Assembly of Subtype 1 Influenza Neuraminidase Is Driven by Both the Transmembrane and Head Domains*

Background: Influenza neuraminidase is thought to function as a tetramer, but what drives tetramerization is unknown. Results: The neuraminidase transmembrane domain (TMD) contributes to the assembly process by tethering the stalk to the membrane in a tetrameric conformation. Conclusion: The enzymatic head and TMD coordinate the proper assembly of neuraminidase. Significance: Single-spanning TMDs can contribute to the assembly of distal domains. Neuraminidase (NA) is one of the two major influenza surface antigens and the main influenza drug target. Although NA has been well characterized and thought to function as a tetramer, the role of the transmembrane domain (TMD) in promoting proper NA assembly has not been systematically studied. Here, we demonstrate that in the absence of the TMD, NA is synthesized and transported in a predominantly inactive state. Substantial activity was rescued by progressive truncations of the stalk domain, suggesting the TMD contributes to NA maturation by tethering the stalk to the membrane. To analyze how the TMD supports NA assembly, the TMD was examined by itself. The NA TMD formed a homotetramer and efficiently trafficked to the plasma membrane, indicating the TMD and enzymatic head domain drive assembly together through matching oligomeric states. In support of this, an unrelated strong oligomeric TMD rescued almost full NA activity, whereas the weak oligomeric mutant of this TMD restored only half of wild type activity. These data illustrate that a large soluble domain can force assembly with a poorly compatible TMD; however, optimal assembly requires coordinated oligomerization between the TMD and the soluble domain.

Proteins are assembled from one or more domains that can function independently, cooperatively, or as part of a complex. Through mutations, domain insertions, or deletions, the function of a protein can be altered by changing its topology, folding, localization, activity, or potential interacting partners (1,2). RNA viruses exploit all of these processes to create variation and functional diversity within the genomes of their progeny for host adaptation (3)(4)(5).
The type A influenza surface antigen neuraminidase (NA) 3 was initially described as a three domain protein based on its appearance in electron micrographs (6). The largest domain is the enzymatic head, which is tethered to the viral membrane by a filamentous stalk domain connected to an N-terminal transmembrane domain (TMD) (7). The enzymatic head domain has been well characterized and is known to facilitate viral release and prevent interparticle aggregation by removing the cell surface receptor (sialic acid) that HA associates with to initiate viral entry (8 -11). The large amount of biochemical and structural data available for NA makes it an ideal substrate to investigate how a TMD contributes to the assembly of a distal ectodomain into a functional enzyme.
Nine different NA subtypes (N1-N9) have been found in type A avian influenza strains and structures have been solved for most of the subtypes, which demonstrate the enzymatic head domain forms a tetramer (12)(13)(14). In contrast, no structural data are available for the NA TMD that is part of the N-terminal signal anchor sequence and is known to link the head domain to the viral envelope by a length-variable stalk (15)(16)(17). The changes in stalk length are thought to regulate the distance of the enzymatic head domain with respect to the cell surface receptor (18 -24). How insertions and deletions occur within the stalk domain without altering the assembly of the head domain is not known.
NA oligomerization occurs rapidly and efficiently within the endoplasmic reticulum, possibly through an N-terminal directed process (25)(26)(27). However, the direct role of the TMD in NA assembly has received little attention. This is likely a result of the initial observation that NA head domains retained ϳ100% of their initial enzymatic activity after they were isolated from influenza particles by proteolytic cleavage within their stalk domain (12,28).
