|Home | About | Journals | Submit | Contact Us | Français|
Microtubules and microtubule-associated proteins are fundamental for multiple cellular processes, including mitosis and intracellular motility, but the factors that control microtubule-associated proteins (MAPs) are poorly understood. Here we show that two MAPs—the CLIP-170 homologue Bik1p and the Lis1 homologue Pac1p—interact with several proteins in the sumoylation pathway. Bik1p and Pac1p interact with Smt3p, the yeast SUMO; Ubc9p, an E2; and Nfi1p, an E3. Bik1p interacts directly with SUMO in vitro, and overexpression of Smt3p and Bik1p results in its in vivo sumoylation. Modified Pac1p is observed when the SUMO protease Ulp1p is inactivated. Both ubiquitin and Smt3p copurify with Pac1p. In contrast to ubiquitination, sumoylation does not directly tag the substrate for degradation. However, SUMO-targeted ubiquitin ligases (STUbLs) can recognize a sumoylated substrate and promote its degradation via ubiquitination and the proteasome. Both Pac1p and Bik1p interact with the STUbL Nis1p-Ris1p and the protease Wss1p. Strains deleted for RIS1 or WSS1 accumulate Pac1p conjugates. This suggests a novel model in which the abundance of these MAPs may be regulated via STUbLs. Pac1p modification is also altered by Kar9p and the dynein regulator She1p. This work has implications for the regulation of dynein's interaction with various cargoes, including its off-loading to the cortex.
Microtubules are critical for a number of basic cellular processes. They are vital to the operation of the mitotic spindle. They can act as ropes that pull on kinetochores to separate the attached chromosomes. Microtubules also function as tracks that guide the transport of various cargoes to distinct destinations within the cell. Many aspects of microtubule function are regulated by distinct classes of microtubule-binding proteins.
In the yeast Saccharomyces cerevisiae, microtubules are critical for positioning the mitotic spindle, a process governed by two genetic systems termed the Kar9 pathway and the dynein pathway. The Kar9 pathway orients cytoplasmic microtubules into the bud, and the dynein pathway pulls on the cytoplasmic microtubules to position the spindle across the bud neck with the onset of anaphase (Kahana et al., 1995 ; Yeh et al., 1995 ; Carminati and Stearns, 1997 ; Miller et al., 1998 ). In the dynein pathway, both Bik1p and Pac1p help localize dynein to the plus end of the microtubule before it is “off-loaded” to the cortex (Sheeman et al., 2003 ; Lee et al., 2005 ; Caudron et al., 2008 ; Markus et al., 2011 ). The microtubule-binding protein Bik1p is a member of the CLIP-170 family, and Pac1p is a member of the Lis1 family.
Bik1p is a multifunctional protein. Bik1p is found on both nuclear and cytoplasmic microtubules (Berlin et al., 1990 ; Sheeman et al., 2003 ; Miller et al., 2006 ). It stabilizes microtubules, and in its absence, microtubules are very short (Berlin et al., 1990 ; Blake-Hodek et al., 2010 ). It also acts to tether microtubules to the cell cortex in mating cells (Molk et al., 2006 ). Whereas the mammalian CLIP-170 binds growing microtubules only, the yeast Bik1p tracks both the growing and shrinking ends of microtubules (Carvalho et al., 2004 ). In the nucleus, Bik1p is also part of the kinetochore, and its activity there has been linked to preanaphase kinetochore separation in polyploid strains (He et al., 2001 ; Lin et al., 2001 ). However, its precise role at kinetochores remains unclear (Westermann et al., 2007 ). Bik1p localization at the plus end of cytoplasmic microtubules is mediated by the kinesin motor, Kip2p, which carries it to the plus end (Carvalho et al., 2004 ; Caudron et al., 2008 ).
Bik1p and Pac1p are each structurally similar to their mammalian counterparts. Bik1p is composed of an amino-terminal head domain, a central coiled-coil domain, and a carboxy-terminal tail (Miller et al., 2006 ). In its head domain, Bik1p has one CAP-Gly domain for microtubule binding, whereas CLIP-170 has two. This is followed by a serine-rich domain, which may be used in phosphoregulation (Miller et al., 2006 ). A 40–amino acid region at the carboxy-terminus termed the cargo-binding domain is required for interaction with dynein and Pac1p (Lin et al., 2001 ; Sheeman et al., 2003 ; Li et al., 2005 ). The structure of Pac1p includes a short coiled-coil domain and a highly conserved WD40 repeat domain. Neither of the two domains alone is sufficient for microtubule binding or plus-end tracking in vivo (Markus et al., 2011 ).
