The expression and crystallization of Cry65Aa require two C-termini, revealing a novel evolutionary strategy of Bacillus thuringiensis Cry proteins

The insecticidal crystal protein (Cry) genes of Bacillus thuringiensis are a key gene resource for generating transgenic crops with pest resistance. However, many cry genes cannot be expressed or form crystals in mother cells. Here, we report a novel Cry protein gene, cry65Aa1, which exists in an operon that contains a downstream gene encoding a hypothetical protein ORF2. We demonstrated that ORF2 is required for Cry65Aa1 expression and crystallization by function as a C-terminal crystallization domain. The orf2 sequence is also required for Cry65Aa expression, because orf2 transcripts have a stabilizing effect on cry65Aa1 transcripts. Furthermore, we found that the crystallization of Cry65Aa1 required the Cry65Aa1 C-terminus in addition to ORF2 or a typical Cry protein C-terminal region. Finally, we showed that Cry65Aa1 has a selective cytotoxic effect on MDA-MB231 cancer cells. This report is the first description of a 130-kDa mass range Cry protein requiring two C-termini for crystallization. Our findings reveal a novel evolutionary strategy of Cry proteins and provide an explanation for the existence of Cry protein genes that cannot form crystals in B. thuringiensis. This study also provides a potential framework for isolating novel cry genes from “no crystal” B. thuringiensis strains.

B acillus thuringiensis is a Gram-positive, spore-forming soil bacterium that can produce parasporal crystals during the sporulation phase. These parasporal crystals consist of proteins (Cry) that exhibit specific toxicity against a variety of insects, such as Lepidoptera, Coleoptera, and Diptera, and against some nematodes, mites, and protozoa 1,2 . Because of its strong and specific toxicity toward a wide range of insects, B. thuringiensis has been developed as a biopesticide and is the leading biopesticide for using as an alternative or supplement to synthetic chemical pesticides 3 . Additionally, the cry genes of B. thuringiensis are also considered to be a key gene resource for generating transgenic crops with pest resistance 4 .
However, because of the extensive and continuous usage of B. thuringiensis-based pesticides, some insect populations have developed resistance to B. thuringiensis [5][6][7] . Several strategies, such as the use of multiple toxins 8 , spatial or temporal refuge 9 , and high or ultrahigh doses 10 , are employed to delay the development of insect resistance to B. thuringiensis-based products. A major principle of management of resistance to insecticidal proteins is the use of combinations of different Cry proteins, especially proteins that have different receptors or different modes of action 11 . Therefore, the discovery of novel holotype crystal proteins in B. thuringiensis is considered to be one of the major approaches to overcoming potential insect resistance problems 1,11 . A series of techniques have been utilized to isolate novel cry genes, such as PCR hybridization 12 , PCR-restriction fragment length polymorphism (RFLP) 13 , and PCR product high-resolution melting (HRM) analysis 14 . The construction of B. thuringiensis DNA libraries in Escherichia coli and then screening by western blotting 15 or hybridization 16 has also been widely used for new cry gene isolation. Additionally, the development of DNA libraries in a crystalliferous mutant of B. thuringiensis, followed by microscopy observations, has been used to detect novel Cry protein genes 17 . Recently, next-generation sequencing technology has also been employed for the discovery of new cry genes 18 .
Using the above approaches, increasing numbers of crystal protein genes have been identified. Thus far, more than 700 cry genes from 73 classes have been cloned (http://www.lifesci.sussex.ac.uk/Home/Neil_Crickmore/Bt/). However, of the reported cry genes, some cannot be expressed or form crystals in B. thuringiensis 13,[19][20][21] . As next-generation sequencing technology has been increasingly employed for the discovery of new cry genes, this kind of cry genes have been found more frequently 17,21 . Previously, we constructed a highthroughput system for isolating new cry genes by combining nextgeneration sequencing and a computational pipeline called BtToxin_scanner 21 . Using this method, we identified a large number of new cry genes. However, only a portion of these cry genes can be expressed in B. thuringiensis.
Why these cry genes cannot be expressed and form crystals in B. thuringiensis are not clear. Here, we report a novel 130-kDa mass range crystal protein gene, cry65Aa1, which was not expressed when introduced alone. We then found that cry65Aa1 exists in an operon that encodes a hypothetical protein, ORF2, followed the Cry65Aa1 ORF. By contrast, cry65Aa1 was expressed when co-expressed with orf2. We further demonstrated that ORF2 is required for Cry65Aa1 expression and crystallization and functions as an mRNA-stabilizing effector and a C-terminal crystallization domain. In addition, Nterminal or C-terminal replacement experiments demonstrated that the expression and crystallization of Cry65Aa1 require the Cry65Aa1 C-terminus and ORF2.