To investigate how the NA TMD contributes to the assembly of the distal head domain into a functional sialidase, we used a systematic molecular and biochemical approach. Our results demonstrate the TMD assembles into a tetrameric conformation and is required for NA to acquire sialidase activity but not for its synthesis and trafficking. Removal, membrane tethering, or facilitating the tetramerization of the stalk domain all restored activity to NA, which explained that the TMD functions to tether the stalk to the membrane in a tetrameric conformation. Taken together, these data illustrate how the assem-bly of the TMD coordinates with that of the ectodomain to promote the proper assembly of a type II membrane protein.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfections, and Harvesting-HEK 293T, HeLa, and MDCK from ATCC were cultured in DMEM with 10% FBS, 100 units/ml penicillin, and streptomycin and maintained at 37°C in a 5% CO 2 humidified incubator. For each transfection, 1.5 g of plasmid DNA was incubated 20 min with 5 l of LT-1 (Mirius) in 500 l of Opti-MEM (Invitrogen). Cells were trypsinized from an ϳ90% confluent 10-cm dish, sedimented (500 ϫ g, 5 min), resuspended in 12 ml of Opti-MEM 10% FBS, and 1 ml of cells were added to each transfection mixture before seeding on 3.5-cm dishes. 48 h post-transfection, cells were washed with cold PBS, pH 7.4, and harvested by scraping in 150 l of lysis buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 1% n-dodecyl ␤-D-maltoside, 10 mM N-ethylmaleimide, 1ϫ protease inhibitor mixture (Sigma)). The whole cell lysates were sonicated on ice for 30 s and sedimented (20,000 ϫ g, 5 min), and the post-nuclear supernatants were retained. For the secreted constructs, cells were cultured in Opti-MEM 1% FBS, and the medium was retained and clarified by sedimentation (20,000 ϫ g, 5 min).
Plasmids and Constructs-NA mammalian expression vectors were created by PCR overlap cloning (29) using pCDNA3.1A-Myc-His (Invitrogen) containing full-length NA from influenza A/WSN/33 (H1N1) fused with the C-terminal Myc-His tag. NA-TM X-453 constructs were created by excluding residues 35-38, 35-46, 35-56, and 35-66 from NA-WT. For the secreted NA (ssNA  ) stalk truncations, the cleavable BiP signal sequence (ss) was fused to residue 32, 39, 47, 57, 67, or 73 in NA (WSN33 numbering). Soluble dimeric, trimeric, and tetrameric GCN4 leucine zipper coiled-coil domains (30) were inserted after the ss cleavage site to create ss Di/Tri/Tet NA  . NA 1-74 was made by excluding residues 75-453. glycophorin A (GpA) and G83I TMD chimeras involved swapping their 17-residue TMDs with residues 7-24 from NA-WT. The GALLEX constructs were derived from the pBLM-GpA/GpA-G83I plasmids kindly provided by Dirk Schneider (Johannes Gutenberg University Mainz). The indicated TMD test segments were amplified from the related NA construct in pCDNA and fused between LexA and MalE by overlap PCR cloning. Point mutations were created by site-directed mutagenesis, and all constructs were verified by sequencing (Eurofins MWG Operon).
Blue Native PAGE and Immunoblotting-Post-nuclear supernatants for blue native PAGE (BN-PAGE) were diluted to 85 l with ACA buffer (750 mM amino-n-caproic acid, 50 mM Bis-Tris, pH 7.0), 15 l of 5% G250 Coomassie stain was added, and the samples were incubated 10 min at 4°C prior to resolving on a 5-15% BN-PAGE gradient gel. A pre-transfer at 15 V for 10 min was performed to remove the Coomassie staining followed by transfer to a PVDF membrane. Denaturing immunoblot samples were diluted in non-reducing or reducing (0.1 M DTT) Laemmli sample buffer, heated to 37°C for 5 min, separated by SDS-PAGE, and transferred to a PVDF membrane. Membranes were blocked with milk/PBST (3% nonfat dry milk, PBS, pH 7.4, 0.05% Tween 20) for 30 min and processed using standard immunoblotting protocols with a monoclonal mouse anti-Myc antibody (Cell Signaling). Immunoblots were developed with ECL-prime (GE Healthcare), images were acquired with a CCD camera (Fuji) and quantified using Multigauge software (Fujifilm).
