HMT-1 Exists in a Protein Complex in C. elegans
We first sought to test whether HMT-1 exists as an oligomer in C. elegans. Towards this goal, we generated transgenic worms expressing functional translational HMT-1::GFP fusions and analyzed the HMT-1::GFP in fractioned worm lysates. As would be expected for integral membrane protein, SDS-PAGE and immunoblot analyses of fractionated lysates from phmt-1-hmt-1::GFP worms identified HMT-1::GFP among total and membrane proteins, but not among soluble proteins (). We also fractioned transgenic worms that express transcriptional phmt-1::GFP fusions. Since the GFP polypeptide does not possess membrane-spanning domains, it localized only in the total and soluble, but not in the membrane fraction of proteins ().
HMT-1 forms oligomeric complexes.
We next performed a gel-filtration separation of HMT-1::GFP by fast protein liquid chromatography (FPLC) on a Superose 6HR column (GE Healthcare). We solubilized HMT-1::GFP with perfluorooctanoate (PFO), a mild detergent that preserves interactions between protein subunits and has been successfully used in studies of ABC transporters 
. We also extracted HMT-1::GFP with SDS, a strong ionic detergent, which disrupts protein interactions 
. If HMT-1::GFP exists as a monomer in vivo
, PFO- and SDS-extracted HMT-1::GFP would have identical migration properties. However, if HMT-1::GFP forms higher oligomeric states, the migration properties would be distinct.
We established that the elution peak of SDS-extracted HMT-1::GFP was at 15 ml, corresponding to an estimated Mr of 125,000 (). Since the calculated molecular mass of HMT-1::GFP is 117,000, the observed elution profile () is consistent with the migration of an HMT-1::GFP monomer. In contrast, PFO-extracted HMT-1::GFP was eluted as a broad peak at 13-15.25 ml (). The highest the anti-GFP antibody immunoreactivity was observed in fractions 13.75–14.25 ml, corresponding to estimate molecular masses of 340–269 kDa (). These estimated molecular masses suggest that HMT-1::GFP is present almost exclusively in a protein complex either with itself or/and with other proteins in C. elegans.
Detection of HMT-1–HMT-1 Interactions Using Mating-Based Split-Ubiquitin Yeast-Two-Hybrid System (mbSUS)
The simplest explanation of the gel-filtration data is that HMT-1 homomerizes. To test this we used a mbSUS that detects binary interactions of membrane proteins in vivo 
. To do so, different HMT-1 fusions with ubiquitin were constructed (, ), and interactions were monitored by the release of the artificial transcriptional factor PLV that activated the expression of lexA
-driven reporter genes, ADE2, HIS3
. Based on the predicted membrane topology (TMHMM software, version 2.0 (http://www.cbs.dtu.dk/services/TMHMM-2.0/
), the HMT-1 NH2-terminus is located outside (Lumen
), whereas the COOH-terminus is inside (Cytosol
) (). HMT-1 “bait” vector was generated by fusing the C-terminus of the full-length of HMT-1 with a CubPLV fusion peptide (). Since interactions can be detected only when CubPLV fusion and NubG fusions are in the cytosol 
, we generated two HMT-1 prey constructs by fusing NubG at NH2- or COOH-termini (“NubG-HMT-1” and “HMT-1-NubG,” respectively (). Therefore, if the topology prediction is correct, NubG of HMT-1-NubG -prey fusions would be localized in the cytosol () and would promote detection of interactions. In contrast, NubG of the NubG-HMT-1-prey construct will be localized intra-organellarly that will prevent interaction read-out (). This experimental design assessed the membrane topology of HMT-1 and provided a negative control for spurious interactions. For other controls we co-expressed the following constructs: for detecting false positives due to self-activation, HMT-1-NubG or HMT-1-CubPLV were co-expressed with the empty pMetYCgate or pXNgate vectors respectively; as a positive control of interactions, the potassium channel, KAT1, from Arabidopsis thaliana
, was used (KAT1-Cub-PVL bait and KAT1-NubG prey 
); finally KAT1 was used as bait or prey for showing the specificity of HMT-1 interactions. The interactions were visualized in diploid cells by their ability to grow on SC medium lacking adenine and histidine, and by β-galactosidase activity assays. We ascertained that interactions occur only due to expression of HMT-1-CubPLV, or in case of the positive control, KAT1-CubPLV, by suppressing their expression with methionine ().
NTE is essential, but not sufficient for protein-protein interactions of HMT-1.