Mutations in Lis1 cause lissencephaly, a devastating disease of abnormal brain development resulting from incomplete neuronal migration (Reiner et al., 2006 ). Lis1 has a number of activities associated with microtubules. Lis1 function is closely linked to dynein, a minus end–directed motor protein that regulates a number of different movements within the cell, including vesicle transport, mitosis, cell migration, nuclear migration, and spindle orientation (Faulkner et al., 2000 ; Dujardin et al., 2003 ; Shu et al., 2004 ; Tanaka et al., 2004 ; Tsai et al., 2005 , 2007 , 2010 ; Lam et al., 2010 ; Zhang et al., 2010 ). As an adaptor to dynein, Lis1 modulates a subset of dynein functions (Faulkner et al., 2000 ). Lis1 also interacts with the motor domain of dynein and affects its motor activity (McKenney et al., 2010 ; Torisawa et al., 2011 ; Huang et al., 2012 ). In doing so, it increases the time that dynein interacts with the microtubule, making dynein more persistent in generating force for transport of heavy loads (McKenney et al., 2010 , 2011 ). Furthermore, Lis1 coordinates the activity of plus end–directed motors with minus end–directed motors (Yi et al., 2011 ). Lis1 can also affect microtubule assembly, both positively or negatively (Sapir et al., 1997 ; Han et al., 2001 ). In the upper layers of the epithelium, Lis1 is important for the organization of cortical microtubules and the stability of desmosomes (Sumigray et al., 2011 ). In yeast, Pac1p functions in the dynein pathway by working with Bik1p to recruit dynein to the plus end of the microtubule (Lee et al., 2003 ; Sheeman et al., 2003 ; Li et al., 2005 ; Markus et al., 2011 ). Although much is known about the various activities of Lis1, little is known about the regulation of Lis1 and/or Pac1p in each of these different contexts.
Proteins in both the Kar9p and dynein pathways have been shown to be regulated by phosphorylation (Choi et al., 2000 , 2002 ; Liakopoulos et al., 2003 ; Maekawa and Schiebel, 2004 ; Moore et al., 2006 , 2008 ). However, little is known about how other protein modifications regulate spindle positioning.
Sumoylation is a small ubiquitin-like modification that regulates many divergent cellular processes and is key to several diseases (reviewed in Dasso, 2008 ; Gareau and Lima, 2010 ; Sarge and Park-Sarge, 2011 ; Praefcke et al., 2012 ). In yeast, the single gene for SUMO is SMT3. SUMO conjugation onto target lysines occurs through a terminal glycine residue, which becomes exposed by the action of the Ulp1p protease, which removes the last three amino acids of SUMO. A cascade of enzymes analogous to, yet distinct from, ubiquitination is responsible for SUMO conjugation to its substrates. The E1 dimer Aos1p–Uba2p activates Smt3p in an ATP-dependent step. The E2-conjugating enzyme is Ubc9p. Four E3 ligases are known in yeast that also confer substrate specificity. These are Nfi1/Siz1p, Siz2p, Nse2p, and Mss21p (Johnson and Gupta, 2001 ; Reindle et al., 2006 ; Duan et al., 2011 ; Heideker et al., 2011 ; Stephan et al., 2011 ).
Removal of SUMO from targets is a dynamic process, and SUMO deconjugation is important for cellular health (Bekes et al., 2011 ). Three proteases remove SUMO from its targets in yeast. These are Ulp1p, Ulp2p, and Wss1p (Mukhopadhyay and Dasso, 2007 ; Kolli et al., 2010 ; Mullen et al., 2010 ). Ulp1p is an essential protease required for cell cycle progression that localizes to the nuclear envelope and septin ring (Li and Hochstrasser, 1999 ; Elmore et al., 2011 ). The Ulp2p has chain-editing activity and can cleave SUMO chains, acting to shorten their length (Bylebyl et al., 2003 ; Mukhopadhyay et al., 2007 ; Yeh, 2009 ). The protease Wss1p displays SUMO-isopeptidase activity and is reported to be associated with the proteasome (Mullen et al., 2010 ). Ulp1p and Ulp2p are SUMO specific, whereas Wss1p may remove both ubiquitin and SUMO (Mullen et al., 2010 ; Su and Hochstrasser, 2010 ).
In 2007, a new mode of cross-talk was identified between the ubiquitination and sumoylation systems (Sun et al., 2007 ; Uzunova et al., 2007 ; Xie et al., 2007 ). SUMO targeted ubiquitin ligases (STUbLs) are ubiquitin ligases that recognize a sumoylated protein and polyubiquitinate it, sending the target to the proteasome for degradation. How STUbLs interact with microtubule-binding proteins is not known.
Recent work suggests that sumoylation regulates the Kar9p spindle-positioning pathway by modifying the Kar9 protein (Leisner et al., 2008 ; Meednu et al., 2008 ). Defects in sumoylation alter the localization of Kar9p to both spindle poles rather than just the pole destined for the bud (Leisner et al., 2008 ; Meednu et al., 2008 ). In the present study, we show that two proteins from the dynein pathway, Bik1p and Pac1p, interact with SUMO and several proteins from the SUMO conjugating and processing system. We also show that the SUMO protease Ulp1p controls the amount of shifted Pac1p. Both Bik1p and Pac1p also interact with the STUbL enzyme dimer Nis1p–Ris1p. Taken together, our findings suggest that these microtubule-associated proteins may be regulated by the SUMO signal transduction system. This could have implications for the regulation of CLIP-170 and Lis1 family members in other systems.
