The full-length CBF3 complementary DNA (cDNA) of Arabidopsis (AtCBF3) was firstly cloned into the pMD18-T vector (TaKaRa Biotechnology, Dalian, China). The released AtCBF3 fragment was then cleaved by restriction endonucleases, KpnI and PstI, and inserted into the multiple cloning sites between CaMV 35S promoter and NOS terminator of a modified pCAMBIA1301 to generate the binary vector pC35S-AtCBF3 (Supplementary Figure 1). The construct was mobilized into Agrobacterium tumefaciens LBA4404, which was used for cassava transformation by friable embryogenic callus-mediated method (Zhang and Gruissem, 2004).

Genomic DNA was isolated from mature leaves of transgenic and wild-type (WT) cassava plants using the CTAB DNA extraction method. Twenty μg of total DNA was digested overnight with EcoRI. The fragmented product was separated on a 0.8% agarose gel and transferred onto a Hybond-N⁺ membrane (Amersham Biosciences, Piscataway, NJ, USA). The blot was hybridized with the DIG-labeled hygromycin phosphotransferase (HPT) gene probe and exposed to X-ray film for signal detection. The HPT probe was amplified by PCR using the following primers: 5′-TGAAAAAGCCTGAACTCACCG-3′ (forward) and 5′-TATTCTTTGCCCTCGGACG-3′ (reverse). The PCR program was as follows: 3 min at 94°C; 28 cycles of 30 s at 94°C, 30 s at 55°C, and 40 s at 72°C; followed by 10 min at 72°C; and finally stored at 4°C. DNA probe preparation, hybridization, and membrane washing were performed using DIG High Prime DNA Labeling and Detection Starter Kit II according to the manufacturer’s instructions (Roche, Grenzacherstrasse, Basel, Switzerland).

One-month-old in vitro seedlings of AtCBF3 overexpressing lines (OE) and WT were planted into soil pots and grown in the greenhouse at 30°C/25°C with 16 h/8 h light/dark cycle and 30–40% relative humidity. For cold stress, 1-month-old plants grown in the greenhouse were exposed to 0°C for 24 h and then recovered for 1 month. For drought stress, stem-propagated pot-grown plants were cultivated in the greenhouse for 3 months under a fully watered regime (approximately 20% water content) and then the water was withdrawn for 21 days until the plants showed distinct symptoms of injury.

For field-grown cassava, about 6-month-old plants were subjected to seasonal changes. At the temperature below 15°C/8°C for day/night in November, the WT plants began to exhibit cold injury. During field growth, six different tissues of transgenic line OE-3 and WT, including apical buds, immature leaves, mature leaves, old leaves, petioles, and young stems, were harvested under three conditions of different day/night temperatures (September: 33°C/25°C as normal temperature; October: 25°C/16°C as cold treatment I, November: 15°C/8°C as cold treatment II), and frozen in liquid nitrogen for RNA extraction. Each treatment had more than three plants per line, and all treatments were conducted in three biological replicates.

One-month-old transgenic OE lines and WT plants were divided into two groups (normal growth and cold treatment), each group containing about 12 plants and each independent line consisting of at least three plants. After the plants had been exposed to 4°C for 12 and 24 h, approximately 0.5 g leaf tissue was harvested for each sample and extracted with various reagents according to different assays. Proline concentrations were measured by the sulfosalicylic acid-acid ninhydrin method (Bates et al., 1973). Malondialdehyde (MDA) equivalent was measured as described previously (Hodges et al., 1999). Electrolyte leakage was assessed following the method by Gilmour et al. (2000).

The RWC of cassava plants under normal growth conditions and drought stress was measured according to Gaxiola et al. (2001). The drought tolerance of these plants was evaluated by growing the plants for 3 months under a fully watered regime and then depleting water for about 3 weeks until these plants displayed obvious injury symptoms. The leaves from OE lines and WT were excised and their fresh weights recorded immediately. Their rehydrated weight was determined after floating the leaves in deionized water at 4°C overnight. Finally, they were dried in an oven at 70°C overnight and weighed. Each measurement was repeated three times. The RWC was calculated according to the formula: RWC = (fresh weight – dry weight)/(rehydrated weight – dry weight) × 100%.

Total RNA was extracted from leaves of 1-month-old cassava plants grown under normal conditions or cold stress (4°C for 12 days) using TRIzol reagent according to the manufacturer’s protocol (Invitrogen). The RNA quality was determined on an agarose gel by ethidium bromide staining. Reverse transcription was performed using the ReverTra Ace qPCR RT Kit according to the manufacturer’s protocol (TOYOBO, Osaka, Japan). The cDNA was directly used as templates for semi-quantitative RT-PCR or diluted fivefold for real-time RT-PCR amplification on a Bio-Rad CFX96 thermocycler (CFX96 thermocycler (Hercules, CA, USA) with gene-specific primers (Supplementary Table 1). The quantitative variation between different samples was evaluated by the 2^-ΔΔCt method, and cassava β-actin was used as an internal control. All of the samples were measured in triplicate, and the experiments were performed on three biological replicates.

