Purification of low-abundance Arabidopsis plasma-membrane protein complexes and identification of candidate components


(fax +1 612 624 6264; e-mail katagiri@umn.edu).


Purification of low-abundance plasma-membrane (PM) protein complexes is a challenging task. We devised a tandem affinity purification tag termed the HPB tag, which contains the biotin carboxyl carrier protein domain (BCCD) of Arabidopsis 3-methylcrotonal CoA carboxylase. The BCCD is biotinylated in vivo, and the tagged protein can be captured by streptavidin beads. All five C-terminally tagged Arabidopsis proteins tested, including four PM proteins, were functional and biotinylated with high efficiency in Arabidopsis. Transgenic Arabidopsis plants expressing an HPB-tagged protein, RPS2::HPB, were used to develop a method to purify protein complexes containing the HPB-tagged protein. RPS2 is a membrane-associated disease resistance protein of low abundance. The purification method involves microsomal fractionation, chemical cross-linking, solubilization, and one-step affinity purification using magnetic streptavidin beads, followed by protein identification using LC-MS/MS. We identified RIN4, a known RPS2 interactor, as well as other potential components of the RPS2 complex(es). Thus, the HPB tag method is suitable for the purification of low-abundance PM protein complexes.


In Arabidopsis, 25% of genes, including more than 600 receptor-like protein kinase (RLK) genes (Shiu et al., 2004), are annotated as encoding integral membrane proteins (Schwacke et al., 2003). Proteins without obvious transmembrane domains can also localize to the PM through post-translational modifications, protein–protein interactions and possibly some unknown mechanisms. For example, members of the coiled-coil/nucleotide-binding site/leucine-rich repeat (CC-NB-LRR) disease resistance (R) family, such as RPM1 (Boyes et al., 1998), RPS2 (Axtell and Staskawicz, 2003) and probably RPS5 (Holt et al., 2005), are localized to the PM through unknown mechanisms.

Formation and subsequent dynamic changes of protein complexes are essential for the functions of many proteins, including PM proteins. Therefore, it is important to know what other proteins are included in protein complexes that contain a particular protein of interest. Tandem affinity purification (TAP) (Rigaut et al., 1999) is often used for purification of protein complexes. The original TAP tag consists of tandem protein A domains for the first affinity purification, a tobacco etch virus protease site for cleavage, and a calmodulin-binding protein (CBP) domain for the second affinity purification (Rigaut et al., 1999). Use of two affinity purification steps enables a high level of purification of tagged proteins and/or protein complexes containing the tagged proteins. The TAP tag approach has been used to purify protein complexes from bacteria (Gavin et al., 2002; Gully et al., 2003) and mammals (Burckstummer et al., 2006; Gregan et al., 2007; Knuesel et al., 2003). A modified TAP tag that lacks a nuclear localization signal in the CBP domain was developed for protein complex purification from plants (Rohila et al., 2004). Application of this modified TAP tag to 41 rice protein kinases showed that more than half of the kinases exist as protein complexes (Rohila et al., 2006). The authors TAP-tagged the intracellular domains of the RLKs rather than the full-length proteins to avoid difficulties associated with membrane proteins (Rohila et al., 2006). It is conceivable that some true components of the protein complexes were not co-purified using such truncated proteins. A different TAP tag, which contains the two IgG-binding domains for the first affinity purification, a rhinovirus 3C protease site for cleavage, and a 6 × His or 9 × Myc tag for the second affinity purification, has also been reported for use in plants (Rubio et al., 2005). Of seven proteins tagged with this TAP tag, two were fully functional and three were partially functional (Rubio et al., 2005).

Methods using a single affinity tag have also been used for protein complex purification. Compared to TAP tag methods, one-step affinity purification methods are generally easier and faster, and result in a higher yield of tagged proteins. A good example is the Strep II tag, which is an 8–9 amino acid peptide that structurally mimics biotin, thus allowing purification using streptavidin or its derivatives (Witte et al., 2004). The Strep II tag has been used successfully in plants and mammals (Junttila et al., 2005; Witte et al., 2004). However, the dissociation constant (Kd) between the Strep II tag and streptavidin is approximately 10−6 m (Witte et al., 2004). This high Kd could make purification of low-abundance protein complexes difficult.

The biotin–avidin or biotin–streptavidin interactions are the strongest non-covalent biological interaction known, with a Kd of 10−15 m (Wilchek and Bayer, 1990). If affinity purification is performed on the basis of this interaction, such high affinity could allow efficient recovery of tagged proteins and stringent purification to reduce non-specific binding of other proteins. Single-step purification of soluble mammalian protein complexes using biotinylated tags has been reported (de Boer et al., 2003; Wang et al., 2006). A TAP tag purification kit utilizing the biotin–streptavidin interaction in mammalian cells is commercially available (NativePure™, Invitrogen, http://www.invitrogen.com/). The biotin carboxyl carrier protein domain (BCCD) of 3-methylcrotonyl CoA carboxylase (MCCA) from tomato can be biotinylated in Escherichia coli (Wang et al., 1994), and was used to tag a soluble TATA-box binding protein (TBP) of rice (Zhong et al., 2003). Using rice suspension cells overexpressing biotin-tagged TBP, candidate components of the TBP complex(es) were co-purified using a streptavidin matrix and identified using mass spectrometry (Zhong et al., 2003). However, purification of PM protein complexes from eukaryotes using a biotin tag has not been reported.

