Efficient infection of tree shrew (Tupaia belangeri) with hepatitis C virus grown in cell culture or from patient plasma

Hepatitis C virus (HCV) has infected more than 170 million people or 3 % of the world population. Most infections become persistent and lead to serious liver disease (Major et al., 2001). Currently, in the USA, most newly acquired cases of hepatitis C (68 %) are associated with injecting drug users, whilst 18 % of cases occur from infected sexual partners (Alter, 2002). HCV is an opportunistic pathogen in about 25 % of human immunodeficiency virus type 1 (HIV-1)-infected people. Our previous study showed that the rate of HCV co-infection among 138 HIV-1-infected injecting drug users from the south-eastern region of Yunnan Province of China was 99.3 % (Zhang et al., 2002). This co-infection causes higher HCV titres and a more rapid progression to cirrhosis (Chisari, 2005).

Since the discovery of HCV in 1989, two major difficulties – virus cell culture and the development of a non-primate small-animal model – have hampered basic research on HCV. Recently, a breakthrough in HCV cell culture was generated by several laboratories (Bukh & Purcell, 2006; Lindenbach et al., 2006; Zhong et al., 2005). Rice and co-workers reported that the infectious virus strain FL-J6/JFH from Huh7.5 cell cultures could establish long-term infections in chimpanzees, and that the virus recovered from these animals was highly infectious in cell culture (Lindenbach et al., 2006). However, the chimpanzee is an expensive and endangered species, and does not develop HCV-induced chronic liver disease (cirrhosis and hepatocellular carcinoma) (Guha et al., 2005). The development of a non-primate small-animal model is urgently needed for research on HCV.

Tree shrews (Tupaia belangeri) are similar to squirrels in their external appearance and habits. The body weight of the adult tree shrew ranges between about 100 and 250 g. Tree shrews can be used in many research fields (Fuchs, 1999; Xu et al., 2005). For human virus study, tree shrews can be infected by human herpes simplex virus (Darai et al., 1978), hepatitis A virus (Zhan et al., 1981), hepatitis B and D viruses (Li et al., 1996; Walter et al., 1996; Yan et al., 1996a), rotavirus (Pang et al., 1983) and Chikungunya virus (Zhang et al., 1991). Several studies have shown that in vivo infection of tree shrews with HCV from patients is possible (Liu et al., 1998; Wang et al., 1997; Xie et al., 1998). Recently, Zhao and others have found that primary hepatocytes of T. belangeri can be infected efficiently by HCV (Ming et al., 2001; Zhao et al., 2002). Scavenger receptor class B type I, CD81 and other cellular molecules on primary Tupaia hepatocytes are important cell-surface molecules for binding of HCV envelope to the hepatocytes (Barth et al., 2005). In previous reports, the in vivo HCV infection rate of tree shrews was low when the normal animal was inoculated with pooled HCV patient plasma containing genotypes 1a, 1b and 3, and infection of tree shrews with HCV grown in cell culture has not been reported. In this study, we demonstrated that normal tree shrews can be infected efficiently with various genotypes of native HCV and with HCV grown in cell culture when the animal is inoculated with an adequate amount of single-genotype HCV.

Patient plasma HCV.
Human plasma HCV (native HCV) was obtained from six patients persistently infected with HCV who were positive for anti-HCV antibody. The plasma from each patient was aliquotted separately and stored at –70 °C until use. All patients from Yunnan Province provided informed consent. Samples from patients co-infected with hepatitis B virus (HBV) and HIV were excluded following serological assays. The concentration of HCV RNA in patient plasma was measured using an HCV RNA Real-time Fluorescent RT-PCR kit (Daeran Biotechnology). The genotype of each patient HCV was determined using a method based on cloning and sequence analysis of a fragment either from the 5'-untranslated region and partial core gene or from the NS5b gene (Laperche et al., 2005).