Results
Identification and cloning of a novel holotype crystal protein gene, cry65Aa1. B. thuringiensis (strain SBT-003) has unique plasmid pattern and crystal morphology, so we finished its genome sequence to find novel cry genes (Genome accession No: AMYJ00000000). A novel cry sequence was identified in the SBT-003 genome sequence by using BtToxin_scanner 21 . This gene encodes a polypeptide of 1064 amino acid residues with a predicted molecular weight of 118,156.4 Da (Fig. S1) and has a typical three-domain-type structure with N-and C-terminal halves reminiscent of the 130-kDa mass range Cry proteins. Domain I has high similarity to the Endotoxin_N domain [pfam03945], and domain III has high similarity to the Delta_endotoxin_C domain [cd04085] (Fig. S2). The C-terminal domain of Cry65Aa1 consists of only 333 amino acids (with a molecular weight of 36.7 kDa), which is shorter than a typical C-terminal domain (approximately 55-65 kDa) (Fig. 1A). Sequence analysis indicates that the Cry65Aa1 C-terminal domain has low homology with a hypothetical protein (GenBank ID: EKU79138) from Veillonella ratti (Fig. S3).
Sequence alignment of the Cry65Aa1 protein with known Cry proteins revealed similarities within the 0-45% range. Therefore, the B. thuringiensis Nomenclature Committee designated the protein as a novel holotype Cry protein, Cry65Aa1. Cry65Aa1 exhibited the greatest similarity (28%) to Cry41Aa1 (accession No: AB116649) (Fig. S4), a parasporin protein (PS3Aa1) that exhibits cytocidal activity toward some cancer cells 22 . In the parasporin dendrogram, Cry65Aa1 proteins appeared grouped in a separate branch (Fig. 1B); thus, we consider the Cry65Aa1 to be a novel class of parasporin proteins.
To express Cry65Aa1, its coding gene was cloned into an E. coli-Bt shuttle expression vector, pBMB1A 23 , to generate the recombinant plasmid pBMB1A-65A and was then introduced into an acrystalliferous B. thuringiensis strain, BMB171. When sporulated cells were examined by phase-contrast microscopy, no crystalline inclusions were produced by BMB171/pBMB1A-65A (Fig. 1D, panel a). SDS-PAGE analysis also revealed no band corresponding to the predicted molecular weight of Cry65Aa1 (Fig. 1F, lane 1).
Why Cry65Aa1 was not expressed and did not form crystals is unclear. When analyzing the SBT-003 genome, we found ORF2, which was annotated as a hypothetical protein located downstream of cry65Aa1, and a typical terminator located downstream of orf2 (Fig. 1C). Interestingly, sequence alignment revealed that ORF2 shared the highest level of identity with the C-terminal regions of Cry21Ba (46%), Cry21Fa (46%), Cry21Ga (45%), Cry12Aa (37%), and Cry5Ab (38%) (Fig. S5). With respect to cry operons, five reported cases (cry10Aa, cry19Aa, cry30Ca, cry39Aa, and cry40Aa) have gene organizations similar to the cry65Aa operon 24,25 . In these five similar operons, the upstream reading frames encode an N-terminal domain or a 65-kDa range Cry protein. However, Cry65Aa1 is a three-domain Cry protein and possesses N-and C-termini similar to the 130-kDa mass range Cry proteins (Fig. 1C).
We then cloned the entire operon into the vector pHT304 and generated the plasmid pBMB1331, which we introduced into BMB171 cells for expression. Phase-contrast (Fig. 1D, panel b) and electron (Fig. 1E) microscopy observations revealed that BMB171/ pBMB1331 formed square-shaped parasporal crystals. SDS-PAGE analysis revealed that BMB171/pBMB1331 produced a major protein of 118 kDa in size (Fig. 1F, lane 2) that corresponds to the predicted molecular weight of Cry65Aa1.
ORF2 is required for expression and crystallization of Cry65Aa1. This study is the first to include orf2 downstream of the 130-kDa mass range Cry protein gene. To determine the function of ORF2 in this operon, a series of recombinant plasmids were constructed to analyze the expression and crystallization of Cry65Aa1 in BMB171 ( Fig. 2A).
When sporulated cells were examined by phase-contrast microscopy, crystalline inclusions were produced only when both cry65Aa1 and orf2 were present as an operon in the constructs (Fig. 2B). The B. thuringiensis BMB171 strain that contained the entire native cry65Aa1 operon (BMB171/pBMB1331 and BMB171/ pBMB1A-65Opn) produced typical square-shaped crystals (Fig. 2B, panels b, and e). However, B. thuringiensis BMB171 strains containing only the intact cry65Aa1 (BMB171/pBMB1332 and BMB171/ pBMB1A-65A) or orf2 alone (BMB171/pBMB1333 and BMB171/ pBMB1A-ORF2) did not produce visible crystals or inclusions during sporulation (Fig. 2B, panels a, c, d, and f). The total protein profiles of BMB171/pBMB1331 and BMB171/pBMB1A-65Opn showed that both strains produced Cry65Aa1 (Fig. 2C, lane 1), which was consistent with phase-contrast microscopy observations. The protein profiles of the other acrystalliferous recombinant strains all lacked a distinct band corresponding to Cry65Aa1 (Fig. 2C). These results demonstrated that ORF2 is required for Cry65Aa1expression and crystallization.