Bik1p interacts physically with Kar9p, and Kar9p is sumoylated (Moore et al., 2006 ; Leisner et al., 2008 ; Meednu et al., 2008 ). We therefore investigated whether proteins in the dynein pathway might also be regulated by sumoylation. We tested for an interaction between Smt3p and two proteins that function in the dynein pathway, Bik1p and Pac1p, by two-hybrid analysis. As shown in Figure 1, both BIK1 and PAC1 interacted with SMT3. BIK1 did not interact with ubiquitin encoded by UBI4 or the small ubiquitin-related protein encoded by URM1. In contrast, PAC1 also displayed a slight but consistently detectable interaction with ubiquitin. As previously reported, the kinesin motor protein encoded by KIP2 did not interact with SMT3 (Meednu et al., 2008 ). Thus two-hybrid analysis of three members of the ubiquitin superfamily suggests that Bik1p and Pac1p interact strongly with Smt3p.
The protease Ulp1p is responsible for removing the carboxyl three amino acids from full-length Smt3p. This exposes a glycine at position 98 for conjugation to substrates. The absence of this terminal glycine precludes conjugation to target proteins (Johnson and Blobel, 1997 ). We therefore tested whether two mutations would abrogate the interaction. The first mutation deletes the last three amino acids (ATY), exposing the terminal diglycine motif (SMT3-GG). The second mutation replaces the terminal glycine with alanine (SMT3-GA), preventing conjugation. As shown in Figure 2A, full-length SMT3 and the truncated form of SMT3, SMT3-GG, interacted with both BIK1 and PAC1, but the SMT3-GA form did not. Previous work demonstrated that the SMT3-GA construct was expressed (Meednu et al., 2008 ). This suggests the possibility that conjugation mediates the two-hybrid interactions of Bik1p and/or Pac1p with Smt3p.
Several enzymes in the sumoylation pathway facilitate the transfer of SUMO to target proteins. Kar9p was previously shown to interact with several of these by two-hybrid analysis (Meednu et al., 2008 ). Both BIK1 and PAC1 interacted with UBC9 encoding the E2 enzyme and NFI1/SIZ2 encoding an E3 enzyme (Figure 2B). Thus two proteins from the dynein pathway interact with multiple enzymes required for sumoylation.
A yeast two-hybrid interaction can be explained most simply by a direct binding of the bait and prey proteins. However, if a third protein binds between the bait and prey, it might activate transcription of the reporter gene and falsely suggest a direct interaction. Bik1p interacts with two other spindle-positioning proteins, Kar9p and Bim1p (Schwartz et al., 1997 ; Moore et al., 2006 ; Wolyniak et al., 2006 ), both of which also interact with Smt3p (Meednu et al., 2008 ). To investigate whether either Kar9p or Bim1p could serve as a bridge in the two-hybrid interaction between Bik1p and Smt3p, we tested the two-hybrid interaction in a reporter strain deleted for either of these genes. As shown in Figure 3, A and andB,B, neither KAR9 nor BIM1 is required for the interaction. We also tested whether PAC1 was required for the interaction of Kar9p with the SUMO pathway. As shown in Figure 3C, no difference was observed for the interaction in the pac1Δ and wild-type strains, suggesting that the interaction of Kar9p with sumoylation proteins does not require Pac1p.
Because Bik1p also interacts with Pac1p (Sheeman et al., 2003 ), we next considered the simple bridging model in which Pac1p might bridge the two-hybrid interaction between Bik1p and Smt3p or vice versa. To test this, we deleted either BIK1 or PAC1 from the two-hybrid reporter strain. To our surprise, the deletion of either gene resulted in the loss of Smt3p's interaction with the reciprocal protein (Figure 3, D and andE).E). This result is not consistent with a simple bridging model in which Bik1p bridges Pac1p and Smt3p or Pac1p bridges Bik1p and Smt3p. Instead, we speculate that the cointeraction of Bik1p with Pac1p promotes a synergistic interaction with Smt3p. Work from the Pellman lab showed that the cargo-binding domain of Bik1p is required for its two-hybrid interaction with Pac1p (Sheeman et al., 2003 ), suggesting that this domain may mediate the Bik1p–Pac1p interaction. To gain insight into the Pac1p– Smt3p interaction, we deleted the cargo-binding domain of Bik1p from the two-hybrid reporter strain. The absence of the cargo-binding domain greatly diminished the interaction between PAC1 and SMT3 (Figure 3F). This finding is consistent with Smt3p interacting with a complex of Bik1p–Pac1p. Further testing of this model will require the identification of a PAC1 allele that does not interact with Bik1p.
To determine whether Bik1p is a target for sumoylation, we analyzed Bik1p in an in vitro sumoylation assay using recombinant Smt3p, Ubc9p, and Uba2p/Aos1p purified from bacteria. Bik1p isoforms were separated on SDS–PAGE and analyzed by Western blotting. Two, and possibly three, higher–molecular weight forms of Bik1p were observed in the test lane (Figure 4A, lane 1) but not in the control reactions, each lacking the respective component of the assay (lanes 2–6). These results are consistent with two possibilities. Bik1p sumoylation may occur on at least two sites or polysumoylation might occur by the formation of SUMO chains.