All data were determined from at least three biological replicates and are presented as mean ± SD. Asterisks indicate significant differences compared with the WT at ^∗P < 0.05 or ^∗∗P < 0.01 by Student’s t-test.

Six AtCBF3 overexpressing (OE) transgenic cassava plant lines regenerated from transformed friable embryogenic calli were multiplied by shoot cultures in vitro. Their transgene integration patterns were verified by the DIG-labeled HPT probe using Southern blotting analysis. Four out of six OE lines were identified as single copy insertion into the genome by EcoRI digestion (Figure 1A). Their AtCBF3 transcript level initially analyzed by semi-quantitative RT-PCR confirmed the high expression of the genes in three lines, OE-1, OE-3, and OE-5 (Supplementary Figure 2). OE-3 and OE-5 lines showed a remarkable overexpression of the CBF3 gene, approximately several 1000-fold higher than that of the OE-1 line, by real-time RT-PCR analysis (Figure 1B). Therefore, these three lines together with WT were used for subsequent studies.

Similar to the most transgenic plant lines overexpressing CBF subfamily members that showed undesired side-effects such as severe growth retardation, the OE cassava lines also displayed downward leaf curling, especially in OE-3 and OE-5 (Figure 1C), and severe retardation under normal growth conditions in the field (Figure 1D). Plant height and internode length of the OE lines were significantly shorter than those of WT (Figure 1E), about 50%. Upon harvest, their phenotypes of storage roots were photographed (Figure 2A). Noticeably, the storage root number and yield of OE-3 and OE-5 were reduced to 30–50% of WT (Figures 2B,C). The OE-1 plants showed less change in these traits. These phenotypic changes were obviously dependent on the transcript abundance of the AtCBF3 gene.

To investigate whether overexpression of AtCBF3 confers abiotic stress tolerance in cassava, 1-month-old OE and WT seedling plants grown in the greenhouse were exposed to cold stress (0°C) for 24 h, followed by recovery of the growth under standard conditions (25°C). The OE transgenic plants recuperated better than the WT (Figure 3A). In particular, after 1 month, the OE-3 and OE-5 lines continued with a vigorous growth, including sprouting new leaves; while OE-1 plants only survived. Contrastingly, WT plants could not be recovered and finally perished.

In the field, under the seasonal transition from autumn to early winter, the phenotypic exhibition of these plants was also different. When the temperature decreased to 15°C/8°C (day/night) at the beginning of November, WT plants displayed obvious cold injury symptoms. Their apical shoots including apical buds and immature leaves were dehydrated and yellowish with a loss of vitality. Moreover, the OE-3 and OE-5 lines showed normal foliage status as before (Figure 3B). The shoots of OE-1 plants were also affected by the low temperature but less damaged as compared to those of WT. The observation indicates that the high overexpression of AtCBF3 in cassava plants increases their tolerance to low-temperature stress in a dose-dependent manner.

Under the normal condition, the content of proline in seedlings of OE lines was similar to that of WT. A significant increase of proline content was detected in the OE lines than that of WT under -2°C stress (Figure 4A, upper panel) with a maximum of 52.9% increase at 12 h and 57% at 24 h. Strikingly, the MDA contents were always lower in OE lines than that in WT. Upon treatment, the minimum level of MDA in OE lines was about 26.9% that of WT; after 24 h treatment, the minimum level of MDA in OE lines only reached about 20.2% that of WT (Figure 4A, lower panel). Similarly, the electrolyte leakage in the leaf disks was also less in OE lines than WT: about half of WT after -5°C treatment (Figure 4B). These results indicate that the overexpression of AtCBF3 protects the cassava with less free radical production and from cellular membrane injury caused by cold stress.

In order to speculate whether the AtCBF3 improves the abiotic stress tolerance of transgenic cassava through CBF cold-responsive pathway and the endogenous CBF-regulon response to stress, the transcript profiles of potential CBF-targeted genes (Table 1) were analyzed. These profiles were assessed using the apical bud, young leaf, mature leaf, old leaf, petiole, and young stem of field-grown WT and OE-3 plants under three different day/night temperatures (September 33°C/25°C as normal temperature, T1; October 25°C/16°C as cold treatment I, T2; November 15°C/8°C as cold treatment II, T3). The homologous genes in the cassava genome sequence from Phytozome¹ were discovered by protein sequence alignment with Arabidopsis stress-related genes. Several homologous genes, including MeRAP2.1, MeERD7, MeCOR47, MeKIN2, MeGolS3, MeLTI30, MeLEA14, MeOEP16, MeCOR78, and MeATDI21, were identified and their promoter regions ca. 2000 bp length were cloned. Six out of 10 genes belonged to the CBF-target genes as they harbored the CBF binding element DRE/CRT (G/ACCGAC, Yamaguchi-Shinozaki and Shinozaki, 1994; Stockinger et al., 1997). The other three genes, whose promoters did not contain this element, might be regulated by CBF3 via other cis-elements or other regulators mediated by CBF3. However, the promoter of MeCOR78, a COR gene, was not found (Table 1).