Neuraminidase Activity Assay and Kinetic Analysis-Sialidase activity was measured using 2Ј-(4-methylumbelliferyl)-␣-D-N-acetylneuraminic acid as described previously (31). Briefly, standardized samples were diluted to equivalent amounts using lysis buffer or medium for secreted samples and brought up to 190 l in reaction buffer (0.1 M potassium phosphate, pH 6.0, 1 mM CaCl 2 ). Reactions were initiated by adding 10 l of 2 mM 2Ј-(4-methylumbelliferyl)-␣-D-N-acetylneuraminic acid, and the activity was monitored at 30-s intervals for 30 min at 37°C with excitation and emission wavelengths of 365 and 450 nm in a SpectraMax Gemini EM. The rates were determined using a methylumbelliferyl standard, and these were then normalized for total protein from reducing SDS-PAGE immunoblots to calculate the percent activity. Na 2 CO 3 Extraction, Endo-H, and PNGase F Treatments-Transfected cells were washed with ice-cold PBS, scraped with 150 l homogenization buffer (PBS pH 7.4, 1 mM EDTA, 0.2 M sucrose, 2X protease inhibitor mixture), and passed 30 times through a 27-gauge needle. Unbroken cells and nuclei were removed by sedimentation (2000 ϫ g, 10 min, 4°C). The isolated vesicle and cytosol fraction was incubated 30 min on ice with 800 l of 0.1 M Na 2 CO 3 (pH 11.3) and then subjected to ultracentrifugation (180000 ϫ g, 20 min). The resulting pellet containing the membrane fraction was resuspended with reducing Laemmli sample buffer and the supernatant containing the soluble fraction was precipitated with 20% TCA before dissolving in sample buffer. Endoglycosidase H (Endo-H) and peptide-n-glycosidase F (PNGase F) treatments were performed on 20 l of transfected cell post-nuclear supernatants according to the manufacturer's instructions (New England Biolabs) except that 5% SDS was used in place of 10ϫ glycoprotein denaturing buffer (32).
GALLEX Assay-SU101 Escherichia coli cells with a constitutively expressed lacZ gene regulated by the LexA operator were kindly provided by Dirk Schneider (Johannes Gutenberg University Mainz) with an established protocol (33). SU101 cells were freshly transformed with each pBLM expression plasmid and grown overnight in the presence of antibiotics. Cultures were back diluted in LB to A 600 nm ϭ 0.1, induced with 0.05 mM isopropyl 1-thio-␤-D-galactopyranoside at 37°C for 2 h until A 600 nm Ϸ 0.6. ␤-galactosidase activity was determined as described previously (34).
Sequence analysis and statistics-All of the NA influenza A protein sequences were retrieved from the NCBI Influenza Virus Resource Database (http://www.ncbi.nlm.nih.gov/ genomes/FLU/) in April 2012. The TMDs (residues 7-34) were determined using the ⌬G predictor for biological hydrophobicity (35). Stalk length was calculated from residue 35 to the first residue not resolved in the available subtype 1 NA crystal structures of the head domain, eight amino acids before the conserved large loop disulfide bond joining the N and C termini (13,14,36). Glycosylation frequencies were calculated by the number of NX(S/T) motifs in the unique N1 stalk domains.
All error bars represent one S.D. from three independent experiments.

RESULTS
A Minimal Stalk Length Is Required for Proper NA Assembly-Influenza neuraminidase is composed of an enzymatic head domain connected to a stalk and TMD that is located in the signal anchor sequence (Fig. 1A). Across and within many neuraminidase subtypes (N1-N9), the stalk region has substantial length variation with respect to host and its corresponding hemagglutinin (H1-H16) subtype. To display the stalk length variation within human and avian N1 sequences, we defined the TMD as residues 7-34 with the ⌬G predictor for biological hydrophobicity (35) and the beginning of the head domain as the first amino acid resolved in the different N1 structures (13,14,36). For each N1 sequence, the unique stalk domain between the TMD and head region was identified, grouped by host, hemagglutinin subtype, and length (Fig. 1B). The N1 stalk lengths varied from 28 -48 residues in human strains and 23-48 residues in avian strains with obvious host and HA subtype specific properties.
