The tomato CONSTANS-LIKE protein SlCOL1 regulates fruit yield by repressing SFT gene expression

Background CONSTANS (CO) and CONSTANS-LIKE (COL) transcription factors have been known to regulate a series of cellular processes including the transition from the vegetative growth to flower development in plants. However, their role in regulating fruit yield in tomato is poorly understood. Result In this study, the tomato ortholog of Arabidopsis CONSTANS, SlCOL1, was shown to play key roles in the control of flower development and fruit yield. Suppression of SlCOL1 expression in tomato was found to lead to promotion of flower and fruit development, resulting in increased tomato fruit yield. On the contrary, overexpression of SlCOL1 disturbed flower and fruit development, and significantly reduced tomato fruit yield. Genetic and biochemical evidence indicated that SlCOL1 controls inflorescence development by directly binding to the promoter region of tomato inflorescence-associated gene SINGLE-FLOWER TRUSS (SFT) and negatively regulating its expression. Additionally, we found that SlCOL1 can also negatively regulate fruit size in tomato. Conclusions Tomato SlCOL1 binds to the promoter of the SFT gene, down-regulates its expression, and plays a key role in reducing the fruit size. Supplementary Information The online version contains supplementary material available at 10.1186/s12870-022-03813-4.


Background
Tomato is one of the most important vegetable crops cultivated worldwide. It also serves as a model plant for research on fruit development and fruit ripening. Breeding for high yield has been one of the ultimate goals for crop breeders. Inflorescence architecture is the main determinant of flower number and crop yield [1]. With the increasing demands for tomato, higher standards for the high-yield tomato varieties have been put forward. Therefore, a better understanding of the key genes that regulate tomato fruit yield is very important for commercial production.
FLOWERING LOCUS T (FT) has been shown to be a key protein at the convergence of several signaling pathways and serves as the key flowering initiation signal, i.e. the florigen, in Arabidopsis. The function of FT as the flowering inducer is conserved among plant species [2][3][4][5]. SINGLE-FLOWER TRUSS (SFT), the tomato ortholog of FT, regulates primary flowering time, sympodial habit, and inflorescence development [6,7]. Tomato sft mutant plants produce flowers later than the wild type, and the inflorescences revert to indeterminate vegetative branches or become a single flower, and the yield of the mutant is significantly decreased [6,7]. Several regulatory factors of FT have Open Access *Correspondence: zhangjunhng@mail.hzau.edu.cn been identified. The trimeric Nuclear Factor-Y (NF-Y) complexes, comprising CO/NF-YB/NF-YC, bind to the CCAAT DNA element of the FT gene promoter and regulate flowering time [8,9]. Arabidopsis CONSTANS promotes FT gene expression, accelerating flowering in the long day condition [10,11]. In rice, Heading date 1 (Hd1), the ortholog of Arabidopsis CONSTANS, promotes flowering under short-day conditions, but delays flowering under long-day conditions by regulating the expression of the rice FT ortholog, Heading date 3a (Hd3a) [12]. Thus, more transcription factors involved in regulating the expression of SFT need to be explored in tomato.
CONSTANS is a B-box (BBX) protein, originally identified in Arabidopsis thaliana [13]. There are 32 BBX family members in Arabidopsis, which can be divided into five structural groups, based on the number and sequence features of the B-box domain and the presence or absence of a CCT domain [14]. CONSTANS has been identified as a mediator of the circadian clock in controlling the flowering time in Arabidopsis [4,5,10,11]. CONSTANS-LIKE (COL) genes have been studied in many other plant species. Overexpression of COL5 can induce flowering in short-day grown Arabidopsis [15]. On contrary, overexpression of COL9 delays flowering by reducing the expression of CO and FT in Arabidopsis [16]. OsCOL3, a rice CONSTANS-LIKE gene, controls flowering time by down-regulating the expression of FT-like genes under short-day conditions [17]. OsCOL13 functions as a negative regulator of flowering downstream of OsphyB and upstream of Ehd1 in rice [18]. In tomato and tobacco, overexpression of COL1 and COL3 has resulted in late-flowering phenotypes [19]. However, it remains unknown if COL1 participates in direct regulation of SFT gene expression in tomato.
Previous studies on CO and COL proteins have been focused mainly on their roles in mediation of the circadian clock and flowering time in plants. Here we show that SlCOL1 binds to the promoter of SFT and negatively regulates its expression. Suppression of SlCOL1 gene expression in transgenic tomato lines increased the flower and fruit numbers and the size of fruits. On the other hand, its overexpression in transgenic tomato lines resulted in increase in the number of vegetative inflorescences, decrease in the numbers of flowers and fruits, and reduction in the size of fruits. Furthermore, yeast one-hybrid experiments and GUS reporter assays showed that SlCOL1 can directly bind to the cis-regulatory elements of the SFT promoter. These findings provide new insight on how SlCOL1 negative regulates tomato fruit yield.