Challenges in purification of low-abundance PM protein complexes include efficient enrichment of protein complexes, solubilization of membrane proteins, and preservation of complex integrity during purification. With the above challenges in mind, we developed a procedure that is suitable for purification of PM-localized low-abundance protein complexes in plants. We devised a biotin tag called the HPB tag. All five Arabidopsis proteins that we tagged with the HPB tag were functional and biotinylated with high efficiency in vivo. Using a plant line expressing RPS2::HPB from the RPS2 promoter at a low level, we optimized a method for purification of PM protein complexes containing an HPB-tagged protein. Using LC-MS/MS, we identified RPS2::HPB, its known interactor RIN4, and other potential RPS2 complex components. Thus, the HPB tag-based purification procedure is suitable for purification of PM-localized protein complexes even when the tagged protein is expressed at a low level. Application of the HPB tag will facilitate discovery of potential components of plant PM protein complexes, which has generally been challenging.

Results and discussion

Construction of the HPB tag

The BCCDs of Arabidopsis MCCA and tomato MCCA share 65% identity and 82% similarity. Based on the successful use of the BCCD of tomato MCCA (Zhong et al., 2003), we chose the 80 amino acid sequence at the C-terminus of Arabidopsis MCCA (At1g03090), which contains the BCCD, as the biotin affinity tag in our study (Figure 1a). An HA epitope tag and a PreScission™ protease (GE Healthcare Life Sciences, http://www.gelifescience.com) recognition site were placed in front of the biotin tag (Figure 1a), resulting in a TAP tag that was termed the HPB (HA–PreScission–Biotin) tag. The HPB tag was designed to create C-terminal fusions to proteins of interest. Biotinylation of the tag facilitates high-affinity purification using a streptavidin matrix. If needed, a TAP strategy can be applied using the HPB tag: tagged proteins bound to a streptavidin matrix can be cleaved using the PreScission™ enzyme and released from the matrix, and the released protein with the HA tag can be further affinity-purified using an anti-HA antibody. However, we used one-step affinity purification based on the biotin–streptavidin interaction because it resulted in sufficient purity and allowed much higher yield of the tagged protein than a TAP procedure.

Figure 1.

 HPB tag sequence and expression cassettes.
(a) Amino acid sequence of the HPB tag. The peptide sequences corresponding to the HA epitope tag, the PreScission™ protease cleavage sequence and the biotin carboxyl carrier protein domain (BCCD) are shown in capital letters and underlined.
(b) Diagrams of the expression cassettes in the Gateway® destination vectors for pMDC32-HPB (top) and pMDC-pRPS2-HPB (bottom). Vector pMDC32-HPB has two copies of the CaMV 35S promoter (2 × 35S), while pMDC-pRPS2-HPB has the RPS2 promoter (proRPS2). The HPB tag is located behind the attR2 site to create a C-terminal fusion. In both vectors, a nos terminator (nos T) was used downstream of the expression cassettes.

To facilitate C-terminal tagging and expression of the tagged protein in plants, we constructed two Gateway® destination vectors, pMDC32-HPB and pMDC-pRPS2-HPB, which have two copies of the CaMV 35S promoter (2 × 35S) or the RPS2 promoter, respectively. The destination vectors were both produced using the backbone sequence of pMDC32 (Curtis and Grossniklaus, 2003) (Figure 1b).

HPB-tagged proteins were functional in vivo

A good tag should not affect the function of the tagged protein. Five Arabidopsis proteins and one E. coli protein, β-glucuronidase (GUS), were tagged and transgenically expressed in Arabidopsis. The five Arabidopsis proteins included three membrane-localized R proteins (RPM1, RPS2 and RPS5) (Bent et al., 1994; Grant et al., 1995; Mindrinos et al., 1994; Warren et al., 1998), one PM-localized RLK (FLS2) (Chinchilla et al., 2007; Gomez-Gomez and Boller, 2000) and one soluble protein (NPR1) that shuttles between the cytoplasm and the nucleus (Mou et al., 2003). RPM1::HPB and RPS2::HPB were driven by the RPS2 promoter. RPS5::HPB and FLS2::HPB were driven by their own promoters. NPR1::HPB and GUS::HPB were driven by the 2 × 35S promoter. An rpm1 rps2 double mutant line was transformed with the RPS2::HPB or RPM1::HPB constructs. The rps5, fls2 and npr1 mutant lines were transformed with the RPS5::HPB, FLS2::HPB and NPR1::HPB constructs, respectively. Col-0 wild-type plants were transformed with the GUS::HPB construct. For each construct (except GUS::HPB, for which only one line was used), two individual T4 or T5 lines, each carrying the transgene at a single locus, were selected for further study.