Sumoylation frequently occurs on a lysine residue contained within the consensus motif, ΨKxD/E, where Ψ is a hydrophobic amino acid, x is any amino acid, and D/E is an aspartic or glutamic acid. Bik1p contains one ΨKxD/E consensus sequence, WKPD at K180, and two examples of a less well-conserved KxD/E motif, GKND at K46 and KKLEE at K373 and K374. Mutation of the four lysines to arginine within these three locations, as well as a mutation in an additional lysine at position K251, did not alter the shift seen in vitro (unpublished data). This indicates that nonstandard sites may be used in the sumoylation of Bik1p.
Next we investigated whether Bik1p is sumoylated in the cell. Bik1p-hexahistidine (his6) was isolated from cells either overexpressing Smt3p or containing an empty vector. Analysis by immunoblotting revealed a shifted form of Bik1p that was present only when Smt3p was overexpressed (Figure 4B, panel 1, lanes 1 and 3). This shifted band corresponded exactly to a band revealed in an identical blot electrophoresed simultaneously and probed with polyclonal Smt3p or anti-hemagglutinin (HA). This suggests that overexpression of Smt3p results in Bik1p sumoylation.
Ulp1p is a SUMO protease that cleaves SUMO from sumoylated targets. Inhibition of Ulp1p is predicted to increase the sumoylation levels of Ulp1p substrates (Li and Hochstrasser, 1999 , 2000 ). To test this idea for Bik1p, we examined the levels of shifted Bik1p in ulp1-ts at the nonpermissive temperature of 37°C. However, despite numerous attempts using a variety of different conditions, we could not detect higher–molecular weight forms of Bik1p in this strain, even when Smt3p was overexpressed for 2 h (unpublished data).
We next investigated whether Pac1p would shift in the ulp1-ts strain. Unlike Bik1p, Pac1p displayed an extensive ladder of slower-migrating forms at the nonpermissive temperature (Figure 5, compare lanes 3 and 6). When additional Smt3p expression was induced, the level of shifted forms of Pac1p was pronounced even at the permissive temperature (compare lanes 1 and 4). Few shifted forms of Pac1p were observed in a wild-type ULP1+ strain prepared under the same conditions (compare lanes 9 and 12). These results suggest that these higher–molecular weight forms of Pac1p are caused either directly or indirectly by the presence of SUMO.
To determine whether the shift of Pac1p mobility was due to multiple monosumoylation events or polysumoylation, we mutated to arginine the three lysine residues within Smt3p that are required for chain formation (Bylebyl et al., 2003 ). Analysis of the Pac1p shift using SMT3-K11R-K15R-K19R revealed that the higher–molecular weight smears were greatly reduced (Figure 6A, compare lanes 1 and 7). This suggests that polysumoylation contributes to the higher–molecular weight shifts of Pac1p.
Although Pac1p shifts in a Ulp1p-dependent manner, this does not eliminate the possibility that other modifications might also play a role. To test whether ubiquitin might also contribute to the shifts, we transformed a ubiquitin plasmid into the ulp1-ts strain containing Pac1p. This resulted in some higher–molecular weight bands of Pac1p, but the shift was not as extensive as when SMT3 was present (Figure 6A, compare lanes 5 and 7). Taken together, these findings suggest that both polysumoylation and ubiquitination contribute to the higher–molecular weight forms of Pac1p.
To further explore whether the Pac1p shifts observed in the ulp1-ts strain are due to SUMO conjugation, we pulled down Pac1p-his6 on nickel beads from a yeast strain expressing Smt3-GG, the processed form (see Materials and Methods). The unmodified form of Pac1p appeared as a doublet, perhaps indicating that Pac1p is phosphorylated. Four to seven higher–molecular weight forms of Pac1p were observed after the pull down (Figure 6B). Two of these cross-reacted with anti-FLAG, which marks SUMO, indicating that these two shifts are due to SUMO conjugation of Pac1p (Pac1-S). To investigate whether the nonreactive bands might be due to another modification, we probed a third identical panel with anti-ubiquitin. This revealed that both of the anti-FLAG bands also reacted with anti-ubiquitin. These findings suggest that Pac1p is altered by at least two types of posttranslational modification—ubiquitin and SUMO. Moreover, two of the bands that were nonreactive with anti-FLAG did react with anti-ubiquitin. This suggests the possibility that some forms of Pac1p may contain only ubiquitin.
