and increasing the branch number can improve the final soybean yield by increasing the number of pods per plant (Dong et al., 2021; Sun et al., 2019). For example, the branch number was significantly increased and flowering time was delayed in the ap1 quadruple mutant and dt2 mutant, improving grain yield in soybean (Chen et al., 2020; Liang et al., 2022). Therefore, modulating the branch number and maturity are crucial for high-yield soybean breeding. However, only a few genes that regulating both branch number and flowering time have been identified.

In total, 26 CONSTANS (CO) homologues have been identified in soybean, but only the functions of COL1a, COL1b, COL2a and COL2b have been reported (Wu et al., 2014). In this study, two independent T₅-generations of transgenic soybean lines that homozygous COL3a-overexpressing (COL3a-OE) were obtained (Figure 1a,b), and used to examine the agronomic traits under natural short-day (SD) and long-day (LD) field conditions in Guangzhou and Shijiazhuang, respectively. The results showed that COL3a-OE transgenic lines flowered and matured significantly later than the wild-type Williams 82 (W82) in the field of Guangzhou (Figure 1c–e) and Shijiazhuang (Figure 1f,g). In addition, COL3a-OE transgenic lines exhibited significantly increased branch numbers and improved overall grain yields compared to that of wild-type W82 (Figure 1d–g). We generated loss-of-function mutants of COL3a (named col3a^CR) on a W82 background using the CRISPR/Cas9-mediated gene editing (Figure S1a–c) to further investigate the function of COL3a. DNA sequencing identified a col3a^CR mutant carrying a 76-bp nucleotide deletion between targets 1 and 2, and a frameshift mutation was introduced (Figure S1a–c). There was no significant difference between the col3a^CR mutant and wild-type W82 under SD or LD conditions in the growth chamber (Figure S1d–g). We speculated that the functionally redundant of duplicated homologous genes are one of the main reasons why the col3a^CR mutant has no phenotype. These results showed that the overexpression of COL3a significantly enhanced grain yield by increasing branch number and delaying maturity in soybean.

The expression pattern of COL3a was firstly investigated in different soybean organs to understand the molecular mechanism of how COL3a regulate flowering and branching in soybeans. The results showed that COL3a was constitutively expressed in flowers, leaves, stems, roots and shoot apexes, but it was highly expressed in the leaves (Figure S2a). The subcellular localization of the COL3a protein was also determined in Arabidopsis protoplasts. We found that the COL3a-GFP fusion protein was located in the nucleus, whereas the GFP control was located primarily in the nucleus and cytoplasm (Figure S2b).

Previous studies have shown that the legume-specific E1 gene plays a central role in photoperiod-regulated flowering and maturity (Xia et al., 2012) by regulating the expression of FT2a and FT5a genes in soybean. We first investigated E1 expression in COL3a-OE and W82 soybean plants to test whether COL3a can regulate the expression of E1. The transcription level of E1 was higher in the COL3a-OE than in W82 plants (Figure 1h), and FT2a and FT5a expression levels were lower in COL3a-OE plants than in W82 plants (Figure S3a,b). Transient expression assays also showed that COL3a induced the expression of the pE1::LUC reporter gene (Figure 1i). Chromatin immunoprecipitation (ChIP)-qPCR revealed that COL3a was directly associated with the E1 promoter regions containing a core-like-motif (CCACA, Figure 1j). We crossed COL3a-OE2 with e1^CR mutant in W82 background to develop COL3a-OE2/e1^CR lines and subjected them to phenotypic evaluation to further explore the genetic interaction of COL3a and E1. The COL3a-OE2 plants showed delayed flowering in both the e1^as and e1^CR genetic backgrounds; however, the effect was weaker in the e1^CR background, implying that the full effect of COL3a on flowering mainly depends on E1 (Figure 1k). These results combined indicated that COL3a directly binds to the promoter of E1 and activates its expression. Notably, this is the first gene to be identified that directly activates E1 expression in soybeans.

SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) transcription factors play a critical role in regulating the number of soybean branches (Bao et al., 2019; Sun et al., 2019). We performed RT-qPCR assay on COL3a-OE2 and W82 soybean plants to test whether COL3a can regulate expression SPL genes to control branch number in soybean shoot apex. The results showed that a large number of SPL genes were down-regulated in COL3a-OE2 transgenic soybean plants compared to wild-type W82 (Figure 1l), including SPL9a and SPL9b, which have been confirmed to increase branching in soybeans (Bao et al., 2019).