R proteins confer gene-for-gene resistance by directly or indirectly recognizing specific effector proteins delivered from pathogens (Jones and Dangl, 2006). RPM1 recognizes the effector AvrRpm1, RPS2 recognizes AvrRpt2, and RPS5 recognizes AvrPphB (Bent et al., 1994; Grant et al., 1995; Mindrinos et al., 1994; Warren et al., 1998). Recognition results in rapid activation of defense responses and consequent limitation of pathogen growth. It is often associated with localized programmed cell death, known as the hypersensitive response (HR), at pathogen infection sites (Greenberg and Yao, 2004). Plants with RPM1, RPS2 or RPS5 limit pathogen growth and display an HR when challenged with pathogens expressing AvrRpm1, AvrRpt2 or AvrPphB, respectively. We tested each of the HPB-tagged R genes for complementation of R-gene mutations by monitoring bacterial growth and the HR. (Figure 2a) shows complementation of rpm1 by RPM1::HPB. The left panel shows that the titer of Pseudomonas syringae pv. tomato DC3000 (Pto) expressing AvrRpm1 was as low in both RPM1::HPB lines as in wild-type Col plants, but the titer in the parental rpm1 rps2 line was more than 100 times higher. The right panel shows that the HR was as strong in both RPM1::HPB lines as in wild-type plants, but was absent in the parental double mutant line. Thus, RPM1::HPB was functional. Analogous experiments shown in (Figure 2b,c) show that RPS2::HPB and RPS5::HPB were also functional.

Figure 2.

 Function of HPB-tagged proteins in vivo.
(a) RPM1::HPB was functional. Two RPM1::HPB lines (RPM1::HPB-1 and -2), the parental rpm1 rps2 line, and the wild-type line (Col) were tested for bacterial growth (left panel) and the HR (right panel). For the results shown in the left panel, bacterial counts were measured 2 days after Pto avrRpm1 was infiltrated at an attenuance at 600 nm of 0.0001. Bacterial counts were measured 2 days after infiltration. The bars represent mean values with standard error. In the right panel, photographs of representative leaves were taken 24 h after Pto avrRpm1 was infiltrated at an attenuance at 600 nm of 0.08.
(b) RPS2::HPB was functional. Similar experiments were performed as in (a) using two RPS2::HPB lines (RPS2::HPB-1 and -3), the parental rpm1 rps2 line, and the wild-type line (Col), except that Pto avrRpt2 was used as the bacterial strain.
(c) RPS5::HPB was functional. Similar experiments were performed as in (a) using two RPS5::HPB lines (RPS5::HPB-1 and -10), the parental rps5 line, and the wild-type line (Col), except that Pto avrPphB was used as the bacterial strain.
(d) FLS2::HPB was functional. Leaves of two FLS2::HPB lines (FLS2::HPB-4 and -19), the parental fls2 line, and the wild-type line (Col) were infiltrated with 1 μm flg22. One day later, the same leaves were infiltrated with Pto at an attenuance at 600 nm of 0.0001. Bacterial counts were measured 2 days after infiltration with bacteria.
(e) NPR1::HPB was functional. Seeds of two NPR1::HPB lines (NPR1::HPB-10 and -15), the parental npr1-1 line, and the wild-type line (Col) were sown on 1× MS agar plates supplemented with 50 μm sodium salicylate. Photographs were taken after 2 weeks.
(f) GUS::HPB was functional. Leaves of one GUS::HPB line (GUS::HPB-4) and the parental non-transgenic line (Col) were histochemically stained for GUS enzyme activity as described by Tsuda and Yamazaki, 2004). All the experiments were repeated, and similar results were obtained.

FLS2 can recognize a 22 amino acid peptide (flg22) from bacterial flagellin and trigger downstream signaling that activates defense responses (Felix et al., 1999; Gomez-Gomez and Boller, 2000; Gomez-Gomez et al., 1999). Pre-treatment of wild-type plants (FLS2) with flg22 induces strong resistance against Pto (Tsuda et al., 2008; Zipfel et al., 2004). We pre-treated FLS2::HPB lines with flg22 and tested induced resistance to Pto. In both FLS2::HPB lines, like wild-type plants, growth of Pto was reduced 300- to 3000-fold compared to the fls2 mutant, indicating that FLS2::HPB is functional.

Seedlings of npr1 plants are more sensitive to exogenous salicylic acid (SA) than those of wild-type (Cao et al., 1997). By growing seedlings on MS agar plates supplemented with 50 μm sodium salicylate, we found that both NPR1::HPB lines, like Col, were more resistant than the parental npr1 mutant to exogenous SA (Figure 2e), indicating that NPR1::HPB was functional.

Histochemical staining of GUS enzyme activity in the GUS::HPB line also indicated that the HPB tag did not affect the GUS enzyme function (Figure 2f). Thus, all of the six proteins that we tagged with the HPB tag at their C-termini were functional in Arabidopsis.

HPB-tagged proteins are efficiently biotinylated in vivo

Next we tested biotinylation of the HPB-tagged proteins in Arabidopsis. One transgenic line for each construct was arbitrarily chosen to produce protein extracts for detection of biotinylated proteins. RPS2::HPB, RPM1::HPB, RPS5::HPB, FLS2::HPB and GUS::HPB proteins were all detected at expected molecular sizes when biotinylated proteins were visualized on a protein blot using peroxidase-conjugated streptavidin (Figure 3a). Two major endogenous biotinylated proteins, MCCA at approximately 78 kDa and the biotin carboxyl carrier proteins at approximately 30 kDa, were also detected (Figure 3a). The size of NPR1::HPB is too close to that of MCCA to allow them to be distinguished. Therefore, we captured biotinylated proteins using streptavidin beads and then subjected the captured proteins to protein blotting with an anti-HA antibody to detect the HA epitope in NPR1::HPB. (Figure 3b) shows that most NPR1::HPB was captured by streptavidin beads, as the amount in the bound fraction was much greater than in the unbound fraction and was comparable to the amount in the total fraction. This indicates that NPR1::HPB was also efficiently biotinylated.