HCV grown in cell culture (HCVcc).
HCVcc was provided by Mr Zhenqiu Chen (Guangzhou Chinawave Biotechnology Co., China). HCVcc was produced using recombinant pHCV-transfected HeLa or Huh7 cells as described previously (Li & Li, 2004; Yao et al., 2004). Briefly, the HCV genome was obtained from a patient with genotype 1b HCV acute infection, and a T7 promoter-containing HCV genome was inserted into plasmid pFK-1. Three specific amino acids (NS3, ¹²⁰²Glu→Gly and ¹²⁸⁰Thr→Ile; NS5A, ²¹⁹⁷Ser→Pro) were mutated by site-directed mutagenesis. This recombinant plasmid was transfected into HeLa cells using Lipofectamine 2000 (Invitrogen). The transfected cells were then infected with recombinant vaccinia virus vTF7-3 for 2 h. After removing free vTF7-3, the transfected cells were cultured for 48 h and then lysed by three cycles of freezing (–70 °C) and thawing. The cell lysate was centrifuged at 400 g for 20 min. The supernatant was filtered with a 0.22 µm filtration membrane. The filtrate, named HCVcc from HeLa cells, was concentrated by ultracentrifugation and used for tree shrew inoculation. Huh7 cells were inoculated with the concentrated filtrate at 37 °C for 24 h and then cultured in complete minimal essential medium for 48 h. The growth kinetics of HCV and expression of HCV structural and non-structural proteins and viral particles in the cell cultures were detected as reported previously (Yao et al., 2004). The Huh7 cell culture supernatant was collected and concentrated, and named HCVcc from Huh7 cells. The number of HCV RNA copies in HCVcc was detected using an HCV RNA Real-time Fluorescent RT-PCR kit as above.

Plasma alanine aminotransferase (ALT) assay.
The ALT activity in tree shrew plasma was detected according to the manual of the commercial ALT kit used (Reitman–Frankel method; Rongsheng Biotech Co.).

Detection of HCV RNA in tree shrew plasma.
Viral RNA in plasma was extracted using TRIZOL reagent (Invitrogen) and amplified by reverse transcription and nested-PCR using a One Step RT-PCR kit (TaKaRa Bio). HCV-specific primers were designed according to GenBank accession no. AB049088 as follows: inner sense primer 5'-CATAGATCACTCCCCTGTGAGGAACT-3' (nt –311 to –286); inner antisense primer 5'-CTGTGGGCGGCGGTTGGTGTT-3' (nt 60–40); outer sense primer 5'-ACTCCACCATAGATCACTCCCCTG-3' (nt –318 to –295); outer antisense primer 5'-AACAAGTAAACTCCACCAACGATCTG-3' (nt 110–85). Amplified products were analysed by agarose gel electrophoresis. In each RT-PCR experiment, two negative samples and one positive sample were added. The assay was considered to be valid when the reactions of the negative- and positive-control samples were correct.

Animals.
Adult tree shrews (Tupaia belangeri chinensis) were captured from Luquan County, Yunnan Province, China, and quarantined for 3 months. All animals underwent a physical examination. Housing, maintenance and care of the animals were performed in accordance with the regulations and recommendations of the Animal Care Committee of Kunming Institute of Zoology, Chinese Academy of Sciences, China.

Inoculation with native HCV from patient plasma.
Before inoculation, all animals were examined using HCV RT-PCR and the ALT kit. As shown in Table 1, there were eight animals (1S1–1S8) in group I and two (1C1 and 1C2) in group II (control group). Each animal in group I was inoculated intravenously (i.v.) on day 0, and intraperitoneally (i.p.) on days 1 and 2 with 0.8, 0.6 and 0.6 ml plasma, respectively, from only one patient. The animals in the control group were inoculated with normal human plasma in the same way as group I. A 1 ml blood sample was collected from each animal at weeks 0, 1, 2, 3, 5, 8, 10, 12, 14, 16, 18 and 20 post-inoculation (p.i.). The plasma was prepared for the ALT assay and stored at –70 °C until used for the HCV RT-PCR and viral load detection. For groups III and IV, the animals were i.v. inoculated on days 0 and 2, and i.p. inoculated on day 1 (Table 1). The tree shrew is a small animal, so i.v. injection is harder than i.p. injection. In order to approach an alternative easier route, we chose the regimes of i.v./i.p./i.p. and i.v./i.p./i.v. in contrast to those used by Xie et al. (1998) (i.v./i.v./i.v.) and Wang et al. (1997) (i.v./i.v.). Due to limitations of individual patient plasma volume, we used the inoculum viral dose sets described above in this study.