Expression patterns of SlCOL1
SlCOL1 (Solyc02g089540), also referred to as SlBBX3, is the ortholog of Arabidopsis CONSTANS (CO) protein.
SlCOL1 has an ORF of 1176 bp, encoding a protein of 391 amino acid residues that contains two B-box domains and a CCT domain. The B-box is a conserved 88-amino acid region and the two B-box domains span the region of the amino acid residues . The CCT domain is a conserved 45-amino acid region (964-1098). Gene expression analysis showed that SlCOL1 was expressed in all tested tissues, with the highest expression in the mature leaves and flowers (Fig. 1A). Analysis of GUS staining in ProSlCOL1::GUS transgenic line 10 plants 90 days post anthesis (dpa) showed high expression of SlCOL1 in the apex (SAM) and flowers, and very low in the stems, young leaves and fruits ( Fig. 1B-C).
The subcellular localization of SlCOL1 protein was determined using confocal laser scanning microscopy. Bioinformatics analysis indicated that a nuclear localization signal is present in the same region where the B-box and CCT domain reside. The nuclear marker protein Ghd7 [20] was fused with the cyan fluorescent protein (CFP) for the identification of the nucleus. We found that the SlCOL1-GFP protein was localized exclusively to the nucleus and its green fluorescence fully overlapped the cyan fluorescence of Ghd7-CFP, when co-expressed in the N. benthamiana protoplasts (Fig. 2). In contrast, the free GFP fluorescence was distributed throughout the cell (Fig. 2B). Thus, SlCOL1 is a nuclear protein, which is consistent with its function as a transcription factor.

Regulation of inflorescence morphology and fruit numbers
To better understand the function of SlCOL1, we generated SlCOL1 RNA interference (RNAi) and overexpression (OE) tomato transgenic lines. Three independent lines (OE-5, OE-6 and OE-8, and RNAi-1, RNAi-10 and RNAi-17) from each transformation were selected for further analysis (Fig. S1A). The transgenic tomato lines were morphologically distinguishable from the wild-type plants under normal growth conditions. Eight weeks after germination, the average number of sympodial units under the first inflorescence was about 8 in the SlCOL1-RNAi lines, as compared to 12 in the WT plants. The number of flowers and fruit yield were also increased in the SlCOL1-RNAi lines. On the other hand, SlCOL1-OE lines produced about 16 leaves in the primary shoot which was significantly more than that of the WT plants. The fruits of the SlCOL1-OE lines were scattered on the inflorescences (Fig. 3A, C). The numbers of flowers and fruits over the entire growth season was also examined. The transgenic plants produced 46-130 flowers and 1-40 fruits in the SlCOL1-RNAi lines and 14-66 flowers and 0-18 fruits in the SlCOL1-OE lines. In contrast, WT plants produced 50-110 flowers and 1-30 fruits (Fig. 3D, E). The total yield of fruit was increased approximately 37% in the SlCOL1-RNAi lines and reduced approximately 42% in the SlCOL1-OE lines as compared to the fruit yield in the WT plants (Fig. 3B). These results illustrated that SlCOL1 plays a major role in the regulation of flowering time, flower and fruit number and yield in tomato.

SlCOL1 represses the expression of the flowering gene SFT
Tomato sft mutant inflorescences revert to indeterminate vegetative branches or become a single fertile flower, thus, the mutant plant has much fewer flowers and a much lower fruit yield than WT [6,7]. The tomato SlCOL1 overexpression transgenic and sft mutant plants were phenotypically similar. Their flower and fruit numbers were significantly reduced (Fig. 3D, E), and as a result, their fruit yield was reduced as well ( Fig. 3A-B). CONSTANS activates FT transcription through binding to the CO-responsive (CORE, CCACA) and CCAAT-box elements in the FT gene promoter in Arabidopsis [8,[21][22][23]. Gene expression analysis revealed that the SFT (Solyc03g063100) transcripts were lower in the SlCOL1-OE lines but higher in the SlCOL1-RNAi lines when compared with those in the WT plants (Fig. S1B). Thus, SlCOL1 gene appears to serve as the core transcription factor that regulates the expression of the SFT gene in tomato.