Figure 3.

 HPB-tagged proteins were efficiently biotinylated in vivo.
(a) Biotinylation of HPB-tagged proteins in Arabidopsis. Protein samples from the indicated plant lines were analyzed using a protein blot with peroxidase-conjugated streptavidin to detect biotin. Two Arabidopsis endogenous biotinylated proteins are indicated by arrows (MCCA and biotin carboxyl carrier proteins (BCCPs)).
(b) Biotinylation of NPR1::HPB in Arabidopsis. Leaf protein extracts from NPR1::HPB-10 were subjected to streptavidin-bead binding. The protein fractions before binding (Total), those not bound to the beads (Unbound) and those bound were analyzed by protein blot using an anti-HA antibody.
(c) Most RPS5::HPB and GUS::HPB molecules were biotinylated. Leaf protein extracts from RPS5::HPB-10 and GUS::HPB-4 lines were subjected to streptavidin-bead binding. Total, unbound and bound fractions were analyzed by protein blot using peroxidase-conjugated streptavidin (upper panel) or anti-HA antibody (middle panel). The Rubisco large subunit was stained with Ponceau S as a measure of the total protein amount loaded in each lane (lower panel). Note that the total protein amounts in the ‘Total’ and ‘Unbound’ lanes were approximately the same but the amount was much lower in the ‘Bound’ lane. HPB-tagged proteins are marked by asterisks. All the experiments were repeated, and similar results were obtained.

We also examined the biotinylation efficiency in one line each for RPS5::HPB and GUS::HPB, which are membrane-associated and soluble, respectively. These lines expressed the tagged proteins at relatively high levels, which enabled the biotinylation analysis. For these HPB-tagged proteins, we visualized biotinylated proteins on a protein blot using peroxidase-conjugated streptavidin (Figure 3c, top panel), as well as HA tag-containing proteins using anti-HA antibody (Figure 3c, middle panel). Not all biotinylated HPB-tagged proteins were captured by the streptavidin beads, as we were able to detect biotinylated proteins in the unbound lanes (Figure 3c, top panel). The important point is that the ratios of band intensities among the total, unbound and bound lanes between biotin-visualized and HA-visualized blots for each of RPS5::HPB and GUS::HPB are very similar. If the unbound fraction contains a significant amount of non-biotinylated HPB-tagged protein, the ratio for the HA-visualized band in the unbound lane should be greater than the ratio for the biotin-visualized band in the unbound lane. Therefore, these results indicate that most RPS5::HPB and GUS::HPB molecules were biotinylated.

Optimization of RPS2::HPB protein complex purification

As we were particularly interested in RPS2, we proceeded with purification of protein complex(es) containing RPS2::HPB. There are several advantages to using RPS2::HPB for technology development. (1) As RPS2 is tightly associated with the PM (Axtell and Staskawicz, 2003), a method developed with RPS2 will probably be applicable to PM proteins in general. (2) As RPS2::HPB was expressed at a low level in the transgenic line used (line 1; Figure 3a), a method developed with RPS2 will probably be highly robust. (3) RIN4 can be used as a positive control because it is known to physically interact with RPS2 (Axtell and Staskawicz, 2003; Mackey et al., 2002, 2003). For improvement of the purification method, we focused on four aspects: reduction of endogenous biotinylated proteins, solubilization of PM proteins, capturing biotinylated RPS2::HPB, and cross-linking.

We found that enrichment of microsomes by centrifugation greatly increased the ratio of RPS2::HPB to the major endogenous biotinylated proteins (Figure 4a, compare lane 2 and lane 1). We also explored the possibility of genetically removing major endogenous biotinylated proteins. A T-DNA insertion line (SALK_137966) in which the MCCA gene was interrupted was morphologically similar to wild-type. As MCCA is mainly involved in leucine degradation in the mitochondria (Che et al., 2002), it is unlikely that it will affect PM protein complex formation. We crossed the mcca mutation into RPS2::HPB line 1 plants. Detection of biotinylated proteins from the microsomal fraction clearly showed that MCCA was eliminated in RPS2::HPB mcca plants (Figure 4a, compare lane 3 and lane 2). Therefore, RPS2::HPB mcca plants were used for all further experiments.

Figure 4.

 Optimization of RPS2::HPB complex purification.
(a) Separation of RPS2::HPB from endogenous biotinylated proteins. The biotinylated proteins were detected using protein blots with peroxidase-conjugated streptavidin in crude leaf extracts (lane 1), microsomal fractions (lanes 2 and 3), and the fraction bound to streptavidin beads under optimized conditions (lane 4). The samples were prepared from RPS2::HPB lines with MCCA (lanes 1 and 2) or mcca (lanes 3 and 4) backgrounds.
(b) Cross-linking was necessary for efficient RIN4 co-purification. RPS2::HPB in the microsomal fraction was treated with various concentrations of DSP, then solubilized and captured with streptavidin beads. Before loading the SDS–PAGE gel, the DSP cross-links were cleaved by 2-mercaptoethanol treatment. Proteins were detected using peroxidase-conjugated streptavidin (Biotin), anti-HA antibody (HA) or anti-RIN4 antibody (RIN4). RPS2::HPB was silver-stained in a separate gel loaded with the same samples. The panels for biotin, HA and silver staining only show the part around the molecular weight of RPS2::HPB, and the panel for RIN4 only shows the part around the molecular weight of RIN4.
(c) DSP efficiently cross-linked RPS2::HPB-containing protein complexes, and the cross-link could be efficiently cleaved afterward. Microsomal fractions prepared from an RPS2::HPB line (in the mcca background; lanes A) and the wild-type line (MCCA; lanes B) were subjected to DSP cross-linking. Part of each sample was subjected to cross-link cleavage with 2-mercaptoethanol (Cleaved) and the other part was left uncleaved (Non-cleaved). Proteins captured with streptavidin beads were analyzed by protein blotting for biotin, HA and RIN4, as described in (b). RPS2::HPB is marked by asterisks. Major endogenous biotinylated proteins (MCCA and BCCPs) are indicated by arrows.