What allows N1 to undergo stalk insertions or deletions and how do they affect production, trafficking, and activity? To investigate this, a quantitative transfection-based system was established using Myc-tagged N1 from the human H1N1 strain WSN33 (supplemental Fig. 2, A-C). The tag did not impair NA-WT plasma membrane trafficking, or its enzymatic properties, as the sialic acid binding defined by the Michaelis constant (K m ) and processing (V max ) were similar to NA in WSN33 influenza particles (supplemental Fig. 1, A and B). Furthermore, the sialidase activity profiles followed the same trends upon Using this system, a series of constructs were analyzed where increasing numbers of residues were removed directly following the TMD at position 34 to examine how stalk deletions affect NA activity (Fig. 1D). The constructs were given the nomenclature NA-TM X-453 were X is the residue in NA from WSN33 (either 39, 47, 57, or 67) that was fused to the TMD.
NA-WT, NA-TM 39 -453 , and NA-TM 47-453 had similar activity levels, and all of these constructs resolved as oxidized dimers on nonreducing SDS-PAGE because they possessed Cys-49 in the stalk that forms an intermolecular disulfide bond ( Fig. 1E and supplemental Fig. 2B). NA-TM 57-453 , which lacks Cys-49, was observed as an oxidized monomer and possessed ϳ55% of NA-WT activity. NA-TM 67-453 , which lacks the entire stalk region, showed substantial misfolding as the head domain formed incorrect intramolecular disulfide bonds and aggregated (Ox Agg ), explaining its lack of activity. On BN-PAGE, all of the constructs possessing sialidase activity predominantly resolved as tetramers slightly below the soluble 440 kDa marker (Fig. 1F). Together, these results indicate that NA assembly is not dramatically affected by changes in the stalk and that a minimal stalk length of 10 amino acids from the membrane is needed for the head domain to fold.
Production of Enzymatically Active NA Is Impaired in the Absence of the TMD-To address the contribution of the NA TMD to the production of functional NA, residues 32-453 were attached to a cleavable signal sequence (ss) (Fig. 2A). ssNA  was appropriately targeted to the endoplasmic reticulum and properly cleaved as it was glycosylated, present in the soluble fraction following Na 2 CO 3 extraction, and accumulated in the medium (Fig. 2B). Compared with equal amounts of NA-WT, the ssNA 32-453 in the medium inefficiently oxidized to a dimer and possessed only ϳ2.7% of the NA-WT activity ( Fig. 2C and supplemental Fig. 2D). As expected, the ssNA 32-453 associated with the cell had even lower activity. Thus, the NA TMD performs a major role in the production of functional NA when the stalk is present but is dispensable for synthesis and trafficking.
Removal of the NA Stalk Domain Restores Activity in the Absence of the TMD-Stalk truncations were examined in the absence of the TMD using ssNA to determine what portion of the stalk contributed to the minimal enzymatic activity of the secreted ssNA   (Fig. 2, A and D). All of the constructs accumulated in their respective medium and resolved on nonreducing SDS-PAGE as intermolecular disulfide-bonded dimers or partially SDS-resistant dimers at ϳ120 kDa (ssNA 57-453 and ssNA 67-453 do not possess Cys-49). The small N-terminal stalk deletions provided subtle but not significant increases in the activity of the secreted ssNA 39 -453 and ssNA  compared with the secreted ssNA  . Surprisingly, larger stalk deletions dramatically increased the activity as the secreted ssNA 57-453 possessed ϳ20-fold higher activity than equal amounts of the secreted ssNA  , and the secreted stalkless NA (ssNA 67-453 ) possessed ϳ12-fold higher activity. For the secreted ssNA 67-453 , this was a dramatic contrast from the nonfunctional NA-TM 67-453 . Only when six additional residues were removed (ssNA 73-453 ), was the majority of activity and secretion lost. Similar, but much lower activity profiles were also obtained from the cell associated fraction for these constructs (supplemental Fig. 2E). These results indicate that in the absence of the TMD, NA either has lower enzymatic activity when the stalk domain is present or is impaired in its proper assembly.