SlCOL1 negatively regulates SFT expression by directly binding to the regulatory cis-elements in the SFT promoter
We searched the 2.5 kb promoter region of SFT and found three CCACA and four CCAAT sequences (Fig. 4A). To examine if SlCOL1 could bind to these cis-elements and drive gene expression, we first selected five SFT promoter fragments that contained different combinations of the conserved cis-DNA elements, including SFT1 (no ciselement), SFT2/3/4 (different combinations of the two cis-elements) and SFT5 (only one CCACA cis-element) (Fig. 4B). Three constructs (SFT2/3/4) were found to confer the antibiotic resistance in the presence of 10-20 mM AbA when SlCOL1 was co-expressed. In contrast, the constructs that contained no cis-element (SFT1) or only one cis-element (SFT5) could not rescue the yeast cell growth in the presence of 10 mM AbA (Fig. 4B), suggesting that the cis-elements of the SFT promoter were required for the SlCOL1 transcription factor to drive the resistance gene expression, and one cis-element (CCACA) was not sufficient to allow the resistance gene expression in this Y1H system. In order to examine the minimum cis-elements that were required for the AbA resistance gene expression, we selected six SFT promoter fragments that contained either the CCACA motif (in SFT2-1 and SFT3-2) or the CCAAT element (in SFT2-2, SFT3-1, SFT4-1, and SFT4-2) from SFT2, SFT3, and SFT4 (Fig. S2). Four constructs (SFT2-2, SFT3-1, SFT4-1, and SFT4-2) were found to confer the antibiotic resistance in the presence of 10 mM AbA, when SlCOL1 was co-expressed. In contrast, the constructs that contained the CCACA motif alone (in SFT2-1 and SFT3-2) could not rescue the yeast cell growth in the presence of 10 mM AbA (Fig. S2). These results suggest that the CCAAT ciselement was necessary and sufficient for the binding of the SlCOL1 transcription factor to the SFT promoter.
To test whether SlCOL1 could regulate the expression of the SFT gene in planta, we co-expressed 35S-SlCOL1 and ProSFT-GUS constructs in tobacco leaves. Our result showed that the GUS reporter (ProSFT-GUS) alone was able to express in tobacco leaves (Fig. S3), suggesting that the host cells had endogenous transcription factors that could drive the reporter gene expression. This background level of the reporter expression was not affected by the co-expression of the empty vector of pHELLS-GATE8 in tobacco leaves (Gate8 + ProSFT-GUS, Fig. 4C left). However, when the 35S-SlCOL1 construct was used to replace the empty pHELLSGATE8 vector, the GUS staining became much weaker (SlCOL1 + ProSFT-GUS, Fig. 4C middle), suggesting that co-expression of SlCOL1 repressed the GUS expression driven by the SFT gene promoter. These results indicated that SlCOL1 acts as a transcriptional repressor of the SFT gene in planta.

SlCOL1 negatively regulates fruit size in tomato
The total yield of fruit was significantly increased in the SlCOL1-RNAi lines and reduced in the SlCOL1-OE lines as compared to the fruit yield in the WT plants (Fig. 3B). In addition to the reduction in the number of fruits, we found the average of fruit weight was 27 to 31% higher in the RNAi lines and 18 to 26% lower in the overexpression lines than the WT plants ( Fig. 5A-B). The length and diameter of the fruits were also reduced in the SlCOL1-OE plants and increased in the SlCOL1-RNAi plants relative to the WT (Fig. 5C). These results illustrated that SlCOL1 also plays a major role in the regulation of fruit size in tomato.