To solubilize RPS2::HPB from the microsomes, various detergent conditions were tested. Lipid rafts, a specialized PM structure, are not solubilized using mild non-ionic Triton X–100 at 4°C (Brown, 1994; Jacobson et al., 2007; Schroeder et al., 1998). Therefore, we included in our tests ionic detergents, such as sodium deoxycholate and sodium dodecyl sulfate (SDS), alone and in combination with non-ionic detergents, such as Triton X–100 and 4-nonylphenyl-polyethylene glycol (NP-40). We found that 0.5% and 1% w/v sodium deoxycholate alone effectively solubilized RPS2::HPB from microsomes (Figure S1a). Interestingly, sodium deoxycholate specifically reduced solubilization and capture of the approximately 30 kDa biotin carboxyl carrier protein (Figure 4a, lane 4). Using HEPES-based resuspension buffer containing 0.5% sodium deoxycholate and 150 mm NaCl for solubilization and the following bead-binding step, RPS2::HPB was the only major biotinylated protein bound to streptavidin beads (Figure 4a, lane 4). Consequently, we adopted these conditions.

To cross-link the components of the RPS2::HPB protein complex(es), we tested two cross-linkers, formaldehyde and dithiobis[succinimidyl propionate] (DSP). Both are membrane-permeable and easily reversible or cleavable. Formaldehyde has been used to recover soluble GVG protein complex from plant extract in a TAP purification approach (Rohila et al., 2004). DSP has been used for identification of protein transport complexes in the chloroplastic envelope membrane (Akita et al., 1997) and an ER transport receptor complex (Appenzeller et al., 1999). We decided to cross-link proteins in the microsomal fraction rather than in whole-plant tissues because a well-controlled cross-linking reaction could easily be performed using the microsomal fraction in microfuge tubes, and because a small amount of toxic cross-linker was sufficient. Solubilization of RPS2::HPB was very difficult after treating microsomes with 1% formaldehyde (Figure S1b). On the other hand, the majority of RPS2::HPB was solubilized after cross-linking with 5 mm DSP (Figure S1c), so we decided to use DSP. We further found that 1 mm DSP worked as well as 5 mm DSP with respect to solubilization of RPS2::HPB (data not shown), capture of RPS2::HPB with streptavidin beads (Figure 4b, two upper panels), and co-purification of the known interactor RIN4 (Figure 4b, third panel from top). More RIN4 was co-purified when cross-linker was used, even though less RPS2::HPB was recovered (Figure 4b). We concluded that a cross-linking step is necessary for purification of RPS2::HPB protein complex(es) under our experimental conditions. We inspected the effectiveness of cross-linking using mobility shift on SDS–PAGE gels. Protein purification using RPS2::HPB mcca and wild-type was carried out under optimized conditions (see Experimental procedures). Both cross-link-cleaved and non-cleaved samples were subjected to SDS–PAGE and protein blotting. As expected, both RPS2::HPB and RIN4 were detected as clear bands in the cleaved samples (Figure 4c, left). In non-cleaved samples, RPS2::HPB was detected as a smear extending towards the high molecular weight side (Figure 4c, right), indicating that RPS2::HPB was cross-linked to other proteins. Furthermore, in the non-cleaved sample, RIN4 was not detected at its expected size, suggesting that RIN4 was also cross-linked (Figure 4c, right).

As mentioned above, we decided that it was necessary to use cross-linking in our procedure. In general, cross-linking is an attractive option for revealing interactions that involve membrane proteins because it allows use of harsh detergent conditions to solubilize membrane proteins without disrupting protein–protein interactions (Gingras et al., 2007; Miernyk and Thelen, 2008). However, cross-linking may reduce the specificity of identification of protein complex components. As the spacer arm length of DSP is 12 Å, proteins in the vicinity of RPS2 could be cross-linked to RPS2 even if they do not form complexes with RPS2. In view of this, protein samples from transgenic plants expressing an unrelated PM protein tagged with HPB would serve as better negative controls than those from plants in which no HPB-tagged protein was expressed. Some proteins that are cross-linked to RPS2 only due to their close proximity could also be cross-linked to the unrelated PM protein. Such proteins could be excluded from the list of potential interactors. In any case, proteins identified by our procedure that uses cross-linking need to be considered as candidates for components of the RPS2 complex until they are verified by other methods.