ORF2 has a stabilizing effect on cry65Aa1 mRNA. The role that ORF2 played in Cry65Aa1 expression and crystallization was unclear. The results presented above demonstrated that the presence of orf2 in cis can aid in cry65Aa1 expression and crystallization. Therefore, we also examined trans expression of cry65Aa1 and orf2. Interestingly, B. thuringiensis strains containing trans cry65Aa1 and orf2 in two different plasmids (BMB171/pBMB13321pBMB1A-ORF2 and BMB171/pBMB13331pBMB1A-65A) did not produce visible crystals or inclusions during sporulation (Fig. 3A, panels b and c). The SDS-PAGE results of the protein profiles of these two recombinant strains also both lacked distinct bands corresponding to Cry65Aa1 (Fig. 3B). These results demonstrate that the presence of orf2 in cis is essential for Cry65Aa1 expression and crystallization, indicating that the orf2 sequence may have a function at the cry65Aa1 transcription level. Therefore, we believe that cry65Aa1 and orf2 may act as a single operon. To confirm this hypothesis, we designed a pair of primers (Coe-F and Coe-R) to amplify the region that spans the entire noncoding region between cry65Aa1 and orf2. We analyzed the transcription pattern of this region at different growth stages of BMB171/pBMB1331 and found that this region can give rise to transcripts at 12 h and 18 h (Fig. 3C). These results demonstrated that cry65Aa1 and orf2 were co-transcribed under the control of the cry65Aa1 promoter and are assembled as an operon.
We then assessed the transcript levels of cry65Aa1 mRNA in the above strains at 12 h. The transcript profiles showed that BMB171/ pBMB1331 and BMB171/pBMB1A-65Opn produced the highest levels of cry65Aa1 mRNA (Fig. 3D) of other strains showed very low levels of cry65Aa1 mRNA (Fig. 3D). These results were consistent with phase-contrast microscopy and SDS-PAGE analysis, which indicated that BMB171/pBMB1331 and BMB171/pBMB1A-65Opn formed crystals and produced the highest levels of Cry65Aa1 protein. These results indicate that ORF2 has a stabilizing effect on cry65Aa1 mRNA and can promote cry65Aa1 mRNA accumulation.
orf2 mediates its mRNA-stabilizing effect on Cry65Aa1 through secondary structure. We provided evidence that ORF2 has a stabilizing effect on cry65Aa1 mRNA. However, its role in Cry65Aa1 transcription was unclear. The results described above demonstrated that the presence of orf2 in cis is essential for high level transcription of the cry65Aa1 gene. However, when orf2 is present in trans or is absent from the cry65A operon, the cry65Aa1 gene cannot be expressed at normal levels. This observation indicated that the sequence of orf2 but not the ORF2 protein may determine its mRNA-stabilizing effect. To confirm this hypothesis, we analyzed the DNA sequence of orf2 and identified a typical stemloop structure in orf2 coding region (Fig. 4A). We then constructed a series of recombinant plasmids to analyze the functional implications of this structure on Cry65Aa1 transcription, synthesis and crystallization (Fig. 4B).
We examined sporulated B. thuringiensis BMB171 strains transfected with plasmids carrying one of four orf2 mutants, by using phase-contrast microscopy. The first mutant plasmid was pBMB1334, in which orf2 had a mutation from ATG to TAA at the original start site. This mutant orf2 retained the stem-loop structure but could not express ORF2 protein. The second mutant was 'broken orf2': the plasmid pBMB1335 contained orf2 with a mutation that removed the stem-loop structure from bases 4254-4277 of the operon. The third mutant was 'intact orf2': the plasmid pBMB1336 carried a truncated orf2 that retained the region from ATG to base 4300. The fourth mutant was pBMB1337, which carried a truncated orf2 that retained the region from ATG to base 4300 and the terminator of the Cry65A operon. None of these mutants produced visible www.nature.com/scientificreports crystals or inclusions during sporulation (Fig. 4C, panels b, c, and d).
To further analyze the function of the stem-loop structure on transcription, we analyzed the transcript levels of cry65Aa1 mRNA in the above strains at 12 h. We found that strains BMB171/ pBMB1334 and BMB171/pBMB1331 produced similarly high levels of cry65Aa1 mRNA. BMB171/pBMB1337 produced the next highest level of cry65Aa1 mRNA. However, the transcript profile of BMB171/pBMB1335 indicated a very low level of cry65Aa1 mRNA production (Fig. 4E). These results were consistent with SDS-PAGE analysis.
Summarize the above results, when there is no ORF2 protein but an orf2 sequence is present (BMB171/pBMB1334), Cry65Aa1 can be transcribed and expressed at high levels. When the stem-loop structure is missing (BMB171/pBMB1335), Cry65Aa1 cannot be transcribed or expressed. When the stem-loop structure is retained, Cry65Aa1 can be transcribed and expressed, but at lower levels than in the presence of the wild type orf2 (BMB171/pBMB1337). These results demonstrated that the stem-loop structure in orf2 determined its mRNA-stabilizing effect on Cry65Aa1.