The NA Signal Anchor Sequence Tetramerizes and Traffics Correctly-How does the TMD contribute to the assembly and activity of NA? To investigate this, a construct was created (NA 1-74 ) where the head domain was removed leaving only the TMD and stalk region. NA 1-74 received the two expected N-linked glycans, and similar to NA-WT, some species were Endo-H resistant, and all were PNGase F-sensitive (Fig. 3A). This indicated that NA 1-74 targets to the endoplasmic reticulum and likely forms a stable conformation that enables it to pass the endoplasmic reticulum quality control system and traffic through the Golgi (37).
The NA TMD is too small for oligomeric analysis by BN-PAGE, so a cysteine cross-linking approach was used. As a test, Ile-48 and Asn-50 in NA-WT were mutated to Cys. On nonreducing SDS-PAGE, both NA-I48C-WT and NA-N50C-WT resolved as intermolecular disulfide-bonded dimers, trimers, and tetramers at the appropriate molecular weight, and these species collapsed to a single band following reduction with DTT (Fig. 3B). This indicated that tetramerization could be observed by this approach, but it was not a quantitative representation of the proportion of tetrameric NA-WT (see supplemental Fig. 2C). NA 1-74 and NA 1-74 -I48C were examined similar to NA-WT and NA-I48C-WT, with the exception that PNGase F was utilized to remove the mixed glycosylation species (Fig. 3C compare with supplemental Fig. 3A). Similar to NA-WT, NA 1-74 appeared as a monomer and oxidized dimer on nonreducing SDS-PAGE, and NA 1-74 -I48C possessed the additional trimeric and tetrameric disulfide-bonded species, which demonstrates that the TMD and stalk region tetramerize independently of the head domain (Fig. 3C).
N-terminal Dimeric, Trimeric, and Tetrameric Domains Progressively Restore Secreted NA Activity-In the absence of TMDs with precisely known oligomeric states, we tested how a soluble N-terminal tetramerization domain contributed to NA assembly in comparison with trimerization and dimerization domains. To do this, well defined 32 residue dimeric (Di), trimeric (Tri), and tetrameric (Tet) GCN4 coiled-coil domains (30) were inserted after the signal sequence cleavage site in ssNA  and fused to the N terminus of the stalk (Fig. 4A).
The secreted ss Di NA 32-453 , ss Tri NA  , and ss Tet NA 32-453 all formed oxidized dimers (Fig. 4B). When equal amounts of secreted ssNA  , ss Di NA  , ss Tri NA  , and ss Tet NA 32-453 were analyzed, a trend of progressive increases in activity were observed with the secreted ss Tet NA 32-453 having ϳ20-fold more activity than ssNA   (Fig. 4B). A similar pattern was also observed in the cell lysates but with lower activity (supplemental Fig. 3B). Upon analysis of these secreted constructs by BN-PAGE, it was evident that ssNA 32-453 aggregated and did not form discrete oligomers, explaining its low activity, and that two oligomeric forms were observed for ss Di NA  , ss Tri NA  , and ss Tet NA   (Fig. 4C). Based on the activity profile and the increasing ratio of the larger oligomeric species from ss Di NA  to the ss Tri NA 32-453 and ss Tet NA 32-453 samples, it was concluded that the larger oligomeric species likely represented properly assembled NA. Together, these results suggest that the tetrameric TMD is likely the optimal conformation for proper NA assembly when the stalk domain is present and that the impairment of NA assembly in the absence of the TMD is related to stalk-induced aggregation.
The NA TMD Oligomerizes in a Heterologous System-TMD interactions can be quantitatively measured using the GALLEX system in E. coli (33). In this system, the strength of the TMD interaction is proportional to the decrease in ␤-galactosidase activity. This is based on the requirement for two LexA repressors to bind to the LexA operator for ␤-galactosidase expression to be suppressed (Fig. 5A). First, the NA signal anchor sequence with (NA 2-74 ) and without (NA 2-34 ) the stalk domain were analyzed together with a positive control, the dimeric GpA TMD, and a negative control, the G83I mutation in the GpA TMD that has weak oligomerization (33). Even though less NA 2-34 and NA 2-74 were synthesized, an ϳ60% decrease in ␤-galactosidase activity was observed indicative of strong interactions between the NA TMD (Fig. 5, B and E).