Identification of potential RPS2::HPB protein complex components

We conducted large-scale purification procedures using RPS2::HPB mcca and mcca control plants. (Figure 5a) shows a flow chart describing the major steps of the procedure. Briefly, microsomes were collected from 30 g Arabidopsis leaf tissue, cross-linked with 1 mm DSP, solubilized with 0.5% sodium deoxycholate and then subjected to binding to streptavidin magnetic beads. The captured proteins were then eluted using SDS sample buffer containing a strong reducing agent, 2-mercaptoethanol, and used for SDS–PAGE. A long run of SDS–PAGE resulted in well-separated protein bands visualized using silver staining (Figure 5b). The mcca control samples contained only a few weak non-specific bands, except for one major non-specific band marked by an arrow (Figure 5b). Compared with a known amount of GUS::HPB protein (35 ng) purified from an E. coli strain expressing it, more than 50 ng of RPS2::HPB was estimated to be purified (Figure 5b). Multiple bands that were specific to the RPS2::HPB sample were observed (Figure 5b), which represent RPS2 complex component candidates.

Figure 5.

 Purification of RPS2::HPB protein complex(es).
(a) Flow chart of the purification procedure.
(b) Silver staining of purified samples. Purified protein samples from RPS2::HPB mcca and non-transgenic mcca lines were separated on a long 7.5% SDS–PAGE gel and subjected to silver staining. As a standard for protein amounts, 35 ng of GUS::HPB was included. A major non-specific band is indicated by an arrow, and RPS2::HPB is marked by an asterisk.
(c) Identification of peptides from RPS2 and the known interactor RIN4 by LC-MS/MS. All the identified peptides are underlined. An ‘M’ labeled with an asterisk indicates oxidation of methionine.

To identify the potential components of the RPS2::HPB protein complex(es), we subjected the RPS2::HPB mcca sample and the mcca control sample to tryptic digestion and LC-MS/MS analysis (Link et al., 1999). A total of eight biological replicates of each sample were analyzed, and an inclusion list for targeted detection of peptides with specific m/z values was used in the last two replicates for protein identification (see below). To decrease the false discovery rate, we applied the following criteria to select potential RPS2 complex components: (1) identified in some replicates of the RPS2::HPB sample but never in any replicates of the control sample, with protein identification probability >50%; (2) identified at least three times with a protein identification probability of 90% or higher; (3) identified at least twice with coverage of two or more peptides (except for RIN4, which was identified once with three peptides and three times with one peptide). This list included RPS2 and its known interactor RIN4. Other proteins on the list were all considered as potential RPS2 protein complex components, although the number of replicates in which they were identified varied (Table 1). RPS2::HPB was identified in every replicate. However, RIN4 was not identified every time. We reasoned that this was due to a limitation of the identification method, which measures ionized trypsin-digested peptides from a complex protein mixture. RIN4 is a rather small protein (approximately 23 kDa) and has relatively few peptides available for detection, thus it has a relatively low chance of being identified.

Table 1.   RPS2 and co-purified proteins
Protein nameAGI no.Molecular weight (theoretical)No. times identifiedMolecular function/biological processc
First six experimentsaLast two experimentsb
  1. The order of proteins is in decreasing order of confidence.
    aThe number of times that a particular protein was identified in the first six experiments.
    bThe number of times that a particular protein was identified in the last two experiments for which the inclusion list was used.
    cThe molecular functions/biological processes listed are mainly according to TAIR (http://www.arabidopsis.org/).
    dIncluded in the inclusion list for the last two experiments.

Disease resistance protein RPS2At4g26090104626.862Protein binding/defense response
RPM1-interacting protein 4 (RIN4)dAt3g2507023353.122Protein binding/defense response
Receptor-like kinase (RLK)dAt4g08850112090.322Kinase/receptor-like kinase
Aquaporin PIP1.2dAt2g4596030580.540Water channel/response to water deprivation and salt stress
Phototropin 1 (PHOT1)dAt3g45780111673.831Kinase, protein and FMN binding/response to blue light
Phototropin 2 (PHOT2)At5g58140102457.922Kinase and FMN binding/response to blue light
Band 7 family proteindAt1g6984031387.421Putative hypersensitive-response protein
Band 7 family proteindAt3g0129031303.130Putative hypersensitive-response protein
Patellin-1 (PATL1)At1g7215064027.121Phosphoinositide binding/membrane trafficking
Epithiospecifier modifier 1(ESM1)dAt3g1421044043.430Carboxylesterase/glucosinolate carbolic process
Heavy metal ATPase 3 (HMA3)At4g3012081933.230ATPase/metal ion transport

To test this possibility, we produced an inclusion list containing the mass/charge (m/z) data for theoretical tryptic digests of RIN4 and a number of other proteins that were arbitrarily chosen based on previous experiments (Table S1). We analyzed the last two biological replicates by LC-MS/MS, in which the ions with the m/z values specified in the inclusion list (Table S2) were preferentially selected based on the first MS results for the subsequent analysis, while other ions detected as peaks in the first MS were also analyzed. In this way, RIN4 was identified specifically in both replicates of the RPS2::HPB sample. Identification of other proteins was not much improved by use of the inclusion list. This could be because only an incomplete set of peptide information compared to RIN4 (Table S1) was included in the inclusion list, or because they were not recovered in sufficient quantity for identification. The number of peptides that can be on the inclusion list is limited because if there are too many peptides on the list it does not sufficiently narrow the m/z selections. Thus, we consider an inclusion list as a means to validate known candidates. We also detected non-specific proteins present in both the RPS2::HPB and control samples, which are listed in Table S4.

Survey of potential RPS2 complex components

The potential RPS2 complex components are listed in Table 1. Some of them may function in RPS2-mediated resistance. Although they are no more than component candidates, this discussion could help prioritize them for further research.