The ORF2 protein functions as a C-terminal crystallization domain for Cry65Aa1. We also noted that, when retaining the stem-loop structure but not expressing the ORF2 protein, the recombinant strains were able to transcribe and express Cry65Aa1 but could not form visible crystals. This finding indicated that, in addition to functioning as an mRNA stabilizer, the ORF2 protein may have an important role in Cry65Aa1 crystallization.
In the above results, B. thuringiensis BMB171 strains containing either intact orf2 (BMB171/pBMB1334) or broken orf2 (BMB171/ pBMB1337) expressed the Cry65Aa1 protein but did not produce visible crystals (Fig. 4C, panels b, e). These two strains contained the orf2 cis factors but did not express ORF2. We considered the possibility that crystal assembly of Cry65Aa1 required ORF2 proteins. To test this hypothesis, we transformed these two recombinant strains with pBMB1A-ORF2 into to add ORF2 protein. Microscopy revealed that the recombinant strains BMB171/pBMB13341 pBMB1A-ORF2 (Fig. 5B, panel a) and BMB171/pBMB13371 pBMB1A-ORF2 (Fig. 5B, panel b) produced typical square-shaped crystals similar to those produced by BMB171/pBMB1331. Additionally, the SDS-PAGE profiles of these two strains revealed the presence of the predicted Cry65Aa1 bands (Fig. 5C, lanes 1, 2). These results indicated that the crystal assembly of Cry65Aa1 requires ORF2 proteins.
We then purified crystals from BMB171/pBMB1331. SDS-PAGE analysis showed that the purified crystals produce two major bands 118 and 58 kDa in size (Fig. 5A). The 118-kDa band corresponds to Cry65Aa1, and the 58-kDa band corresponds to the predicted molecular weight of ORF2. To confirm whether the 58-kDa band was ORF2, the band was excised from SDS-PAGE gels and examined by mass spectroscopy. A total of 80% of the peptides were encoded by ORF2, as identified from matrix-assisted laser desorption/ionization tandem mass spectrometry (MALDI-TOF/TOF-MS) results (Fig.  S6). These results suggested that the ORF2 protein assembles with Cry65Aa1 forms crystals.
Sequence alignment revealed that ORF2 shares the highest level of similarity with the C-terminal regions of some reported Cry proteins (Fig. S5). Therefore, we hypothesized that ORF2 may function as a Cterminal domain to aid in Cry65Aa1 crystallization. To test this hypothesis, we constructed a recombinant plasmid (pBMB21B-N::ORF2) that contained the Cry21Ba N-terminus (Cry21Ba-N) and orf2 in-frame. Using this construct, we analyzed whether ORF2 can help other crystal proteins to assemble into crystals, as the C-terminal domain does in 130-kDa crystal proteins. We chose Cry21Ba1 because ORF2 has the highest similarity to the C-terminal domain of Cry21Ba1. Microscopy of sporulated cells of the recombinant strain BMB171/pBMB21B-N::ORF2 (Fig. 5B, panel c) produced typical bipyramidal crystals, similar to those of the wild-type Cry21Ba in BMB171. Additionally, the SDS-PAGE profiles of sporulated cultures showed the presence of the predicted bands (Fig. 5C, lane 3). Taking these findings together, we concluded that ORF2 is required for the expression and crystallization of Cry65Aa1 and functions as a C-terminal crystallization domain.
The crystallization of Cry65Aa1 requires two C-termini. We have shown that intact cry65Aa1 cannot be expressed or produce visible crystals under the control of either P65 or P1Ac ( Fig. 1 and Fig. 2). Structural analysis revealed that the Cry65Aa1 C-terminal domain has 333 aa, which is shorter than the typical C-terminal domain (Fig. 1C). It is possible that the short form C-terminal domain is degenerate and cannot aid the N-terminal domain of Cry65Aa1 in expression and crystallization. To test this, we constructed 3 recombinant plasmids, as illustrated in Fig. 6A. The plasmid pBMB65A-N::1Ac-C contained the Cry65Aa1 N-terminus (Cry65A-N) and the C-terminal domain of Cry1Ac in-frame, and pBMB65A-N::21Ba-C contained the Cry65Aa1 N-terminus (Cry65A-N) and Cry21Ba in frame. These two constructs were used to analyze whether a typical C-terminal domain contributes to the crystallization of Cry65Aa1. In addition, pBMB65A-N::ORF2 contained the Cry65Aa1 N-terminus (Cry65A-N) and ORF2 in frame and was used to analyze whether ORF2 contributes to the crystallization of Cry65Aa1.
Microscopy of sporulated cells showed that none of these strains produced visible crystals or inclusions during sporulation (Fig. 6B, panels a-c). Additionally, the SDS-PAGE profiles of sporulated cultures showed that all analyzed recombinant strains lacked distinct bands corresponding to Cry65Aa1 (Fig. 6C, lanes 1-3). These results indicated that Cry65A-C, ORF2, or a typical C-terminal domain does not contribute to the expression or crystallization of Cry65Aa1 when fused in-frame, such as those in 130-kDa crystal proteins.