Our data show that regardless of whether interactions were monitored as colony formation on selective media or by β-galactosidase activity, interactions did not occur between HMT-1-CubPLV and NubG lacking the HMT-1 insert (). As would be expected for membrane proteins participating in different biological processes, interactions did not occur between C. elegans HMT-1 and A. thaliana KAT-1, regardless of the vector combination used in the study (). Furthermore, HMT-1-HMT-1 interactions were not detected when NubG was placed at the HMT-1 amino terminus. Instead, we detected interactions only when NubG was placed on the carboxyl terminus.
Based on these results we propose that: first, HMT-1 at a minimum can form homodimers; second, the COOH-terminus of HMT-1 localizes in the cytosol since interactions were detected only when CubPLV and NubG were fused at the C-termini. Our data also suggest that the NH2-terminus of HMT-1 may be in the lumen. However, it is also possible that the amino terminal fusion is on the same side of the membrane but is not accessible to the Cub-PLV bait. Additional studies will determine the precise topology of HMT-1.
HMT-1 of C. elegans Increases Cd Tolerance of S. cerevisiae
We showed previously that HMT-1 functions independently of PC synthases in heavy metal detoxification 
. Therefore, HMTs are expected to increase heavy metal tolerance of organisms that lack the capacity to produce PC. Consistent with this, Preveral and colleagues showed that SpHMT-1 increases heavy metal tolerance of E. coli
and S. cerevisiae
, whose genomes lack PC synthase homologs 
. Here we tested whether CeHMT-1 is able to increase Cd tolerance of S. cerevisiae
, and in doing so, test if the interacting construct, HMT-1-NubG, is functional. In this assay, we compared Cd sensitivity of THY.AP5 yeast expressing the empty pNXgate 32/33-3HA (NubG) vs.
THY.AP5 expressing HMT-1-NubG.
We first established that THY.AP5 cells expressing the empty vector were sensitive to 50 µM CdCl2 and their sensitivity increased with increasing Cd concentration in the culture medium (). In contrast, THY.AP5 cells expressing HMT-1-NubG were more tolerant to Cd and were able to grow at a concentration of Cd (75 µM), that blocked growth of yeast cells expressing vector without HMT-1 cDNA insert ().
These findings show that: first, HMT-1-NubG construct is functional because it increases Cd tolerance in S. cerevisiae; second, the ability of CeHMT-1 to increase heavy metal tolerance in S. cerevisiae, whose genome lacks PC synthase homologs, complements our previous findings and observation of others that HMTs act independently of PC synthases, reinforcing the remarkable conservation of HMTs' function in metal detoxification.
The N-terminal Extension Domain (NTE) is Essential, but not Sufficient for HMT-1-HMT-1 Interactions
Since the N-terminal extension domain (NTE) is a conserved structural feature of HMT-1 proteins, unique to this half-transporter family, we expected that it may be critical for HMT-1 function. Therefore, we used the mbSUS approach to test if deletion of the NTE domain would affect the ability of HMT-1 to interact with itself. Truncated HMT-1, lacking NTE (designated ΔNTE) or possessing NTE only (designated NTE), were fused at their C-termini with NubG of pXNgate21-3HA vector (ΔNTE-NubG, NTE-NubG, respectively, ) and co-expressed with HMT-1-CubPLV as described above. Interactions were assayed by monitoring colony formation of serially-diluted cell inocula on SC medium lacking adenine and histidine, or by β-galactosidase activity. Interactions were suppressed by supplementing SC medium with methionine ().
We found that regardless of the approaches used for the analysis, interactions occurred only when a full-length HMT-1-NubG or -CubPLV fusions were co-expressed in diploid yeast cells. Interactions were significantly suppressed when a full-length HMT-1-CubPLV was co-expressed with HMT-1-NubG lacking NTE (). Minimal, methionine-repressible cell growth relative to the negative control (HMT-1-CubPLV + NubG) was observed when ΔNTE-NubG prey construct was co-expressed with the full-length HMT-1-CubPLV bait. However, the β-galactosidase activity was detectible only when a full-length bait and prey constructs were co-expressed (). Cell growth and the β-galactosidase activity assays can be interpreted to mean that the full-length HMT-1 and HMT-1 lacking NTE do not interact, or that interactions are very weak and are at the limit of detection.