Pac1p functions in the dynein pathway to help recruit dynein to the plus end of the cytoplasmic microtubule before its off-loading to the cortex (Li et al., 2005 ; Markus et al., 2011 ). To determine whether the integrity of the dynein pathway affects the modification of Pac1p, we examined Pac1p in several deletion strains related to various aspects of dynein function. KIP2 encodes a kinesin that transports Bik1p to the plus end of the microtubule (Carvalho et al., 2004 ). DYN1 encodes the dynein motor itself, and JNM1 encodes a component of the dynactin complex (McMillan and Tatchell, 1994 ). NDL1 encodes a NudE homologue that helps localize dynein to the plus end (Li et al., 2005 ; Moore et al., 2008 ). NUM1 encodes a cortical protein important for the off-loading of dynein (Heil-Chapdelaine et al., 2000 ; Farkasovsky and Küntzel, 2001 ). In a NUM1-deleted strain, the off-loading of dynein to the cortex is blocked (Lee et al., 2005 ). In each of these delete strains, little or no additional shift in Pac1p was observed (Figure 7A). We next tested a strain deleted for SHE1, which encodes a regulator of the dynein–dynactin interaction (Woodruff et al., 2009 ; Markus et al., 2011 ; Bergman et al., 2012 ). Surprisingly, Pac1p shifted significantly. This finding suggests that She1p is a novel inhibitor of Pac1p modifications.
To investigate further, we analyzed the shift of Pac1p in strains deleted for members of the Kar9 pathway, which orients the cytoplasmic microtubule into the bud (Miller and Rose, 1998 ; Miller et al., 1998 ). BIM1 encodes an EB1 homologue and microtubule-binding protein that binds Kar9p (Lee et al., 2000 ; Miller et al., 2000 ). KAR9 encodes a linker protein between BIM1 and the type V myosin Myo2p (Miller et al., 1999, 2000 ; Hwang et al., 2003 ). The movement of Myo2p along an actin cable serves to guide the attached microtubule into the bud (Beach et al., 2000 ; Yin et al., 2000 ). KIP3 encodes a type 8 kinesin in the Kar9 pathway that controls microtubule length (Miller et al., 1998 ; Su et al., 2011 ). Surprisingly, an increase was observed in the shift of Pac1p (Figure 7B). This suggests that the Kar9 pathway inhibits the interaction of Pac1p with Smt3p and/or ubiquitin. This is consistent with earlier findings that Bim1p interacts with She1p by two-hybrid analysis (Wong et al., 2007 ).
The results from Figure 7A could in theory be explained by She1p blocking the addition of either SUMO or ubiquitin. To distinguish between these two possible mechanisms, Pac1p-his6 from she1Δ and wild-type strains was enriched on nickel beads, and identical immunoblots were analyzed by probing with anti-SUMO or anti-ubiquitin (Figure 7C and 7C′). Consistently, larger amounts of Pac1p were purified from equal amounts of protein extract in the she1Δ and ulp1 strains than from wild type. This indicates, but does not prove, that these modifications may affect the steady-state stability of Pac1p. Furthermore, when anti-SUMO was used, nearly identical amounts of reactivity were observed in the she1Δ and wild-type (WT) strains. This suggests that increased sumoylation is not the cause of the higher–molecular weight forms of Pac1p observed in the she1Δ strains. However, when probed with anti-ubiquitin, additional reactivity was observed. This suggests that the addition of ubiquitin contributes to the higher forms of Pac1p seen in she1Δ. How She1p blocks this addition of ubiquitin remains an active avenue of further investigation.
SUMO-targeted ubiquitin ligases can link the sumoylation and ubiquitination pathways by ubiquitinating proteins already modified by sumoylation, thus promoting their removal by the proteasome (Sun et al., 2007 ; Uzunova et al., 2007 ; Xie et al., 2007 ). Wss1p was originally identified as a weak suppressor of a temperature-sensitive allele of Smt3p (Biggins et al., 2001 ). Recently Wss1p was shown to be a SUMO-dependent isopeptidase that promotes the targeting of SUMO-conjugated proteins to the proteasome (Mullen et al., 2010 ). Our previous work showed that Kar9p interacted with both WSS1 and the complex encoded by the RIS1 and NIS1 genes (Meednu et al., 2008 ). Therefore we tested whether Bik1p or Pac1p would also interact with the STUbL enzyme Nis1p–Ris1p and the SUMO iospeptidase Wss1p by two-hybrid analysis. Indeed, both did (Figure 8A).
These findings suggest the possibility that the shifts of Pac1p may be modulated by a STUbL enzyme and the proteasome. If such a model is correct, higher–molecular weight forms of Pac1p should be present when the STUbL is absent from the cell. To assess this, we examined the levels of shifted Pac1p in strains deleted for WSS1, RIS1–NIS1, or another STUbL enzyme complex, SLX5–SLX8. As shown in Figure 8B, the shifted forms of Pac1p were significantly increased in strains lacking the catalytic subunit of the STUbL Ris1p and the Wss1p isopeptidase. No difference in the shift pattern was detected in strains deleted for SLX5 or SLX8, suggesting that the shift in Pac1p is specific to only one of the two STUbLs present in yeast. These findings also indicate that Pac1p may be a substrate of the Ris1p STUbL enzyme. Further work is required to determine whether STUbLs can exert an effect on the unmodified levels and/or steady-state levels of Pac1p.