Owing to the complex nature of these results, we have summarized all of the former constructs that can form crystals (Table 1). We found that the recombinant strains that can form crystals all contain cry65A and orf2 in trans or cis. This finding indicated that the crystallization of Cry65Aa1 might require two C-termini: the Cry65A Cterminal domain in addition to that of ORF2. To test this hypothesis, we constructed 3 additional recombinant plasmids: pBMB1338 (containing the Cry65A N-terminus [Cry65A-N] and the C-terminal domain of Cry1Ac in-frame and ORF2 in an operon), pBMB1339 (containing Cry65A and Cry1Ac-C in an operon), and pBMB1340 (containing Cry65A and Cry21Ba-C in an operon). Microscopy of sporulated cells of the recombinant strains BMB171/pBMB1338 (Fig. 6B, panel c) showed typical bipyramidal crystals. Additionally, the SDS-PAGE profiles of sporulated cultures of BMB171/pBMB1338 showed the presence of the predicted bands (Fig. 6C, lane 4). Microscopy of sporulated cells of the strains BMB171/pBMB1339 and BMB171/pBMB1340 (Fig. 6B, panels d, f) identified typical square-shaped crystals similar to those produced by the wild-type operon in BMB171/pBMB1331. Additionally, the SDS-PAGE profiles of the sporulated cultures of these two recombinant strains showed the presence of the predicted Cry65Aa1 bands (Fig. 6C, lanes 5, 6).
Taking these findings together, we concluded that the crystallization of Cry65Aa1 requires two C-termini: its own C-terminal domain in addition to ORF2 or a typical C-terminal domain.
Cry65Aa1 exhibits specific cytotoxicity toward MDA-MB231 cancer cells. Sequence analysis indicated that Cry65Aa1 is a novel class of parasporin; thus, we examined the cytotoxic effect of Cry65Aa1 on cancer cell lines. It is known that some Cry proteins must be activated by proteinase before they exhibit activity; therefore, solubilized Cry65Aa1 proteins were treated with proteinase K and trypsin. We found that Cry65Aa1 can be activated by trypsin, as a 55-kDa stable band was observed (Fig. 7B), which is similar to that observed for other characterized Cry three-domain toxins. The cytotoxic results are shown in Fig. 7. We found that there was no cytotoxic effect of non-activated Cry65Aa1 on any tested cell line (Fig. 7C). Although activated Cry65Aa1 exhibits high cancer cellkilling activity toward MDA-MB231 cells, it has no cytotoxic effect on HepG2 and L2 cells. Additionally, MDA-MB231 cytotoxicity mediated by activated Cry65Aa1 exhibited a significant dose-dependent response (Fig. 7D). We suggest that the activated Cry65Aa1 protein has a selective cytotoxic effect on MDA-MB231 cancer cells.
To analyze whether the Cry65Aa1 protein possesses other properties, its effects on insect and nematode activities were determined, using solubilized and activated Cry65Aa1. Neither form of Cry65Aa1 had any killing activity toward the tested insect and nematode larvae.

Discussion
Many parasporin proteins have been isolated from B. thuringiensis. These proteins have been divided into six types (PS1-6) by the Committee of Parasporin Classification and Nomenclature 26 . Here, we report a novel parasporin protein, Cry65Aa1, which has a selective cytotoxic effect on MDA-MB231 cancer cells (Fig. 7D). This novel parasporin protein may therefore be useful for the diagnosis and treatment of breast cancer. It also adds to the substantial number of identified parasporins and may be useful in comparative studies that aim to elucidate the mode of action of parasporins.
cry65A-intact orf21orf2 cry65A-cry21Ba-C 111 1 / Note: 1, representative the expression or transcription lever were below the 10% as that of wild type cry65A operon; 11, representative the expression or transcription lever were locate 10% to 50% as that of wild type cry65A operon; 111, representative the expression or transcription lever were similar as that of wild type cry65A operon;-, representative it has no expression or transcription products;/, representative not detected. Since the employment of next-generation sequencing technology to discover new genes, the number of new cry genes has increased rapidly 21 . However, an increasing number of cry genes cannot be expressed when cloned 13,[19][20][21] . Here, we found that Cry65Aa1 cannot be expressed when solely under its own control or the control of the Cry1Ac promoter ( Fig. 1E and Fig. 2C). Interestingly, the orf2 gene is located downstream of cry65Aa1 (Fig. 1B). We found that ORF2 assisted with Cry65Aa1 expression (Fig. 2C, lanes 2, 5) and crystallization (Fig. 2B, panels b, e). Our result indicated that cry65Aa1 cannot be expressed without ORF2. Because this observation may apply to other cry genes, the Cry65Aa1 expression system provides an appropriate strategy to promote the expression and crystallization in other cry genes.