One aquaporin PM intrinsic protein (PIP) (At2g45960) was identified. Some other aquaporin PIPs were also detected, but did not pass the strict criteria used for the candidate list in Table 1 (Table S3). Aquaporin PIPs transport water (Kaldenhoff and Fischer, 2006; Maurel, 2007). As plant cells undergoing an HR lose water rapidly, aquaporin PIPs may play a direct role in the HR response. Alternatively, the pathogens may try to manipulate aquaporin PIPs for their own benefit. For example, stomata are the main passage for entry of bacterial pathogens into plants, and aquaporins may be involved in ABA-induced closure of stomata (Cui et al., 2008). Pathogen effectors might modify aquaporins to keep the stomata open. RPS2-mediated resistance has a positive effect in maintaining closure of stomata (Melotto et al., 2006). Aquaporins complexed with RPS2 might be ‘guarded’ (Dangl and Jones, 2001) by RPS2, and modification of the aquaporins by pathogen effectors might trigger RPS2-mediated resistance.

Two band 7 proteins (At1g69840 and At3g01290; 74% identical in amino acid sequences) were identified. Both of them contain the stomatin/prohibitin/flotillin/HflK/C (SPFH) domain which is present in band 7 proteins of both prokaryotes and eukaryotes (Morrow and Parton, 2005; Rivera-Milla et al., 2006). All members of the SPFH domain family are enriched in lipid rafts (Langhorst et al., 2005; Morrow and Parton, 2005). Lipid rafts are specialized PM microdomains that are enriched with sphingolipids and cholesterol, and their possible functions in plants have been discussed recently (Bhat and Panstruga, 2005; Grennan, 2007; Martin et al., 2005). SPFH domain proteins tend to oligomerize (Browman et al., 2007), and are thought to function as scaffolds (Browman et al., 2007; Langhorst et al., 2005). Thus, the band 7 proteins may provide a scaffold for the RPS2 complex in lipid rafts. In plants, band 7 genes are also referred to as hypersensitive-induced reaction (HIR) genes because some members are highly induced in plant tissue that is undergoing an HR (Nadimpalli et al., 2000; Rostoks et al., 2003). However, it is not known how band 7 proteins are involved in R gene-mediated resistance. Aquaporin PIPs (Borner et al., 2005; Mongrand et al., 2004; Morel et al., 2006; Shahollari et al., 2004, 2007) and band 7 family proteins (Borner et al., 2005; Morel et al., 2006) are localized in lipid rafts. As we used chemical cross-linking, identified candidates may not actually be present in RPS2 complexes but merely in the vicinity of RPS2. Even if this is the case, our results suggest possible enrichment of RPS2 in lipid rafts, in which pathogen entry and signal transduction events often occur (Bhat and Panstruga, 2005).

Blue-light receptor kinases phototropin 1 (PHOT1, At3g45780) and phototropin 2 (PHOT2, At5g58140) were identified. PHOT1 and PHOT2 have overlapping functions in phototropism, stomatal opening, chloroplast accumulation, and leaf expansion and movement (Christie, 2007; Lin, 2002). The HR response and SA accumulation triggered by bacterial and viral pathogens are light-dependent, and phytochrome is involved (Chandra-Shekara et al., 2006; Genoud et al., 2002; Griebel and Zeier, 2008; Zeier et al., 2004). It is not known whether phototropin-signaling is also involved in the light dependence of R gene-mediated responses.

Epithiospecifier modifier 1 (ESM1) is involved in glucosinolate breakdown and defense against insects (Barth and Jander, 2006; Zhang et al., 2006). ESM1 appears to have a regulatory function that inhibits production of nitrile while promoting production of isothiocyanate (Zhang et al., 2006). Although no evidence has been observed for involvement of RPS2 in insect defense, co-purification of ESM1 suggests such a possibility.

The potential components that we identified are not necessarily localized in a single protein complex. RPS2 may be a component of multiple complexes. Our approach cannot distinguish different complex species. However, the identities of the components in these multiple species, some of which are specific to particular species, will provide us with a means to distinguish them. As an R protein may guard multiple pathogen effector targets, and as R protein complexes may undergo dynamic changes upon activation, knowledge regarding multiple R protein complex species is important for understanding how the mechanisms of gene-mediated resistance are organized.

Experimental procedures

All the materials generated in this study are freely available for non-profit research.

Plasmid constructs

The sequences of pMDC32-HPB and pMDC-pRPS2-HPB (Figure 1) have been deposited in Genbank as FJ172534 and FJ172535, respectively. Details of construction of the plasmids are described in Appendix S1. All the primers used in the study are listed in Table S1.

Plant lines and bacterial strains

Both rps5 (SALK_015294) and mcca (SALK_137966) are T-DNA insertion mutants (http://signal.salk.edu/cgi-bin/tdnaexpress). SALK_015294 was genotyped using primers LBe, LP1 and RP1, and SALK_137966 was genotyped using primers LBe, LP2 and RP2 (Sessions et al., 2002). The fls2–17 mutant (SAIL_694_C4) is also a T-DNA insertion mutant (Zipfel et al., 2004). The npr1–1 mutant is an EMS mutant with a histidine-to-tyrosine change at residue 334 of the NPR1 protein (Cao et al., 1997). The rpm1 rps2 double mutant was derived from a cross of rps2–101C (Bent et al., 1994) and rpm1–3 (Grant et al., 1995). The Pto bacterial strains carrying avrRpm1 (Dangl et al., 1992), avrRpt2 (Whalen et al., 1991), avrPphB (Simonich and Innes, 1995) or the empty vector pLAFR3 (Staskawicz et al., 1987) were cultured in King’s B medium with rifampicin (25 μg ml−1) and tetracycline (10 μg ml−1). Plant growth, flg22 treatment and bacterial growth assays were performed as described by Tsuda et al. (2008).