Together these results suggest the possibility that ubiquitination of Pac1p by the Nis1p–Ris1p STUbL modulates the turnover of sumoylated Pac1p. Wss1p may aid in the delivery to or the processing of Pac1p by the proteasome. To investigate this, we treated cells with a proteasome inhibitor, MG132. Because yeast cells can pump the MG132 drug out of the cell, this study was conducted in a strain deleted for PDR5, which encodes a drug pump (Fleming et al., 2002 ). As seen Figure 8C, higher–molecular weight forms of Pac1p are present in the stacker portion of the gel when the proteasome is inhibited with MG132 (compare lanes 2 and 3, asterisk). Taken together, these results suggest that the STUbL enzyme Nis1p–Ris1p controls the levels of posttranslational modifications attached to Pac1p.
In this study, we show that the Lis1 homologue Pac1p and the CLIP-170 homologue Bik1p interact with SUMO and several enzymes of the SUMO modification pathway. Pac1p also interacts with ubiquitin. We also show that Bik1p and Pac1p interact with the STUbL enzyme Nis1p–Ris1p and that the posttranslational modifications of Pac1p are controlled by this STUbL enzyme. In summary, we showed that two new classes of conserved microtubule-associated proteins interact with and are likely to be regulated by SUMO. With this work, four different spindle-positioning proteins have now been shown to interact with SUMO.
We use several different approaches to show that Pac1p interacts with SUMO. First, the two-hybrid analysis shows that Pac1p interacts with the GG but not the GA form of SUMO. This implies, but does not prove, that the interaction is due to a conjugation event. Second, inhibition of the SUMO-specific protease Ulp1p results in higher–molecular weight forms of Pac1p. This suggests that at least some of the higher–molecular weight forms are caused by SUMO moieties conjugated onto Pac1p. Indeed, when an additional SUMO is provided to the cell on a plasmid, a similar but significantly stronger banding pattern is seen. The higher–molecular weight forms of Pac1p are likely due to poly-SUMO chain formation, because the shifts are greatly diminished by the presence of a nonchainable form of SUMO. Pac1p also interacts with the STUbL enzyme Nis1p–Ris1p, an enzyme that recognizes sumoylated proteins. A pull-down assay with Pac1p suggests that the higher–molecular weight forms of Pac1p contain covalently attached SUMO. Taken together, these experiments strongly support our assertion that Pac1p is conjugated with SUMO.
Our conclusion that the interaction of Pac1p with SUMO may also involve cross-talk with ubiquitin is derived from four lines of investigation. First, we see a weak two-hybrid interaction of PAC1 with ubiquitin, encoded by UBI4. Pac1p is also likely to be modified by ubiquitin, because higher–molecular weight forms of Pac1p can be generated by the inclusion of a plasmid encoding ubiquitin. Third, ubiquitin copurifies with Pac1p, with some shifted bands cross-reacting with both anti-ubiquitin and anti-SUMO. Fourth, Pac1p interacts with the ubiquitin ligase complex, Ris1p–Nis1p, and deletion of RIS1 results in higher–molecular weight forms of Pac1p. This is consistent with our previous finding that Kar9p and Bim1p interact with Wss1p and the same STUbL (Meednu et al., 2008 ). The finding that multiple spindle-positioning proteins interact with SUMO suggests the possibility that the Nis1p–Ris1p STUbL enzyme may regulate spindle positioning.
This is the first report that a member of the Lis1 family or a member of the CAP-Gly domain family is modified by SUMO. This has significant implications for the regulation of these two classes of microtubule-associated proteins, which are widely conserved across evolution. Prior to this work, only four other microtubule-associated proteins have been shown to be modified by SUMO. These are Tau (Dorval and Fraser, 2006 ; Takahashi et al., 2008 ), Ndc80p (Montpetit et al., 2006 ), CENP-E (Zhang et al., 2008 ), and Kar9p (Leisner et al., 2008 ; Meednu et al., 2008 ). We also showed that the EB1 homologue Bim1p interacts with SUMO by two-hybrid analysis, but details of this interaction and whether this interaction represents an actual conjugation by SUMO remain to be elucidated (Meednu et al., 2008 ). Our findings are also notable in that Pac1p was not found in any of the previous genome-wide screens for sumoylated proteins (Zhou et al., 2004 ), indicating that yeast SUMO-ome may not yet be complete.
Previous reports suggest that the Kar9 pathway for spindle positioning is regulated by sumoylation (Leisner et al., 2008 ; Meednu et al., 2008 ). Leisner et al. (2008) showed that the spindle-positioning defect seen in the nonsumoylatable kar9-4K-R mutant is not as severe as the smt3-331 defect. This suggests that there are additional proteins required for spindle positioning that are also regulated by sumoylation. The findings reported here suggest that Pac1p and/or Bik1p may be the additional protein(s). Thus it is likely that both spindle-positioning pathways in yeast may be regulated by sumoylation. Future studies should elucidate whether this signal transduction system regulates each pathway separately or whether they are coordinated as a unit.
Our two-hybrid bridging data suggest that Bik1p and Pac1p are both required for their mutual interaction with SUMO and Ubc9p. However, this relationship was not observed in the in vitro shift assay using purified Bik1p. This suggests that Pac1p is not required for the in vitro shift of Bik1p. The apparent discrepancy between the two assays may be reconciled if Pac1p enhances the sumoylation of Bik1p. Consistent with this idea is our observation that Bik1p from partially fractionated extracts shifted better in vitro than purified Bik1p (unpublished data). Future work to test this hypothesis is warranted.