Five reported cry genes (cry10Aa, cry19Aa, cry30Ca, cry39Aa, and cry40Aa) have gene organizations similar to the cry65Aa operon 24,25 (Fig. 1C). In particular, Barboza-Corona et al 25 demonstrated that ORF2 could enhance the synthesis and crystallization of Cry19A by functioning as a C-terminal crystallization domain. Our results also demonstrate that ORF2 in the cry65Aa operon has a C-terminal crystallization domain function and promotes Cry65Aa1 crystallization. However, there are differences between the cry65Aa and cry19A operons. First, we found that the orf2 sequence of the cry65A operon has an mRNA-stabilizing effect function on cry65Aa mRNA (Fig. 3). Furthermore, we showed that there is a typical stem-loop structure in the orf2 coding region and that this structure affects the mRNAstabilizing effect on cry65Aa mRNA (Fig. 4). Most likely, the translation of ORF2 prevents the stem-loop formation that results in mRNA stability, instead of ORF2 having a direct effect on mRNA stabilizing activity. The mechanisms that can improve Cry protein expression in B. thuringiensis have been studied at the transcriptional, posttranscriptional, and posttranslational levels. Factors containing promoters 27 , stable mRNA factors (such as STAB-SD sequences 28 ), and accessory proteins (such as P19 or P20 29 ), have been identified. Our results provide a novel understanding of the function of ORF2. We also reveal a novel mechanism that can improve Cry protein expression in B. thuringiensis.
In addition, the upstream reading frames of five similar operons all encode an N-terminal domain or a 65-kDa-range Cry protein (Fig. 1C). For the cry19A operon, it was speculated that mutations that accrued in the extant intergenic region between cry19A and orf2 resulted in this two-gene operon and that the Cry19A (75 kDa) and ORF2 (60 kDa) proteins together have features that are similar to those of 135-kDa Cry proteins 25 . However, Cry65Aa1 is a threedomain Cry protein and possesses N-and C-termini similar to the 130-kDa mass range Cry proteins. It is known that most Cry proteins that have a mass in the 130-kDa range typically do not require other proteins for synthesis and crystallization. Thus, the genes that encode these proteins occur alone rather than in operons 1 . However, the cry65A operon includes an orf2 downstream of the Cry proteins in the 130-kDa mass range. We do not have a clear explanation for why the 130-kDa mass range Cry65Aa1 exists as a two-gene operon that also encodes the crystallization domain-like ORF2. The C-terminal domain of Cry65Aa1 is shorter than the typical C-terminal domain (Fig. 1A). Given that Cry65Aa1 did not crystallize in the presence of its own C-terminal domain alone (Fig. 6), it is tempting to speculate that some mutations have accrued in its C-terminus and have resulted in C-terminal degradation and the partial loss of the capacity to form crystals. Thus, the protein requires an additional C-terminal domain to complement its own crystallization capacity. However, when the sequence was analyzed, we found that the Cry65Aa1 Cterminal domain has low homology with a hypothetical protein from V. ratti ACS-216-V-Col6b, but not with any reported C-terminal domains (Fig. S4). Consequently, another possibility is that, earlier in evolution, Cry65Aa1 was an N-terminal domain or a 65-kDarange Cry protein member. In the evolutionary process, the N-terminal domain may have fused with other proteins to produce the novel protein Cry65Aa1. However, at the synthesis and crystallization levels, this novel protein still requires the help of a C-terminal domain such as ORF2, which accounts for the current composition of the operon.
The ORF2 in the cry65A operon has highest homology with the Cterminal domain of some nematicidal Cry proteins (Fig. S5), but Cry65Aa1 has highest homology with a parasporin, Cry41Aa (Fig.  S4). We found that Cry65Aa1 has a selective cytotoxic effect on MDA-MB231 cancer cells (Fig. 7) but no anti-nematode activity. These findings suggested that Cry65Aa1 and ORF2 might have originated from different B. thuringiensis strains that had different ecological niches. Here, we suggest that the coexistence of cry65Aa1 and orf2 in SBT-003 is the result of a long evolutionary process. It is possible that cry65Aa1 and orf2 may exist separately in some strains that cannot express and form crystals and therefore have previously been overlooked because of a B. thuringiensis strain-screening strategy that is based on crystal formation. The downstream orf2 arrangement is not unique; rather, it is present in several genes 24,25 . We are unaware of the exact significance of this arrangement in the Cry protein family. However, we envisage that this configuration may be an evolutionary strategy of some groups of Cry proteins, such as Cry65Aa1. That these groups of proteins cannot be expressed or form crystals in their mother cells may be due to the loss of ORF2 or some other similar factor. Furthermore, the existence of strains that contain cry genes but do not form crystals creates challenges for the phenotype-based B. cereus and B. thuringiensis classification system 30,31 ; using the existing strategy, some B. thuringiensis strains containing cry genes with ''no crystals'' phenotype may be classified as B. cereus.