SlBBX24 functions to regulate tomato fruit size
The fruit size was significantly increased in tomato sft mutant plants as compared with WT [7]. In the SlCOL1-OE plants, the fruit size was significantly decreased (Fig. 5). Therefore, we believe that SlCOL1 regulates fruit size in tomato not directly through regulating SFT gene expression. Tomato SlBBX24 (Solyc06g073180), as a CONSTANS-LIKE protein, has two B-BOX domains and interacts with SlCOL1 [19,24]. For these reasons, we tested whether SlBBX24 may regulate tomato fruit development. SlBBX24 expression was determined using real-time RT-PCR on total RNA extracted from various tomato organs. The transcripts of SlBBX24 were detected in all tissues tested, with the highest expression level in leaves and flowers, which is similar to the expression patterns of SlCOL1 (Fig. S4). To better understand the function of SlBBX24, we next generated SlBBX24 overexpression (OE) and CRISPR/ cas9 (CR) tomato lines. Three lines of each transformation experiment were selected for further analysis, including OE-2, OE-3 and OE-9 from the overexpression lines and CR-1, CR-2 and CR-3 from the CRISPR/ cas9 transformation. We found that the SlBBX24 gene could affect the fruit size based on our transgenic functional analysis (Figs. 6A, S5). Overexpression of the SlBBX24 significantly reduced the fruit size (Fig. 6). However, no significant phenotype in fruit size and other plant morphological traits was observed in the three CR-slbbx24 lines as compared to those in the WT plants. We tested the fruit weight and fruit length and diameter of transgenic lines and WT plants and found that fruits of the overexpression lines were smaller than those of the WT plants ( Fig. 6B-C). These results suggest that SlBBX24 may interact with SlCOL1 to form a heterodimer of transcription factor and plays a role to regulate fruit size in tomato.
In this work, we found that the transcript levels of SlCOL1 and SFT were not affected in SlBBX24-OE lines as compared with their expression levels in WT (Fig. S6). We also found that the transcript levels of SlBBX24 were not affected in SlCOL1-RNAi lines as compared with that in WT (Fig. S7, right). These results showed that SlBBX24 may not regulate flowering time in tomato, but participates in fruit size regulation by interacting with SlCOL1.