Table 1. HCV RNA change in tree shrew plasma p.i. with native HCV

Inoculation with HCVcc.
The inoculated animals were divided into five groups (A, B, C, PC and NC). As shown in Table 2, animals in groups A and B were inoculated with HCVcc prepared from HeLa cell culture, and in group C, with HCVcc from Huh7 cell culture. Each animal was i.v. inoculated on day 0 with 0.8 ml virus solution, and i.p. inoculated on days 1 and 2 with 0.6 ml virus solution. The positive-control animal (PC1) was inoculated with 1.17x10⁶ copies of genotype 2c native HCV. Negative-control animals were inoculated with cell culture medium instead of virus, in the same way as the experimental group. A blood sample from each animal was collected at weeks 0, 1, 2, 3, 5, 8, 10 and 12 p.i.

Table 2. HCV RNA changes in the plasma of tree shrews infected with genotype 1b HCVcc from HeLa or Huh7 cell culture

Quantification of HCV viral load in tree shrew plasma.
For some animals with chronic or persistent HCV infection, the viral load in plasma was detected using an HCV RNA Real-time Fluorescent RT-PCR kit (Shenzhen PG Biotech Co.) following the manufacturer's instructions. Detection of HCV-RNA in tree shrew plasma following inoculation
In this study, the presence of plasma HCV RNA was determined by nested RT-PCR in all animals pre- and p.i. Based on the brightness of the HCV-specific band in agarose gel, the amount of plasma HCV RNA was scored semi-quantitatively as +++, ++, + or – (Fig. 1a). The general high and low trends in this semi-quantitative data were in accordance with the viral load data detected using a commercial kit in five animals with persistent HCV infection (Fig. 1b–f). These results suggested that the semi-quantitative data basically predicted the changes in viraemia levels in the absence of assay-derived viral load measurements.

Table 1); 1S1–1S8, animals inoculated with native HCV (Table 1); N, negative control (normal uninfected tree shrew plasma); P, positive control (human plasma from an HCV-infected patient). The corresponding semi-quantitative scores +++, ++, + and – represent a very bright band, a bright band, a weak band and no visible band, and are indicated beneath the lanes. (b–f) Relationship among ALT level, viral load and HCV nested RT-PCR score in five representative tree shrews following inoculation with native HCV or HCVcc. Results are shown for animals 1S4 (b), 1S7 (c), 2S3 (d), PC1 (e) and SC2 (f), respectively. U, Units.

In vivo infection of tree shrews by native HCV
In order to understand better the characteristics of HCV infection in tree shrews, we propose several definitions of HCV infection in tree shrews based on the criteria of HCV infection in humans and chimpanzees, as well as on the biological features of the tree shrew. If the HCV-specific RT-PCR from the plasma of a tree shrew was positive more than 1 week p.i., the animal was considered to be infected. If HCV RNA was not detected in the plasma of an animal for more than 7 weeks until termination of the experiment, or HCV RNA had disappeared from the liver of a deceased animal after a positive reaction in the plasma, the HCV infection was classified as cleared or transient. If viraemia occurred intermittently or persistently for more than 10 weeks, the animal was defined as chronically infected. If the high viraemia continued for more than 8–10 weeks before termination of the experiment, the infection was regarded as persistent.

As shown in Tables 1 and 3, of a total of 18 animals inoculated with native HCV, 16 (89 %) became infected and infection was cleared in four of these (2S4, 2S7, 2S8 and 2S9) (25 %). Thus, chronic infection was observed in 12 animals (75 %): seven of these (58 %) had a persistent infection (1S4, 1S5, 1S7, 1S8, 2S2, 2S3 and 2S10), whilst the remaining five animals (42 %) had intermittent infection. We used four HCV genotypes (1b, 2c, 3b and 6) and two different inoculation routes (i.v./i.p./i.p. and i.v./i.p./i.v.) for injecting the tree shrews. We did not find an obvious correlation among infection rate, HCV genotype and inoculation route (Tables 1 and 3). Two animals (2S11 and 2S12), inoculated with the lowest dose of HCV (0.03x10⁶ RNA copies), were not infected, indicating that the inoculated viral dose may be a key factor for successful HCV infection of tree shrews.