As expected, both NA 2-34 and NA 2-74 integrated into the E. coli membrane (supplemental Fig. 3C). More impressively, NA 2-74 , which possesses Cys-49 in the stalk, formed an oxidized dimer that resulted in the observation of SDS-resistant trimers and tetramers under nonreducing conditions (Fig. 5C). Thus, oligomerization of the NA TMD is conserved in a heterologous system that lacks several eukaryote-specific lipids.
The NA Stalk Can Contribute to TMD Oligomerization-The strength of the NA TMD interaction did not differ when the stalk domain was present. However, the stalk domain could contribute to oligomerization to a lesser extent than the TMD. Thus, residues 7-23 in NA 2-74 were exchanged with the 17 residue dimeric GpA TMD or the G83I mutant TMD (Fig. 5A). The GpA-NA 24 -74 chimera displayed a similar interaction strength as NA 2-74 , lower than GpA, but the G83I-NA 24 -74 chimera interacted stronger than the G83I TMD alone (Fig.  5E). Furthermore, under nonreducing conditions, the GpA-NA 24 -74 and G83I-NA 24 -74 chimeras formed SDS-resistant tetramers similar to NA 2-74 . This suggested that the stalk can contribute to oligomerization in the presence of a TMD (Fig.  5D).
Tethering the Stalk with an Oligomeric TMD Optimizes NA Assembly-To differentiate between the affect of membrane tethering and TMD oligomerization on the synthesis of enzymatically active NA, NA chimeras containing strong (GpA) and weak (G83I) oligomerizing TMDs were created. Analysis of these constructs in 293T-transfected cell lysates revealed that both chimeras formed oxidized dimers and GpA-NA 24 -453 possessed ϳ85% of NA-WT activity, whereas G83I-NA 24 -453 possessed ϳ55% of NA-WT activity (Fig. 6A).
On BN-PAGE, GpA-NA 24 -453 and G83I-NA 24 -453 produced in 293T cells formed tetrameric oligomers but also a stable higher order oligomer above the soluble 669-kDa marker was observed (Fig. 6B). The higher ratio of the non-native oligomeric structure for the G83I-NA 24 -453 chimera potentially explains the difference in activity. In summary, these data show that the NA head domain can properly assemble on its own, but when the stalk domain is present, its N terminus must be stabilized to prevent it from aggregating. Thus, the NA assembly process is optimized such that the TMD tethers the stalk to the membrane in a tetrameric conformation that coordinates with the assembly of the tetrameric head domain.

DISCUSSION
The first type A influenza NA structure was determined using subtype 2 (N2) head domains that were isolated from influenza particles with a protease that cleaved in the stalk and removed the TMD (12,28). The purified head domains retained ϳ100% of the original activity and this structure, combined with later variants, aided in the design of the NA inhibitors zanamivir and oseltamivir, which are the main drugs against influenza (38,39). These results likely explain why many previous studies did not investigate the role of the TMD in NA assembly but instead focused on how mutations within the TMD affect trafficking, apical sorting, viral incorporation, particle shape, propagation, and virulence (40 -44).
Prior to solving the structure of the NA head domain, it was discovered that the stalk length varies within and across NA subtypes (15,45). In nature, stalk deletions of up to 25 amino acids have been found, and these are thought to be an adaptive feature as they are linked to host range, cell tropism, and are heterogeneously present during a single outbreak (17, 18, 20 -23, 46). In support of this, the stalk is believed to position NA with respect to its receptor as stalk insertions can rescue detrimental deletions that do not affect enzymatic activity but instead prevent elution from erythrocytes (16,19,20,24,45). Thus, NA must be able to adapt to various insertions and/or deletions without negatively impacting its assembly and function.
Our results demonstrate that the tetrameric NA head domain from subtype 1 (N1) can properly assemble on its own. However, when the stalk region is present, this intermediate domain requires the stabilization of its N terminus to prevent it from inhibiting proper NA assembly. The stabilization of the stalk is achieved by tethering it to the membrane, ideally in a tetrameric state that matches the head domain. Together, the interactions within the head and TMD likely compensate for any structural alterations stalk insertions, deletions, or mutations could cause during NA maturation. On the other hand, when the TMD is attached to the head domain, a minimal stalk length (ϳ10 amino acids) is required to prevent the TMD from obstructing the head domain assembly.