Discussion
The BBX transcription factor family is known to be involved in a wide range of cellular processes, including the resistance to abiotic stresses [25][26][27], control of the Asterisks indicate statistically significant differences relative to the wild type as determined using t-tests. **, P < 0.01 circadian clock [28] and regulation of flowering time [17,29]. Several BBX genes have been shown to play key roles in the regulation of flowering time and flower development in different plant species. Within a plant species, several BBX genes are known to participate in flowering regulation through different mechanisms [12,15,16,19,30,31]. As the first identified BBX protein, CONSTANS is known to activate FT transcription through binding to the CORE (CCACA) and CCAAT-box cis-elements in the FT promoter in Arabidopsis [8,[21][22][23]31]. Hd1, the rice ortholog of CO, also regulates the expression of Hd3a, the rice ortholog of Arabidopsis FT, by binding to the CORE (CCACA) DNA element of the Hd3a promoter [12,[32][33][34]. In this study, we demonstrated that tomato SlCOL1 regulates flower time, flower number and yield by binding to the SFT gene promoter, repressing its expression (Figs. 3, 4 and S1B). These results illustrate that CO and its orthologs play conserved roles in flowering regulation through binding to the FT promoter to regulate its expression.
CO and its homologs have been shown to regulate the expression of downstream target genes by modulation of DNA methylation. In Arabidopsis, overexpressing CO can change the chromatin status in the FT locus, such as a decrease in binding of LIKE HETEROCHRO-MATIN PROTEIN1 (LHP1) and an increase in the acetylation of H3K9 and K14 [31]. In addition, Nuclear Factor-Y (NF-Y) can interact with CO to modulate H3K27me3 levels of the SOC1 promoter and regulate the transcription of SOC1 in Arabidopsis [35]. In rice, the DTH8 (NF-YB) transcription factor plays a critical role in mediating the Hd1 regulation of Hd3a transcription in photoperiodic flowering through its interaction with Hd1 to shape epigenetic marks. The DTH8-Hd1 module enhances H3K27 trimethylation at Hd3a and represses Hd3a expression in long day conditions, but reduces the H3K27me3 levels at Hd3a and enhances Hd3a expression in short day conditions [12]. In our previous study, we have illustrated that NF-YBs bind to the CCAAT element of the CHS1 promoter and regulate the levels of H3K27me3 at the CHS1 locus during tomato fruit ripening. Suppression of the expression of NF-YB significantly reduces the expression level of CHS1 and leads to the development of pink-colored fruits with colorless peels [36]. Previous studies have revealed that CONSTANS may replace NF-YA in the NF-Y complex to form a trimeric CO/NF-YB/ NF-YC complex [19,21]. Therefore, we hypothesized that SlCOL1 represses the expression of SFT possibly through regulating the levels of H3K27me3 at the SFT promoter by interacting with the NF-Y complex.
The BBX gene family comprises 29 members in tomato and can be divided into five structural groups based on the number and sequence features of the B-box domain and the presence or absence of a CCT domain [24]. In this study, we found that down-regulation of the expression of SlCOL1 by RNAi led to drastic phenotypes of flower development, while knocking out SlCOL1 by CRISPR/cas9 did not display any visible phenotype in plant growth and reproduction (Fig. S8). This implies that there could be redundancy in the BBX genes. Sequence analysis indicated that SlCOL2 (Solyc02g089500) and SlCOL3 (Solyc02g089520) share high similarities with SlCOL1, and they are grouped into the same branch in the BBX family. We assume that SlCOL2 and SlCOL3 may play redundant roles with SlCOL1 in the regulation of flowering time and fruit yield. In fact, the expression levels of SlCOL2 and SlCOL3 were both reduced in the SlCOL1-RNAi lines (Fig. S1A).
Our previous studies have shown that overexpression of SlBBX20 results in transgenic tomato plants with smaller leaves and plant size as compared with those of the WT plants [37]. The fruit size has also been found to be reduced in the SlBBX20-OE lines. This implies that BBX genes from different groups of the BBX gene family may play a similar function in regulating organ size in tomato. In the present work, the SlBBX24 gene was shown to regulate the tomato fruit size as well. Moreover, the fruit size of SlBBX24 overexpression lines was found to be smaller than that of WT (Fig. 6). SlBBX20 and SlBBX24 genes belong to the same branch in the BBX family [24]. It is interesting to point out that SlBBX20 and SlBBX24 are not grouped to the same branch with SlCOL1 in the BBX family [24]. In this work, the transcript levels of SlBBX20 and SlBBX24 were not affected in SlCOL1-RNAi lines (Fig. S7, left). These results imply that SlBBX20 and SlBBX24 may exert their biological functions in the regulation of fruit size through interacting with SlCOL1 or through regulating the expression of other genes. It is also likely that SlCOL1, SlCOL2, SlCOL3, SlBBX20 and SlBBX24 all play a role, either uniquely or redundantly, in regulating fruit size in tomato.

Conclusion
Based on our findings, we propose a model in which SlCOL1 controls the tomato yield traits by regulating the expression of SFT, and regulates tomato fruit size by modulating the expression levels of downstream genes (Fig. 7). There are at least two distinct pathways: SlCOL1 may act as a transcriptional repressor that controls the production of fruit by down-regulating SFT expression (Fig. 7); and SlBBX24 and SlCOL1 may control tomato fruit size by regulating the expression levels of downstream genes (Fig. 7). Thus, the fine tuning of the expression of SlCOL1 will have the potential for improving tomato fruit yield (fruit number and size) and a better understanding of this pathway may eventually lead to similar genetic improvements in other crops.

Plant materials and growth conditions
The tomato (Solanum lycopersicum) variety Ailsa Craig (AC, LA2383A) was used as the wild-type (WT) control and for genetic transformation experiments in this study. The seeds of AC were originally obtained from the Tomato Genetics Resource Center, UC Davis, USA

RNA isolation and gene expression analysis
Total RNA was extracted from various tissues of the transgenic lines or WT plants using the TRIZOL reagent (Invitrogen, USA). Complementary DNAs (cDNAs) were synthesized using an M-MLV reverse transcriptase kit (Toyobo, Japan). The LightCycler480 SYBR Green I Master Kit (Roche Applied Sciences, Germany) was used for qPCR analysis. Three biological replicates from each genotype were carried out and analyzed for statistical differences. The Actin gene (BT013524, Solyc11g005330) was used as the internal control. The primer sequences used in real-time PCR are listed in Table S1.