Table 3. Number of animals and rates of infection of infected, chronically infected and viraemia-cleared tree shrews inoculated with various native HCV genotypes

In vivo infection of tree shrews by HCVcc
As shown in Tables 2 and 4, 10/12 inoculated tree shrews (83 %) were infected by HCVcc. Transient or cleared infection was found in eight of these animals; chronic infection occurred in two of the ten infected animals, whilst the remaining two (SC2 and SC3) were considered to be persistently infected. In group B, two animals (SB3 and SB4) inoculated with a low dose (3.8x10⁶ RNA copies) of HCVcc from HeLa cells were not infected. HCVcc from Huh7 cells appeared to have a higher infectivity and fewer animals cleared infection following inoculation than animals infected with HCVcc from HeLa cells (Tables 2 and 4).

Table 4. Number of animals and rates of infection of tree shrews inoculated with HCVcc from HeLa or Huh7 cell culture

Viral load in tree shrew plasma
In order to measure the viral load in infected animals in a cost-effective manner, the viral loads of five animals (1S4, 1S7, 2S3, SC2 and PC1) with persistent infection following inoculation with native HCV or HCVcc were determined. The viral load in most plasma samples was less than 1000 international units (IU) ml^–1, although occasionally the viral load reached 10⁴–10⁵ IU ml^–1. For example, the peaks of viral load in tree shrews 2S3 and PC1 inoculated with native HCV reached 1.26x10⁴ and 5.42x10⁵ IU ml^–1, respectively, and the peak viral load in animal SC2 inoculated with HCVcc from Huh7 cells reached 2.46x10⁴ IU ml^–1 (Fig. 1b–f).

Change in ALT levels in tree shrews following inoculation with HCV
During the experimental period, the means±SEM of ALT levels in 17 animals inoculated with native HCV plus four negative-control animals (Table 1), and 12 animals inoculated with HCVcc plus two negative-control animals, plus the positive control animal PC1 (Table 2) were measured. The results are shown in Fig. 2(a) and (b), respectively. In comparison with the ALT levels prior to inoculation, the ALT levels in both the experimental and control groups, with the exception of animal PC1, did not show any obvious change during the study period and the fluctuations were within the normal range (less than the mean±2.5SD). Thus, there was no bulk death of liver cells in these tree shrews following inoculation with HCV.

(17K): Fig. 2. Changes in mean ALT level in tree shrews before and after inoculation with native HCV (a) or HCVcc (b). Results are shown as means±SEM. , Experimental groups; , negative-control groups. U, Units.

Correlation of ALT level, viral load and HCV RNA score
In order to explore the correlation among ALT level, viral load and HCV RNA score, data from five animals (1S4, 1S7, 2S3, SC2 and PC1) with chronic infection following inoculation with native HCV or HCVcc were examined (Fig. 1). Changes in ALT level did not correlate with either viral load or HCV RNA score. As described above, the viral loads of the animals correlated well with the HCV RNA scores.

In summary, these data showed that tree shrews without immunosuppression can be infected efficiently by both native HCV and HCVcc when the animals are inoculated with an adequate amount of single-genotype HCV. The peak viral load reached 10³–10⁵ IU ml^–1 in chronically infected tree shrews. The ALT level did not increase significantly in most of the infected animals.

Infection of tree shrews with HCV from patients (native HCV) has been described in several papers. Wang et al. (1997) reported that intermittent HCV viraemia was observed in 3/6 tree shrews inoculated with HCV from patient serum using an HCV-specific nested RT-PCR. HCV core antigens were demonstrated in the liver tissue by immunohistochemistry. A publication by Xie et al. (1998) reported that transient or intermittent viraemia was detected in 8/23 tree shrews (34.8 %) inoculated with pooled sera from four HCV patients in one study in China, and in another study in Spain, transient and intermittent viraemia was found in 2/10 (20 %) tree shrews inoculated with mixed HCV sera (containing >2x10⁷ RNA copies of HCV genotypes 1a, 1b and 3) from five patients; in addition, four X-ray-irradiated tree shrews had a higher infection rate (2/4) (50 %). In another publication (Liu et al., 1998), each tree shrew was inoculated with HCV serum from either an HCV-infected blood donor or a patient. Intermittent viraemia was observed in 4/4 tested tree shrews, and viral antigens (NS3, NS5 and core) appeared in the livers of 6/10 tree shrews. HCV RNA was found in the livers of 3/3 tree shrews using a negative-strand RNA hybridization assay.