Furthermore, the bridging data in which Bik1p is required for the Wss1p's interaction with Pac1p (but not vice versa) suggest that Bik1p may recruit Wss1p to the Bik1p–Pac1p complex (Figure 9). This is consistent with our data that Pac1p is hypermodified in a strain deleted for WSS1.
We propose a model in which Pac1p sumoylation and its subsequent interaction with a STUbL promotes its degradation via the proteasome (Figure 9). In this model, Wss1p aids in debranching of polysumoylated and/or polyubiquitinated Pac1p at the proteasome. Because She1p regulates the association of dynactin with dynein (Woodruff et al., 2009 ), our finding that She1p inhibits the modification of Pac1p suggests the possibility that sumoylation and/or ubiquitination may regulate the interaction of dynein with its accessory proteins. In our model, She1 protein blocks the ubiquitination of Pac1p. It might also promote the proteasome-mediated degradation of modified Pac1p. We speculate that the sumoylation of Pac1p may play a role in regulating the off-loading of dynein to the cell cortex.
Several questions remain. SUMO is induced by various types of cellular stress. Future work is needed to show whether the sumoylation of these key classes of microtubule-associated proteins represents a mechanism by which microtubule-dependent processes are inactivated under conditions that are inhospitable for cell division. Furthermore, both Pac1p and Lis1 serve as adaptors of dynein. Lis1 regulates some but not all of dynein's functions. It remains to be determined whether Lis1, either in neurons or epithelial cells, is also sumoylated. Our findings have implications for how this class of dynein adaptor may regulate dynein function. Because dynein is critical to a number of fundamental processes important for life, it will be important for future studies to elucidate the widespread utility of this modification in other systems.
Two-hybrid analysis was carried out as previously described (Moore and Miller, 2007 ; Moore et al., 2008 ; Meednu et al., 2008 ). All analysis was carried out after 2–3 d of growth at 30ºC. PAC1-DNA-binding domain (BD) was constructed by amplifying PAC1 from pRM3574/pJG13 with BamHI and PstI ends and ligating it into pRM1154/GBDU-C1 cleaved with the same enzymes. This generated pRM3604 (Table 1), which was sequenced to confirm the absence of unintended PCR errors.
PAC1 was deleted from the two-hybrid reporter strain yRM1757 by one-step gene replacement using a PCR product derived from genomic DNA from the pac1Δ::KANR deletion strain (Open Biosystems, Huntsville, AL), yRM3138. This generated pRM6249, which was confirmed by PCR.
BIK1-BD was generated by excising the BIK1 coding fragment with EcoRI and BamHI from pRM2627 and ligating it to the BD vector, pRM1156, cut with the same enzymes. This generated pRM4924.
Bik1p was sumoylated using a previously described protocol (Meednu et al., 2008 ). Strains expressing Bik1p-V5-his6 were grown at 30°C in synthetic complete (SC) media lacking uracil (–ura) containing 2% sucrose, and Bik1p expression was induced by the addition of 2% galactose for 2 h. To prepare yeast extracts, cells were washed once with water, resuspended in cold 1× binding buffer (20 mM Tris-HCl, pH 7.9, 5 mM imidazole, 0.5 M NaCl) containing 0.2% protease inhibitor P8215 (Sigma-Aldrich, St. Louis, MO) and 1 mM phenylmethylsulfonyl fluoride (PMSF; Sigma-Aldrich). Cells were lysed by vortexing with glass beads. Crude extracts were clarified by centrifugation at 13,000 rpm for 30 min, applied to nickel resin (Novagen/EMD, Darmstadt, Germany), and washed with 25 column volumes of 1× binding buffer containing 50 mM imidazole. Bound protein was eluted with 1× binding buffer containing 300 mM imidazole and dialyzed overnight into sumoylation assay buffer (50 mM Tris, pH 7.6, 5 mM MgCl2, 15% glycerol) at 4°C.
Sumoylation enzymes were purified from bacteria as previously described (Meednu et al., 2008 ). Briefly, the processed form of SUMO his6-Smt3p-GG (pRM6713) and Ubc9p-his6 (pRM5169) were purified from bacteria using nickel affinity chromatography (Johnson and Blobel, 1997 ). The E1 components, Aos1p and Uba2p, located on a polycistronic plasmid (pRM6760) were copurified from bacteria using glutathione affinity chromatography (Bencsath et al., 2002 ). The assay was carried out using purified sumoylation enzymes (2 μg of Smt3p-his6, 1 μg of Ubc9p-his6, 1 μg of GST-Uba2p/Aos1p) mixed with Bik1p-V5-his6 (1 μg) in the presence and absence of ATP and an ATP regeneration system at 30°C for 2 h.