The discovery and application of novel crystal protein genes is considered to be an important approach to controlling and overcoming the potential resistance of certain pests. As a result, a growing number of cry genes have been identified using various strategies 4,11 . To discover novel cry genes, the typical starting point is to screen a novel B. thuringiensis strain. Thus far, a practical method of screening B. thuringiensis strains is to detect parasporal crystals by microscopy 30,31 . Using this method, a large number of B. thuringiensis strains have been isolated and maintained after detection of the parasporal crystals. All of the reported cry genes were cloned from B. thuringiensis strains that can form parasporal crystals. The Cry65A phenotype indicates that many ''no crystal'' B. thuringiensis strains that harbor novel cry genes may have been overlooked. The presence of these ''no crystal'' cry genes presents challenges to the new cry gene discovery strategy, which is based on either assessing crystals or SDS-PAGE profiles 17,18,21 . It can be envisaged that these overlooked B. thuringiensis strains contain a wealth of undiscovered cry genes. This situation has not been previously reported and requires future consideration when searching for novel cry genes. Our discovery provides a potential framework for isolating novel crystal protein genes from ''no crystal'' B. thuringiensis strains.

Methods
Bacterial strains, plasmids, and medias. The strains and plasmids used in this study are listed in Table S1. Escherichia coli and B. thuringiensis strains were maintained on Luria-Bertani (LB) medium and supplemented with appropriate antibiotics at 37uC and 28uC, respectively. For crystal protein preparation, B. thuringiensis strains were grown in ICPM liquid medium at 28uC until cell lysis. The transformation of B. thuringiensis was performed by electroporation, as described previously 32 .
Crystal protein preparation, solubilization, activation and quantification. Crystal preparation, solubilization, and quantification were performed as previously described 33 . Solubilized Cry65Aa1 proteins were treated with proteinase K and trypsin (final concentrations: 0, 1, 10, 100 and 200 mg/ml) at 37uC for 90 min. The samples were then dialyzed overnight against 50 mM Tris (pH 9.0) at 4uC for subsequent bioassay.
Sequence analysis. DNA sequences were determined, and primers were synthesized by AuGCT Biotechnology Co., Ltd (AuGCT, Beijing). Amino acid sequence alignments and phylogenetic trees were produced using MEGA 4.1 software. Analyses of primary and secondary protein structure predictions were performed using PHD software (http://www.predictprotein.org/). The protein sequences were compared to other proteins using BLAST.
Microscopy. The sporulating cultures were monitored using an optical microscope with an oil immersion lens. For transmission electron microscopy observations, samples were treated using methods described previously 33 and were examined using a Hitachi 7000 FA electron (Hitachi, Japan) microscope.
Construction of recombinant plasmids. The primers used in this study are listed in Table S2, and the plasmid construction strategy is illustrated in Figures. 2A, 4B, and 6A.
The cry65A gene was amplified with primers 65A-F and 65A-R, and a 3.2-kb DNA fragment was then cloned into the BamHI-XhoI site of pUC18-T to generate the plasmid pEMB1330. Then, the 3.2-kb cry65A DNA fragment was cloned into the BamHI-XhoI site of the E. coli-Bt shuttle expression vector pBMB1A to generate pBMB1A-65A for expression. To clone the entire Cry65A operon, primers 65Opn-F and 65Opn-R were designed based on the SBT-003 genome sequence, for amplifying cry65Aa1 together with orf2 and its promoter and terminator. Then, a 5.3-kb PCR fragment was purified and cloned into the BamHI-HindIII site of the E. coli-Bt shuttle vector pHT304 to generate pBMB1331. The primers 65A-F and ORF2-R-1 were used to amplify the cry65Aa1 and orf2 region, and the 4.7-kb PCR fragment was then digested with BamHI-XhoI and ligated into pBMB1A to generate pBMB1A-65Opn. The primers 65Opn-F and 65A-R were used to amplify the 3.5-kb putative promoter region of the cry65A operon and cry65Aa1 orf region. The primers T65-F and 65Opn-R were used to amplify the 0.2-kb terminator region of the cry65A operon. Then, the 3.5-kb and 0.2-kb DNA fragments were digested with XhoI and ligated together, and the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB1332. The primers ORF2-F-1 and ORF2-R-1 were used to amplify orf2, and the 1.5-kb PCR fragment was then digested with BamHI-HindIII and ligated into pBMB1A to generate pBMB1A-ORF2. The primers 65Opn-F and P65-R (splice overlap extension [SOE] primers contain orf2 1-20-bp regions) were used to amplify the 0.4-kb putative promoter region of the cry65A operon. The primers ORF2-F-2 (SOE primers containing 20-bp regions of putative promoter regions) and 65Opn-R were used to amplify the 1.7 kb orf2 and the terminator region of the cry65A operon. Then, the 0.4-kb and 1.7-kb DNA fragments were ligated by SOE PCR, and the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB1333.