In the present study, transient (cleared) and chronic infections were found in 16/18 tree shrews (89 %) inoculated with native HCV. Among the infected animals, 12/16 animals (75 %) were chronically infected, and the viraemia in 4/16 animals (25 %) was transient or cleared. The efficiency of the native HCV infectivity in both our study (89 %, 16/18) and the study by Liu et al. (1998) (60 %, 6/10) is much higher than that in the studies by Xie et al. (1998) (34.8 %, 8/23) and Wang et al. (1997) (50 %, 3/6). The differences among infection rates may be caused by several factors. First, when each tree shrew was inoculated with plasma or serum from a single patient or donor, such as in this study and the study by Liu et al. (1998), the infection rate was higher than in tree shrews inoculated with pooled or mixed sera or with plasma, as in the study by Xie et al. (1998). In contrast to HIV-1, a recent study showed that two or more HCV virions rarely enter one cell, and that intra- or intergenotypic recombination of HCV is rare following superinfection (Bernardin et al., 2006). Secondly, the inoculum viral dose appears to be an important factor for successful infection. In our study, animals 2S11 and 2S12 were inoculated with the lowest dose (0.03x10⁶ HCV RNA copies) and were not infected by HCV; in contrast, animals 2S9 and 2S10 were inoculated with the same genotype of HCV at a dose of 1.12x10⁶ HCV RNA copies and became infected (Table 1). TCID₅₀ values should be used as a measure of inoculation viral dose in future studies, as reported recently in chimpanzee (Lindenbach et al., 2006), as the number of HCV RNA copies does not reflect viral infectivity accurately. Thirdly, it is reported that use of the subspecies T. belangeri chinensis might be a factor in efficient in vivo infection (Lindenbach & Rice, 2001). However, we suggest that a comparative large-scale study with different subspecies is needed for confirmation of this point of view. In addition, similar to HCV infection of humans and chimpanzees, the present study demonstrated that recipient gender, HCV genotype and inoculation route were not relevant to the infection rate of tree shrews, and that the tree shrew is susceptible to infection with adequate levels of single-genotype HCV in the absence of X-ray irradiation (Tables 1–4).

Compared with native HCV, HCVcc has a uniform single sequence and can be readily produced in large amounts. Therefore, HCVcc is the best viral resource for the study of an animal model of HCV infection. In the present study, we used HCVcc from HeLa and Huh7 cells to infect tree shrews. Higher doses (1x10⁸ copies) of HCVcc from HeLa cells had a higher infection rate (4/4) (Table 2, group A) than a lower dose (3.8x10⁶ copies) (2/4) (Table 2, group B). In group C, HCVcc from Huh7 cells demonstrated high infectivity: all animals were infected, and in two of the four animals a chronic infection was established (Tables 2 and 4; SC2 of Fig. 1). The success with this inoculum might be related to quasi-species generation during replication in Huh7 cells.

The normal range of ALT in an adult healthy tree shrew is 12–67 U (Reitman–Frankel method): the mean±SD was shown to be 30.8±18.0 for a female (n=17) and 37.8±14.0 for a male (n=16) (Ding, 1984). From Figs 1 and 2, we can see that the ALT level in most HCV-infected tree shrews varied within the normal range. In one case (PC1), the animal had slight ALT elevation (approximately fourfold) for several weeks and the level then declined to the normal range. These results suggest that most tree shrews did not undergo severe liver inflammation and bulk hepatocyte death following HCV infection. The ALT data in this study are different from other reports and are in accordance with in vitro infection of primary hepatocytes with HCV. Positive staining of HCV proteins occurs in only 1–5 % of Tupaia hepatocytes and the cells do not show any cytopathic changes (Zhao et al., 2002). In contrast, very high ALT elevation (ten times the normal level) often occurs in the acute stage of HCV infection in humans and chimpanzees (Major et al., 2004). Changes in ALT levels and viral loads are not correlated in HCV-infected humans and chimpanzees (Major et al., 2004). Similarly, in this study, viral load and/or HCV RNA score did not correlate with ALT levels (Fig. 1b–f). The ALT levels in the two animals from the negative-control group in the HCVcc infection experiment were lower than those in the other groups but remained within the normal range (Fig. 2b), according to the study by Ding (1984).