Failure to detect HMT-1 self-association was not due to decreased expression or stability of the truncated HMT-1 since it was detected in microsomal membranes by immunoblot analysis (). However, it was possible that the lack of interactions was due to fact that the NTE domain was necessary for targeting, but not for interactions. Therefore, HMT-1 without NTE and full-length HMT-1 might localize to different subcellular compartments. If the latter suggestion is correct, and NTE is dispensable for interactions, than co-expressed in yeast ΔNTE-NubG and ΔNTE-CubPLV constructs might interact since both would be in the same subcellular compartment. However, the interactions occurred only in cells co-expressing a full-length HMT-1-NubG or -CubPLV fusions, but not in cells co-expressing HMT-1-CubPLV and HMT-1-NubG constructs lacking NTE (). These results further support our suggestion that the NTE domain is necessary for interactions, but do not rule out the possibility that it is also important for targeting.
NTE is essential, but not sufficient for the ability of HMT-1 to confer Cd tolerance.
Interactions were not observed when a full-length HMT-1-CubPLV was co-expressed with NTE-NubG construct (). Failure to detect HMT-1 self-association was not due to decreased expression or stability of the NTE domain since it was detected in microsomal membranes by immunoblot analysis ().
Since the NTE alone did not interact with the full-length HMT-1, we concluded that the NTE domain is not sufficient for HMT-1-HMT-1 interaction, and that other structural feature(s) must associate to form a functional transporter.
NTE and Oligomerization are Essential for HMT-1 Function in Cd Detoxification
Since formation of at least a four-domain structure is a pre-requisite for the activity of half-molecule ABC transporters 
, and truncated HMT-1 does not self-associate, we expected that truncated HMT-1 would not confer Cd tolerance in S. cerevisiae
. Consistent with this prediction, the C. elegans
HMT-1 deletion mutant, gk155
, lacks two cytosolic loops and a transmembrane domain of the NTE and is hypersensitive to heavy metals 
To test our hypothesis we compared Cd sensitivity of THY.AP5 yeast expressing the empty pNXgate32/33-3HA (NubG) vector, the full-length HMT-1-NubG, the HMT-1-NubG lacking the NTE domain, or the NTE domain fused to NubG. Cd sensitivity was monitored as described above. We found that only the full-length HMT-1 conferred Cd tolerance of yeast cells (). Since only a full-length HMT-1 is capable of interacting with itself in mbSUS (), we concluded that self-association is essential for HMT-1 functional activity and that the NTE domain plays a structural role, preserving HMT-1 self-association and function in metal detoxification.
HMTs are acutely required for heavy metal detoxification in different species, act by unknown mechanisms, and are distinct from other ABC transporters due to their unique topology: they are half-molecule ABC proteins, and are the only half-transporters with an NTE. Since functional activity of half-molecule ABC transporters requires homo- and/or hetero-oligomerization, in this manuscript we addressed two basic questions. First, does functional HMT-1 oligomerize? Second, what is the role of the NTE domain in HMT-1 function? In this manuscript we present the following observations: first, HMT-1 exists in a protein complex in C. elegans and, as determined by mbSUS, at least homodimerizes in S. cerevisiae. Accordingly, the molecular mass of 269 kDa of PFO-extracted HMT-1-GFP, detected by gel-filtration chromatography, is consistent with its being a dimer (). Nevertheless, since higher molecular mass species were also detected (e.g. 340 kDa, ), it is also possible that HMT-1 trimerizes, or that HMT-1 homodimers associate with other cellular protein(s). Future studies will discriminate between these possibilities and will establish the identity of proteins present in the HMT-1-associated protein complex. Second, the NTE domain plays a structural role and is necessary, but is not sufficient for HMT-1 homomerization, suggesting that other structural components must associate with HMT-1 to form a functional transporter. Third, the ability of HMT-1 to homomerize is needed for its function in Cd detoxification. Since the presence of NTE and heavy metal detoxification function of HMTs are conserved across species, we speculate that in addition to serving a structural role, the NTE domain possesses transport activity, and directly contributes to heavy metal detoxification.
The functional significance of the NTE has been studied in some full-molecule ABC transporters of the ABCC subfamily 
. For instance, the NTE domain of a full-molecule ABC transporter, S. cerevisiae
YCF1, is needed for vacuolar membrane trafficking and transport 
. In contrast, the NTE domain of ABCC1/MRP1 has redundant trafficking signals with the COOH-region 
, but regulates its homodimerization 
. Unlike CeHMT-1, the NTE of ABCC1/MRP1 is both essential and sufficient for oligomerization. We do not know whether the NTE domain of CeHMT-1 is required only for self-association or, as shown for full-molecule ABC transporters of ABCC subfamily, is also involved in membrane trafficking and transport activities. The prominence of heavy metals as environmental toxins and the remarkable conservation of HMT-1 structural architecture and function in different species reinforce the value of continued studies of HMT-1 in model systems for identifying functional domains in HMTs of humans.