The natural variation of the COL3a coding sequence was analysed in 617 previously resequenced soybean accessions, including 177 wild, 28 landrace and 412 cultivar soybeans (Dong et al., 2022; Kou et al., 2022) to explore the evolutionary origin of the different alleles in COL3a. Two unique, high-confidence haplotypes were identified in COL3a gene. The 6-bp deletion in haplotype 2 (COL3a^H2) was identical to that of all other COL3 homologues in legumes, suggesting that COL3a^H2 is the original haplotype in soybean (Figure 1m,n). A dual-luciferase transient expression assay showed that COL3a^H2 has a stronger ability to activate the expression of E1 than COL3a^H1 (Figure S4). Varieties carrying COL3a^H2 showed delayed flowering compared to that of COL3a^H1 (Figure 1o,p). Next, we examined the percentages of the different alleles in the improved cultivars, landraces and wild soybeans in our panel of 617 resequenced accessions. The COL3a^H1 allele was present in 43.7% of the wild soybeans, whereas COL3a^H2 was present in 56.3%, indicating that the COL3a^H2 allele is a major genetic variant in wild soybeans (Figure 1m). The frequency of COL3a^H1 increased to 96.4% and 99.8% in the landraces and cultivars, respectively, suggesting that COL3a^H1 have been underwent strong artificial selection during post-domestication (Figure 1m). We further identified strong evidence of selection in a region of 108 kb that contains COL3a gene and 17 other genes (Figure 1p and Table S1). These results suggested that the COL3a^H1 allele is targeted by selection, thereby causing its rapid accumulation in domesticated soybeans.

In conclusion, we identified that the COL3a gene play a key role in regulating maturity and branch number to control grain yield in soybean and that the earlier flowering alleles of COL3a^HI have undergone artificial selection in modern cultivar soybean in high-latitude regions. Our findings also provide a biotechnological strategy for introducing COL3a^H2 allele into modern soybean to create high-yielding soybean in low latitude by delaying flowering and increasing branch number.

低纬度地区的大豆产量占全球总产量的 50% 以上（美国农业部，2023 年）。因此，提高低纬度地区的大豆产量非常重要。分枝数和开花时间是影响大豆籽粒产量的主要因素 （Fang et al.， 2024）。延迟开花和成熟，增加分枝数可以通过增加每株植物的豆荚数量来提高大豆的最终产量（Dong et al.， 2021;Sun等 人，2019 年）。例如，ap1 四重突变体和 dt2 突变体的分支数显著增加，开花时间延迟，从而提高了大豆的产量（Chen等 人，2020 年;Liang et al.， 2022）。因此，调节分枝数和成熟度对于高产大豆育种至关重要。然而，目前仅鉴定出少数几个同时调节分枝数和开花时间的基因。

在大豆中总共鉴定出 26 个 CONSTANS （CO） 同源物，但仅报道了 COL1a、COL1b、COL2a 和 COL2b 的功能（Wu等 人，2014 年）。在本研究中，获得了两个独立的 T₅ 代纯合 COL3a 过表达 （COL3a-OE） 的转基因大豆品系 （图 1a、b），并分别用于检查广州和石家庄自然短日照 （SD） 和长日照 （LD） 条件下的农艺性状。结果表明，在广州（图 1c-e）和石家庄（图 1f，g）领域，COL3a-OE 转基因品系的开花和成熟时间显著晚于野生型 Williams 82 （W82）。此外，与野生型 W82 相比，COL3a-OE 转基因系表现出分支数显著增加，总籽粒产量提高（图 1d-g）。我们使用 CRISPR/Cas9 介导的基因编辑在 W82 背景上生成了 COL3a 的功能丧失突变体（命名为 col3a^CR）（图 S1a-c），以进一步研究 COL3a 的功能。DNA 测序鉴定出一个 col3a^CR 突变体，在靶标 1 和 2 之间携带 76 bp 核苷酸缺失，并引入了移码突变（图 S1a-c）。在生长室的 SD 或 LD 条件下，col3a^CR 突变体和野生型 W82 之间没有显着差异（图 S1d-g）。我们推测，重复同源基因的功能冗余是 col3a^CR 突变体没有表型的主要原因之一。 这些结果表明，COL3a 的过表达通过增加大豆的分枝数和延迟成熟来显著提高籽粒产量。