At present, it is well known that nested RT-PCR generally has a higher sensitivity than real-time RT-PCR and that the semi-quantitative method is not applicable at very high HCV RNA levels. In this study, nested RT-PCR was run in several batches; each batch included samples collected from different animals and at different time points. The semi-quantitative scores were recorded in single batches by eye, which could have resulted in some differences among batches. For example, in animal SC2 (Fig. 1f) the viral load at week 10 was higher than that at week 12, but the score at week 10 was + (run in one batch), and was thus lower than ++ (run in another batch) at week 12. In contrast, real-time RT-PCR was run using the same quantitative reference among batches. Therefore, the general trends of semi-quantitative HCV RNA scores basically reflected the levels of viral load (Fig. 1). In order to monitor HCV changes in plasma from HCV-inoculated animals cost-effectively, the semi-quantitative nested RT-PCR was used in all animals and the commercial real-time RT-PCR kit was used only in five representative tree shrews with chronic or persistent HCV infection. The highest peak in viral load in an HCV-infected tree shrew (PC1) reached 5.42x10⁵ IU ml^–1, and the peak viral load in other tree shrews generally ranged from 10³ to 10⁴ IU ml^–1 (Fig. 1). In HCV-infected humans and chimpanzees, the peak viral load usually reaches greater than 10⁵–10⁶ copies or IU ml^–1 (Folgori et al., 2006; Major et al., 2004). Thus, consistent with a previous study (Xie et al., 1998), HCV-infected tree shrews appear to have lower HCV viral loads than humans and chimpanzees.

In contrast to the chimpanzee HCV model, there are few biological and immunological research reagents available for tree shrews. For example, in the present study we were not able to detect HCV-specific antibody levels in infected tree shrews as sensitive and reliable tree shrew antibody determination assays are not available. The commercial anti-HCV ELISA kits for humans and some primates (chimpanzees and macaques) cannot be used for the tree shrew because of the absence of cross-reaction between tree shrew and human antibodies. Therefore, previous reports on the existence of tree shrew anti-HCV antibodies are not reliable and the development of new immunological reagents is needed urgently for further investigation into tree shrew humoral and cellular immune responses to HCV.

In two reports from Chinese journals, HCV NS3, NS5 and core antigens were found in the livers of some tree shrews using an immunohistochemical assay (Liu et al., 1998; Wang et al., 1997). This assay is supported by in vitro infection of primary Tupaia hepatocytes (Zhao et al., 2002). This assay is also being used for continuation of the present study; preliminary results have yielded a positive reaction (data not shown).

To date, the chimpanzee is the only recognized HCV animal model with an infection rate of almost 100 %. The rate of chronic or persistent infection ranges from 40 to 60 % and the rate of transient or cleared infection is about 50 %. However, HCV-induced cirrhosis and hepatocellular carcinoma has not been found in chimpanzees chronically infected with HCV (Bukh, 2004; Guha et al., 2005; Lanford et al., 2001). In the present study, except for the tree shrews inoculated with a low viral dose, 100 % of animals were infected by native HCV and HCVcc (Tables 1–4). For native HCV, 12/16 tree shrews (75 %) were chronically infected, whilst 4/16 animals (25 %) resolved the infection (Table 3). For HCVcc, 4/4 animals were infected by HCVcc from Huh7 cells and two of these were classified as persistent infection. Recent studies on infection of tree shrew primary hepatocytes with native HCV further supports the research on in vivo infection (Barth et al., 2005; Guitart et al., 2005; Ming et al., 2001; Zhao et al., 2002). As hepatocellular carcinoma can occur in HBV-infected tree shrews (Yan et al., 1996b), it will be interesting to see whether hepatocellular carcinoma can also be induced in tree shrews by HCV. Taken together, these results are promising and should encourage an expanded and comprehensive investigation into the development of a practical and reliable tree shrew model for HCV infection.

We are grateful to Mr Zhenqiu Chen for providing concentrated HCVcc, Mrs Xueshan Xia and Wenhua Zhao for assistance in HCV genotyping, Drs Yanwei Qi and Junxin Zhang for patient consultation, Mrs Yunzhen Wei for animal maintenance and care, and Mr Weiyun Li for partial support from the China 863 Program. This work was supported in part by the Science Innovation Program of the Chinese Academy of Sciences (1999–2004) and China 863 Program (no. 2002AA216031).

Footnotes

^,‡, Hongbo Chen¹ †These authors contributed equally to this work. ‡Present address: Institute of Microbiology, Chinese Academy of Sciences, Zhongguancun, Beijing, China.