Yeast strains (yRM3350) containing pGAL-BIK1-his6 (pRM5487) and 3HA-FLAG-SMT3 (pRM5251) or an empty pGAL plasmid (pRM2908) were grown to mid–exponential phase in SC media lacking uracil and histidine (–ura –his) containing 2% sucrose and then induced with 2% galactose. Extracts were prepared by disrupting cells with glass beads, followed by centrifugation at 13,000 rpm in a 4°C cold room for 30 min. Bik1p was enriched from 3 mg of extract on a nickel column resin (Novagen/EMD). Beads were washed eight times with B150 buffer and twice with B150 buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 0.2% Triton X-100) containing 50 mM imidazole (Sigma-Aldrich). Bound proteins were eluted by the addition of 50 μl of 3× Laemmli buffer and boiled for 5 min. Samples were analyzed by SDS–PAGE (8%) and immunoblotting. Bik1-his6 was detected with mouse anti-his6 (Novagen/EMD). HA-tagged proteins were detected using mouse anti-HA (Santa Cruz Biotechnology, Santa Cruz, CA), and Smt3p was detected with rabbit anti-Smt3p. Secondary antibodies were obtained from Santa Cruz Biotechnology.
Strains expressing SMT3 under the control of a copper-inducible promoter (pRM8023) were grown to mid–exponential phase to an OD600 of <0.4 in liquid SC media lacking uracil and leucine (–ura –leu). Cells were shifted to either 30 or 37°C and simultaneously induced for SMT3 expression by the addition of 1 mM CuSO4 for 2 h. Cells were collected by low-speed centrifugation, washed, and resuspended in cold B150 breaking buffer; excess fluid was removed, and cells were resuspended in B150 breaking buffer containing 1 mM PMSF, 20 mM N-ethylmaleimide, 20 mM 2-iodoacetamide, and 1% Sigma-Aldrich protease inhibitor. Cells were lysed by vortexing with glass beads for 10 min. It should be noted that the inclusion of the alkylating agent 2-iodoacetomide was crucial for obtaining consistent results. Extracts were clarified by centrifugation at 13,000 × g for 20 min at 4°C. Protein concentrations were determined by Bradford Protein Assay (Bio-Rad, Hercules, CA) using bovine serum albumin as a standard.
Yeast extracts were prepared as described earlier by bead beating cells in binding buffer (20 mM Tris-HCl, pH 7.9, 0.3 M NaCl, 10 mM imidazole, 1% protease inhibitor [#P8849; Sigma-Aldrich]). The volume was increased to 14 ml with binding buffer. Extracts were incubated with nickel beads at 4°C on a rotisserie for 1 h. Beads were collected and washed 7× with cold binding buffer containing 50 mM imidazole and 0.5 M NaCl. Samples were prepared for SDS–PAGE and Western blot analysis. The his6 epitope was detected using mouse anti-his6 (Novagen, Madison, WI) at 1:1500 in Tris-buffered saline (TBS) for 3 h at RT. Smt3p was detected using anti-Smt3p (Rockland, Gilbertsville, PA) at 1:2000 in phosphate-buffered saline (PBS) for 3 h at RT. Mouse anti-ubiquitin was obtained from Covance (Princeton, NJ) and used at 1:1500 in TBS for 3 h at room temperature. FLAG-tagged Smt3p was detected using mouse anti-FLAG (#3165; Sigma-Aldrich) at 1:5000 in TBS for 3 h at room temperature.
Western blotting was carried out using 8% SDS–PAGE as previously described (Moore et al., 2006 ; Meednu et al., 2008 ), except that 0.2% Tropix I-Block reagent (Applied Biosystems, Foster City, CA) with 0.1% Tween-20 in TBS was used as the blocking agent.
Point mutations to generate lysine-to-arginine substitutions were introduced into HA-his6-Smt3 plasmid (pRM8023) by site-directed mutagenesis using the QuikChange Kit (Stratagene, La Jolla, CA). The mutant K11R was generated using primers #837 and #838. K15R was generated using primers #839 and #840. K19R was generated using primers #841 and #842. The final construct (pRM8836) was sequenced to confirm the absence of additional mutations. The sequences of these primers are given in the Supplemental Material.
An integration vector for Pac1p was constructed with a his6 tag fused in-frame at its 3′ prime end. A PCR fragment starting at PAC1 bp699 was synthesized by PCR with XhoI and BamHI restriction sites at its termini. This was ligated into the corresponding sites of the vector, pRM2194. This created plasmid pRM9232, which was sequenced to confirm the intended sequence. For integration by one-step gene tagging, the plasmid was linearized with ClaI, and transformants were selected on SC media lacking histidine.
We thank Jake Kline for his artistic interpretations and creativity in designing the artwork for the model. This work was supported in part by funding from the National Science Foundation (MCB-0414768 and MCB-1052174), the Oklahoma Health Research Program of the Oklahoma Center for the Advancement of Science and Technology (OCAST HR09-150S), and start-up funds to R.K.M from Oklahoma State University and the Oklahoma Agricultural Experiment Station. A.A. is supported in part by a fellowship from the Sloan Foundation. J.W.P.K. was supported by a Niblack Research Scholarship and a Wentz Research Scholarship. We also thank the manuscript reviewers for their insightful comments.
This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E12-03-0195 on October 3, 2012.