Construction of ORF2 mutants. The primers Morf2-1 and Morf2-2 were used to construct the ORF2 mutant by SOE with primers 65Opn-F and 65Opn-R, which changed the start codon ATG to the stop codon TAA. Then, the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB1334. The primers 65Opn-F and Morf2-3 were used to amplify the 1-4254 site fragment of the www.nature.com/scientificreports cry65A operon. The primers Morf2-4 and 65Opn-R were used to amplify the 4277-5366 bp fragment of the cry65A operon. Then, the 4.2-kb and 1.1-kb DNA fragments were digested with XhoI and ligated together, and the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB1335, which removed the stem-loop structure of the cry65A operon. The primers 65Opn-F and Morf2-5 were used to amplify the 1-4300 bp fragment of the cry65A. Then, the 4.3-kb DNA fragment was digested with BamHI-XhoI and ligated into pHT304 to generate pBMB1336. The primers T65-F and 65Opn-R were used to amplify the 0.2-kb terminator region of the cry65A operon. Then, the 0.2-kb DNA fragment was digested with XhoI-HindIII and ligated into pBMB1336 to generate pBMB1337.
Chimeric protein construction. The C-terminal region of Cry1Ac contains 567 aa (from 612 to 1178), and the C-terminal region of Cry21Ba contains 606 aa (from 680 to 1286). The N-terminal region of Cry65Aa1 contains 731 aa (from 1 to 731), and the N-terminal region of Cry21Ba contains 680 aa (from 1 to 680). The primers 65Opn-F and 65A-N-R were used to amplify a 2.6-kb DNA fragment containing the putative promoter region of the cry65A operon and the N-terminal coding region of Cry65Aa1. The primers 1Ac-C-F and 1Ac-C-R were used to amplify a 1.9-kb DNA fragment containing the C-terminal coding region and the terminator of Cry1Ac. Then, the 2.6-kb and 1.9-kb DNA fragments were digested with SalI and ligated together, and the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB65A-N::1Ac-C. Similarly, the promoters of Cry21Ba and Cry65A and their N-terminal half coding regions were amplified using the primers 21Ba-N-F, 21Ba-N-R, 65A-F and 65AN-R, respectively. Additionally, the coding regions of ORF2, C-terminal half of Cry21Ba, and their terminators were amplified using the primers ORF2-F-3, 65Opn-R, 21Ba-C-F and 21Ba-C-R, respectively. Then, the PCR DNA fragments were digested with SalI and ligated together, and the products were digested with BamHI-HindIII and ligated into pHT304 to generate pBMB21B-N::ORF2, pBMB65A-N::ORF2, and pBMB65A-N::21Ba-C. To test the hypothesis that the crystallization of Cry65Aa1 requires two C-termini, we constructed 3 other recombinant plasmids: pBMB1338 (containing the Cry65A Nterminus [Cry65A-N] and the C-terminal domain of Cry1Ac in-frame and orf2 in an operon), pBMB1339 (containing Cry65A and Cry1Ac-C in an operon) and pBMB1340 (containing Cry65A and Cry21Ba-C coding regions in an operon).
All PCR products were purified using an OMEGA D2500-01 gel extraction kit (OMEGA, Shanghai), and recombinant plasmids were transformed in E. coli DH5a cells. The fragment integrity of all constructed plasmids was confirmed by restriction enzyme digestion and sequencing analysis (AuGCT, Beijing).
Peptide identification. The peptides corresponding to SDS-PAGE bands were examined using MALDI-TOF/TOF-MS system (4700 Proteomics Analyzer, Applied Biosystems). Database searching of spectral data was conducted using the MASCOT search engine (www.matrixscience.com, Matrix Science).
RNA extraction and expression analyses. Total RNA was extracted from B. thuringiensis strains using a total RNA isolation system (Promega, Madison, WI, USA). Analyses of cry65Aa1 gene expression were performed using qPCR with the primers Qrt-F and Qrt-R; the 16s RNA gene was amplified as a quantitative control. The qPCR was conducted using IQ SYBR Green Supermix (BioRad, Hercules, CA, USA) according to the manufacturer's instructions. At least three independent biological samples were used for each individual strain. The data were normalized by the value of 16s RNA signal, and fold changes in expression levels were calculated and compared with those of strain pBMB1331/BMB171.
Insect and nematode bioassays. The bioassays of toxicity toward insects, including Plagiodera versicolora, Tribolium castaneum, Helicoverpa armigera, Plutella xylostella, and Bombyx mori larvae, were conducted using methods described previously 33 . The nematode activity assays against Caenorhabditis elegans and Meloidogyne hapla were conducted using methods described previously 17 .
Cytotoxicity assays. The cytotoxic effects of solubilized and activated Cry65Aa1 were tested on MDA-MB-231 cells, HepG2 cancer cell lines, and L2 cells by MTT assay 34 . The cell lines were obtained from CCTCC (Wuhan) and were maintained in RPMI 1640 medium (Gibco). Some of the other cells were in DMEM medium (Gibco) supplemented with 10% fetal bovine serum (FBS), 100 units/ml penicillin, and 100 mg/ml streptomycin in 5% CO 2 at 37uC. Activated and non-activated Cry65Aa1 proteins were added to cultured mammalian cells at different concentrations (0, 1, 10, 20, 100 mg/ml). After preincubation at 37uC, cell proliferation was measured 24 h after administration.