Compared to WT, all tissues of the OE-3 line detected an increased expression of CBF3, the highest in young leaf, petiole, and young stem, whereas the apical bud showed the lowest expression of the gene (Figure 5). Interestingly, all the stress-related genes exhibited little expression changes in the apical bud of both WT and OE-3 plants indicating an unknown upstream factor that might be inhibiting the expression induction of these stress-response genes in cassava meristematic tissues. In other tissues, most of the genes showed upregulated expression by the cold treatment I (T2) and maintained similar levels as WT by cold treatment II (T3) in the OE-3 line. Nevertheless, specific response of these genes to cold treatment varies among different tissues. For example, in the young and mature leaf of OE-3 line, the expression of MeRAP2.1, MeERD7, MeCOR47, and MeKIN2 was obviously induced by the cold treatment I (T2); however, the expression of these genes was less affected by the treatment in WT. The expression of MeGolS3 only changed in the young stem by T2. Several genes whose promoters lacked the DRE/CRT element, especially in MeERD7, MeLTI30, and MeCOR78 showed less difference in expression between the WT and OE-3 plants under the normal or cold treatments. The expression of other genes such as MeLEA14 and MeATDI21, containing the DRE/CRT element in their promoters was not affected by the overexpression of AtCBF3 (Figure 5), indicating the variations between Arabidopsis and cassava CBF regulons. Taken together, cassava CBF regulon differed from that of Arabidopsis and its regulation showed temporal and spatial patterns in response to heterologous CBF and cold treatment.

Since cold treatment promoted excessive proline accumulation in OE lines (Figure 4A), the expression profiles of P5CS (Δ1-pyrroline-5-carboxylate synthase) and ProDH (proline dehydrogenase), two key genes controlling the balance of biosynthetic and catabolic pathways (Sharma et al., 2011), were also checked in these different tissues. Upregulation of MeP5CS and downregulation of MeProDH were detected in the transgenic OE-3 line compared with those of WT under normal and cold treatment conditions (Figure 6). This suggested that ectopic overexpression of AtCBF3 might favor the modulation of stress-related gene expression in response to abiotic stress in transgenic cassava.

To examine the drought stress tolerance of OE transgenic cassava, the 3-month-old plants of OE lines and WT grown in pots were subjected to drought treatment by water depletion. Before drought stress, all the plants exhibited active growth status (Figure 7A). After water depletion for 3 weeks, WT plants showed distinct injury phenotype including severe wilting and dehydrated leaves. Contrastingly, the OE plants consisted of more green leaves and relatively less dehydration phenotype (Figure 7A), especially in OE-3 and OE-5 plants. Their RWC was dramatically higher than WT, about ninefold after the drought treatment, although no difference was recorded between the OE and WT plants under normal growth conditions (Figure 7B). Furthermore, the electrolyte leakage of leaves in OE plants was reduced significantly than that in WT under normal growth conditions. Under drought stress, WT showed approximately 90% electrolyte leakage, while that of OE lines was only 15–40% (Figure 7C). The phenotype indicates that the integrity of the membrane was dramatically injured in WT but preserved in OE lines. These results demonstrate that the overexpression of CBF3 resulted in improved drought stress tolerance in transgenic cassava plants.

As illustrated in Figure 3B, the absence of anthocyanin accumulation was noticed in the OE plants, especially OE-3 and OE-5. In cassava TMS60444, apical buds, young leaves, and the base region of petioles show the purple color of anthocyanins. A quantitative analysis of anthocyanins from apical buds and young leaves demonstrated a significant decrease of anthocyanins in the OE lines (Figure 8A). In their top leaves, the expression level of several anthocyanin biosynthetic genes in cassava was substantially decreased, including 4CL, CHS, CHI, and ANS. In addition, only the BZ1 gene showed an upregulated expression, possibly due to the feedback regulation (Figure 8B). In response to the downregulated expression of these genes, the expression of several TFs, such as MYB, MYC, WD-40-like that have been reported to regulate anthocyanin biosynthesis (Petroni and Tonelli, 2011), was also analyzed. Among these TFs, MePAP2 and MeTTG1 showed about 50% reduction in their expression (Figure 8C), although their promoter regions did not harbor the DRE/CRT element (data not shown).