Tetramerization of the NA TMD was directly investigated by introducing a cysteine residue into a small TMD construct that enabled endoplasmic reticulum-mediated intermolecular disulfide bond formation. An earlier study investigated oligomerization of the NA TMD using a chimera with transferrin. However, only a slight fraction of transferin shifted in density, which was suggestive of tetramerization (47). This observation FIGURE 5. Quantification of the stalk and TMD oligomerization. A, schematic representation of the GALLEX system for measuring TMD interactions with schematics of the TMD segments. The N-and C-terminal orientations are depicted, and the asterisk indicates the G83I mutation in GpA. B, anti-MBP immunoblots illustrating the expression levels of the TMD constructs that were quantified using the GALLEX system (in E). Expressed constructs are indicated with filled circles. C, NA 2-74 forms an intermolecular disulfide bonded dimer (Ox di ) in the bacterial periplasm through Cys-49, resulting in SDS-resistant trimers (SDS R tri ) and tetramers (SDS R tet ) that are sensitive to reduction with DTT. The monomers (mon) are also indicated. D, representative immunoblots displaying the expression levels and SDS R tet formation of NA 2-74 , GpA-NA 24 -74 and G83I-NA 24 -74 from the SU101 E. coli used to quantify their interactions (in E). E, bar graph displaying the relative ␤-galactosidase activity, which inversely correlates to the strength of the TMD segment interaction (n ϭ 3). supports our conclusion that in the presence of a less compatible TMD, a large ectodomain can force the assembly process, but optimal assembly involves coordination between the TMD and the large ectodomain.
Previous work demonstrated that NA rapidly dimerizes within ϳ2.5 min and this can occur co-translationally (25,27). Additionally, conformation specific antibodies revealed a unique time lag of ϳ5 min between NA oligomerization and when NA reached its native tetrameric conformation (25). These findings, combined with our results, suggest that the TMD likely aids in the assembly process both co-translationally and post-translationally. Co-translationally, the TMD could facilitate the dimerization by driving the initial interaction between a nascent and fully synthesized monomer. Post-translationally, the time lag between NA tetramerization and folding into its native conformation could be explained by the requirement for two dimers to meet and undergo conformational changes in both the head and the TMD. To distinguish between these possibilities the faces or residues that drive the TMD, interactions must first be identified.
There are nine NA subtypes, and all of these are believed to function as a tetramer. This raises the question of whether the contribution of the TMD to NA assembly is conserved, and if so, can an analysis of the different subtypes reveal what mediates the TMD interaction? It is possible that the requirement for the TMD is specific to N1 as a secreted N2 was previously shown to tetramerize and possess activity, but a direct comparison to the wild type version was not made (48). Currently, no NA TMD structures are available to aid in understanding these questions.
It is interesting to speculate what selective advantage oligomerization of the NA TMD provides influenza. The data presented here suggests the strong TMD interactions would increase NA stability and hence activity in different host envi-ronments. On the negative side, the increased TMD stability could make the stalk more immunogenic and antibodies to the stalk region could possibly disrupt NA tetramerization and function similar to removing the TMD. If this is true, it would explain the high mutation rate within the stalk domain and the correlation between stalk length and the number of glycosylation sites (about two N-linked glycosylation sites are found for every 10 amino acids above a length of 30), which can mask antibody binding domains (Fig. 6C).
More and more evidence is accumulating that single-spanning TMDs perform multiple functions such as specific lipid binding that can alter protein function, trafficking, or processing (49,50). This study alone raises several questions regarding whether the NA TMD interaction varies between strains and subtypes, does TMD oligomerization or the strength of the oligomerization change enzymatic or viral properties? Or more generally, how conserved is the principle of matching oligomeric states between the TMDs from type I and type II membrane proteins with that of their ectodomain? By continuing to utilize the significant sequence and biochemical data associated with multifaceted viral proteins, it is likely that more functions attributed to TMDs will become evident.