Vectors constructs and tomato transformation
The full-length ORF and RNAi fragments for SlCOL1 and SlBBX24 were amplified from tomato cDNA using the KOD-Plus DNA polymerase (Toyobo, Japan) and cloned into the effector vector pHELLSGATE8. The CRISPR/ cas9 (PTX041) vector targeted two sites in the first exon of the ORF of SlBBX24 were designed at CRISPR-PLANT (http:// www. genome. arizo na. edu/ crispr/ CRISPR search. html). The sequences of primers used in these experiments are listed in Table S1. The vectors were introduced into the Agrobacterium tumefaciens strain C58. This strain was used for plant transformation in tomato Ailsa Craig (AC) as described previously [38]. Genomic DNA was extracted from transgenic plants using the CTAB method as described by Murray and Thompson (1980). The genomic DNA was analyzed using PCR-based markers to identify transgenic plants. The transgenic materials (SlCOL1-OE, SlCOL1-RNAi, CR-bbx24 and SlBBX24-OE) have been deposited in Key Laboratory of Horticultural Plant Biology, Ministry of Education, Huazhong Agricultural University (Hubei, China).

Yeast one-hybrid assay
The yeast one-hybrid (Y1H) assay was used to test whether COL1 could bind to the SFT promoter. The fulllength SlCOL1 ORF sequence was amplified from tomato cDNA and cloned into pGADT7 (Clontech). Five promoter fragments (− 2528 to − 2246 bp, − 289 to − 0 bp, − 2087 to − 1791 bp, − 862 to − 501 and − 519 to − 0 bp relative to the translation initiation codon of the SFT) were amplified from tomato genomic DNA and cloned into pAbAi (Clontech). The transformed yeast strains were picked and diluted in 0.9% NaCl to an OD 600 of 0.1, and 2 μL of the suspension was spotted on a SD/−Leu medium, with or without aureobasidin A (AbA, Clontech). The plates were incubated for 3 to 7 days in an incubator at 30 °C.

GUS staining
For GUS staining in tobacco, the full-length ORF of SlCOL1 and was amplified and cloned into the effector vector pHELLSGATE8. The cauliflower mosaic virus (CaMV) 35S promoter was used to drive gene expression in pHELLSGATE8 vector. The 2.53-kb promoter region of SFT was amplified and cloned into the effector vector pHELLSGATE8 (with the GUS gene, but without the 35S promoter). A. tumefaciens GV2260 was separately transformed with the effector and reporter vectors. For GUS staining in tomato, a DNA fragment of 3013 bp from the SlCOL1 promoter region was amplified by PCR and cloned into the effector vector pHELLSGATE8 (with the GUS gene, but without the CaMV35S promoter). Agrobacterium tumefaciens strain C58 was transformed with the vector. This strain was used for plant transformation in tomato Ailsa Craig (AC) as described previously [38]. Transgenic tomato seedlings, floral buds, and developing fruits at different stages were selected for GUS staining. The selected seedlings and tissues were incubated at 37 °C for 24 h in staining buffer (100 mM sodium phosphate, pH 7, 0.1% Triton X-100, 0.1% N-laurylsarcosine, 10 mM Na 2 EDTA, 1 mM K 3 Fe (CN) 6 , 1 mM K 4 Fe (CN) 6 , and 0.5 mg mL − 1 5-bromo-4-chloro-3-indolyl-β-D-glucuronic acid), followed by washing with 70% (v/v) ethanol. The expression of the GUS gene was quantified using qRT-PCR. All primers used for the construction of the vectors are listed in Table S1.

Transient expression in tobacco protoplasts and microscopy
The SlCOL1 CDS without the stop codon was amplified by PCR and fused to the 5′ end of the open reading frame encoding GFP in pCAMBIA 1302, which uses the CaMV 35S promoter to drive gene expression, generating 35S:SlCOL1-GFP. Ghd7-CFP was used as the marker for the nucleus. Tabaco leaf protoplasts were prepared and transient transcriptional activation was assayed as described previously [39]. Fluorescence from the transformed protoplasts was imaged using a confocal laser scanning microscope (Leica TCS SP2). The pertinent primer sequences are listed in Table S1.

Statistical analysis
Statistical analyses were conducted using SigmaPlot, Excel and the SPSS (IBM, SPSS 22) software. Comparisons between pairs of the groups were performed using the Student's t-test. Statistically significant differences were categorized into two groups: P < 0.05 and P < 0.01.