Biotinylation efficiency test

For each sample, 0.2 g leaf tissue of 5- to 6-week-old Arabidopsis was flash frozen using liquid nitrogen and ground with a mortar and pestle. Then 400 μl of Extraction buffer [50 mm HEPES, 150 mm NaCl, 10% glycerol (w/v), 1% NP-40 (w/v), 0.5% sodium deoxycholate (w/v), 1 mm DTT, 1 mm PMSF, 1 μg ml−1 leupeptin, 1 μg ml−1 pepstatin, 1 μg ml−1 E-64, Complete Mini protease inhibitor cocktail, pH 7.6 (all the protease inhibitors were obtained from Roche Applied Science, http://www.roche-applied-science.com) was added, and the tissue was ground further. After centrifugation at 16 000 g for 30 min at 4°C, 100 μl supernatant was incubated with 100 μl pre-washed streptavidin-coated Dynabeads® M-280 magnetic beads (Invitrogen) at 4°C for 1 h. After incubation, the beads were washed three times, each of which was performed by a brief vortexing of the tube, with extraction buffer, and the bound proteins were eluted by boiling in 1× SDS sample buffer (62.5 mm Tris/HCl, 2% w/v SDS, 10% glycerol, 1% w/v 2-mercaptoethanol, 0.01% bromophenol blue, pH 6.8) for 10 min.

Purification of RPS2::HPB protein complex(es) and identification of their potential components

All the procedures described in this section were performed at 4°C. For each sample, 30 g of fresh leaf tissue from 5- to 6-week-old Arabidopsis was cut into pieces with scissors and ground with a mortar and pestle in 150 ml grinding buffer (50 mm HEPES-KOH pH 7.5, 10 mm EDTA, 330 mm sucrose, 0.6% polyvinylpolypyrrolidone, 1 mm DTT, 1 mm PMSF, 1 μg ml−1 leupeptin, 1 μg ml−1 pepstatin, 1 μg ml−1 E-64). The homogenate was filtered through double layered Miracloth (Calbiochem, http://www.emdbioscience.com). The filtered fraction was centrifuged at 18 000 g for 10 min, and the supernatant was further centrifuged at 100 000 g for 1 h. The microsome pellet was resuspended in resuspension buffer (20 mm HEPES-KOH pH 7.5, 1 mm EDTA, 330 mm sucrose, 1 mm PMSF, 1 μg ml−1 leupeptin, 1 μg ml−1 pepstatin, 1 μg ml−1 E-64) to a total volume of 6 ml. DSP (100 mm; Pierce, http://www.piercenet.com) was added to the microsome suspension to a final concentration of 1 mm, and the suspension was incubated on a rotator for 30 min. The cross-linking reaction was quenched by adding 1 m Tris/HCl (pH 7.5) to a final concentration of 50 mm, and incubated on a rotator for 30 min. The cross-linker-treated microsomes were collected by centrifugation at 100 000 g for 30 min, and resuspended in resuspension buffer to a final volume of 6 ml. Proteins were solubilized from the microsomes by adding 10% w/v sodium deoxycholate to a final concentration of 0.5%, and incubated on a rotator for 30 min and then centrifuged at 100 000 g for 30 min. Bead binding was conducted by incubating 300 μl pre-washed Dynabeads® M-280 with solubilized supernatant on a rotator for 1 h. The streptavidin beads were washed three times with RIPA buffer 1 (50 mm Tris/HCl pH 7.4, 150 mm NaCl, 1 mm EDTA, 1% NP-40, 0.5% sodium deoxycholate, 1 mm PMSF, 1 μg ml−1 leupeptin, 1 μg ml−1 pepstatin, 1 μg ml−1 E-64) and then three times with RIPA buffer 2 (same as RIPA buffer 1 except 50 mm NaCl is used instead of 150 mm NaCl). The bound cross-linked proteins were cleaved and eluted by boiling the beads in 120 μl 1× SDS sample buffer with 5% v/v fresh 2-mercaptoethanol for 10 min. Details of the complex component candidate identification and electrophoresis and blotting of proteins are described in Appendix S1.


We thank John Ward, Anton Sanderfoot and William Gray for technical advice, Jeff Dangl (Department of Biology, University of North Carolina) for anti-RIN4 antibody and the rpm1 rps2 double mutant, David Mackey (Department of Plant Cellular and Molecular Biology, Ohio State University) for the fls2 mutant, Brian Staskawicz (Department of Plant and Microbial Biology, University of California) for an immunoblotting protocol, Bruce Witthuhn and the University of Minnesota Center for Mass Spectrometry and Proteomics for technical assistance on LC-MS/MS, Jane Glazebrook for critical reading of the manuscript, and members of the Glazebrook and Katagiri laboratories for discussion. This work was supported by a grant from the National Science Foundation (Arabidopsis 2010 grant number IBN-0419648) to F.K.