首先研究 COL3a 在大豆不同器官中的表达模式，以了解 COL3a 如何调节大豆开花和分枝的分子机制。结果表明，COL3a 在花、叶、茎、根和芽尖中组成型表达，但在叶中高度表达（图 S2a）。在拟南芥原生质体中也确定了 COL3a 蛋白的亚细胞定位。我们发现 COL3a-GFP 融合蛋白位于细胞核中，而 GFP 对照主要位于细胞核和细胞质中（图 S2b）。

以前的研究表明，豆科植物特异性 E1 基因通过调节大豆中 FT2a 和 FT5a 基因的表达，在光周期调节的开花和成熟中起核心作用 （Xia et al.， 2012）。我们首先研究了 E1 在 COL3a-OE 和 W82 大豆植株中的表达，以测试 COL3a 是否可以调节 E1 的表达。E1 在 COL3a-OE 中的转录水平高于 W82 植物（图 1h），FT2a 和 FT5a 在 COL3a-OE 植物中的表达水平低于 W82 植物（图 S3a、b）。瞬时表达测定还显示 COL3a 诱导 pE1：：LUC 报告基因的表达（图 1i）。染色质免疫沉淀 （ChIP）-qPCR 显示 COL3a 与包含核心样基序的 E1 启动子区域直接相关 （CCACA，图 1j）。我们将 W82 背景下的 COL3a-OE2 与 e1^CR 突变体杂交，开发 COL3a-OE2/e1^CR 系，并对其进行表型评价，以进一步探索 COL3a 和 E1 的遗传相互作用。COL3a-OE2 植株在 e1^as 和 e1^CR 遗传背景中均表现出延迟开花;然而，在 e1^CR 背景下效果较弱，这意味着 COL3a 对开花的全部影响主要取决于 E1（图 1k）。 这些结果综合表明，COL3a 直接与 E1 的启动子结合并激活其表达。值得注意的是，这是第一个被鉴定的直接激活大豆中 E1 表达的基因。

鳞状启动子结合蛋白样 （SPL） 转录因子在调节大豆分枝数量中起关键作用（Bao等人 ，2019 年;Sun等 人，2019 年）。我们对 COL3a-OE2 和 W82 大豆植株进行了 RT-qPCR 检测，以检测 COL3a 是否能调节 SPL 基因的表达以控制大豆芽顶端的分支数。结果表明，与野生型 W82 相比，COL3a-OE2 转基因大豆植株中大量 SPL 基因下调（图 1l），包括 SPL9a 和 SPL9b，已被证实会增加大豆的分枝（Bao et al.， 2019）。

在 617 个先前重测序的大豆种质中分析了 COL3a 编码序列的自然变异，包括 177 个野生大豆、28 个地方品种和 412 个栽培品种大豆（Dong等人 ，2022 年;Kou et al.， 2022）探索 COL3a 中不同等位基因的进化起源。在 COL3a 基因中鉴定出两种独特的、高置信度的单倍型。单倍型 2 （COL3a^H2） 中的 6 bp 缺失与豆科植物中所有其他 COL3 同源物的缺失相同，表明 COL3a^H2 是大豆中的原始单倍型（图 1m，n）。双荧光素酶瞬时表达测定显示，COL3a^H2 激活 E1 表达的能力比 COL3a^H1 强（图 S4）。与 COL3a^H1 相比，携带 COL3a^H2 的品种表现出延迟开花（图 1o，p）。接下来，我们在 617 个重测序种质的面板中检查了改良品种、地方品种和野生大豆中不同等位基因的百分比。COL3a^H1 等位基因存在于 43.7% 的野生大豆中，而 COL3a^H2 存在于 56.3% 的野生大豆中，表明 COL3a^H2 等位基因是野生大豆中的主要遗传变异（图 1m）。COL3a^H1 的频率增加到 96.4% 和 99。在地方品种和栽培品种中分别为 8%，表明 COL3a^H1 在驯化后经历了强烈的人工选择（图 1m）。我们进一步确定了在包含 COL3a 基因和 17 个其他基因的 108 kb 区域中选择的有力证据（图 1p 和表 S1）。这些结果表明，COL3a^H1 等位基因是通过选择靶向的，从而导致其在驯化大豆中快速积累。

综上所述，我们发现 COL3a 基因在调控大豆成熟度和分枝数以控制籽粒产量方面起关键作用，并且 COL3a^HI 的早期开花等位基因在高纬度地区现代品种大豆中经历了人工选择。我们的研究结果还为将 COL3a^H2 等位基因引入现代大豆提供了一种生物技术策略，通过延迟开花和增加分枝数在低纬度地区创造高产大豆。