A putative amphipathic α-helix in BMV 1a is sufficient for membrane association
Previously, using membrane affinity and protease sensitivity assays, we showed that BMV 1a strongly localizes to the cytoplasmic face of the ER membrane despite lacking any detectable trans-membrane domain 
. Membrane flotation assays of 1a deletion derivatives and GFP-fusion to truncated versions of 1a showed that a 105 amino acid (aa) region (aa 368–472, previously designated region E, ) plays a major role in 1a-ER membrane binding 
. In this region, a stretch of 35 amino acids (aa 388–422) is predicted to be predominantly α-helical. Within this helical region, a putative amphipathic α-helix core peptide of 18 amino acids (aa 392–409) can be recognized, which we will refer to as “helix A”. One indication that helix A is likely important is that its amino acid sequence is evolutionarily highly conserved among the equivalent 1a replication proteins of other bromoviruses ().
Evolutionarily conserved helix A is sufficient to direct membrane association of GFP.
To test the functionality of helix A for membrane association, the 105, 35 and 18 aa regions described above were fused to the N-terminus of GFP to produce E-GFP, 35H-GFP, and 18H-GFP, respectively (). Lysates of yeast cells expressing these fusion proteins were loaded under flotation gradients, which upon centrifugation were fractionated and analyzed by SDS PAGE and western blotting using anti-GFP antibodies. As a measure of membrane association, flotation efficiency was determined as the percentage of total GFP or 1a-GFP fusion protein in the gradient that was present in the top two fractions. In these assays, less than 3% of wild type cytosolic GFP floated to the top of the gradient with the membrane fraction. Fusing the 35 aa region to GFP greatly increased membrane association up to 45%, which was as efficient as membrane association directed by the full 105 aa E region fused to GFP. The smaller 18H-GFP fusion protein retained about 30% flotation efficiency (). Thus, the 35 aa segment 388–422 accounts for essentially all of 1a's membrane association mediated by domain E, and the 18 aa helix A region retains most of this function and is sufficient to direct membrane association of GFP.
NMR spectroscopy and mutational analysis confirm the α-helical and amphipathic nature of helix A
A helical wheel projection of the 18 aa helix A core region shows that it has the potential to form an amphipathic α-helical cylinder with one side (the right side in ) having a cluster of hydrophobic, non-polar residues including three leucines (L396, L400, L407) and two nearby positive-charged lysines (K403, K406), and the other (left) side of the helix mostly hydrophilic and polar residues (, see also marked aa in ). To test these predictions, we used NMR to resolve the structure of an 18 aa peptide with the core sequence (aa 392–409) of helix A. NMR spectra of this peptide dissolved in water did not reveal a long term stable structure. However, upon including 100 mM SDS to provide lipid bilayer-mimicking micelles 
, the peptide showed NMR spectral changes consistent with a stable conformation (). Based solely on 13
C chemical shifts, NMR showed that aa 397–406 in the peptide had a >80% probability to be in a helical structure (). To elucidate this further, the three dimensional structure of the peptide was calculated based on NOE distance constraints arising from spatial contact of hydrogen atoms observed to be closer than ~5×. Additional dihedral angle constraints were derived from chemical shifts using the TALOS program 
. The resulting structure () shows an α-helical conformation for aa 397–406, indicating that an amphipathic helix formed upon binding to the lipid membrane-mimicking SDS micelle. The constraints and overall quality of the structure are shown in .
Linear and helical wheel projections of helix A.
NMR structure of helix A on SDS micelles.
Table 1 Statistics for the structure determination by NMR from PSVS .
shows that 65% of the observed NMR signals were assigned to specific atoms in the peptide. Of these assigned signals, 80% were affected by the addition of 16-doxyl stearic acid (DSA), a paramagnetic molecule whose presence in SDS micelles causes nearby atoms' NMR signals to broaden and lose intensity, thus serving as an internal probe for the extent to which atoms on the surface of a labeled structure are immersed in the micelles 
. In parallel with the distribution of hydrophobic amino acid residues ( and ), the N-terminal half of the peptide had a larger percentage of assigned atoms that showed DSA contact than the C-terminal half, i.e. 91% vs. 69%, respectively ( and ).
Titration results for BMV-1a helix A bound to SDS micelles with 16-DSA.
Since the structure and DSA results implied that L396, L400, and L407 were positioned in the face of helix A most deeply immersed into the bilayer-mimicking micelle (, bottom view, and ), we tested the importance of these three leucines for helix A-mediated membrane association. We introduced L to A mutations in the 18H-GFP fusion protein expression plasmid and tested their effects on membrane flotation efficiency. As shown in , the wt18H-GFP again had 30–35% flotation efficiency, while single L to A mutations reduced this to ~7–15%. Of the three leucines, mutating the more N-proximal L396 and L400 more severely reduced membrane association than mutating L407, which paralleled the stronger micelle contact of the N-terminal half of the peptide ( and ). These results might also explain in part the tolerance for an isoleucine at the 407-equivalent position in other bromovirus replicase proteins (). A fusion protein with a combination of all three L to A mutations had near background level flotation, implying a complete loss of function of helix A in targeting cytosolic GFP to membranes. In contrast, K to E mutations reversing the charge of lysines 403 and 406 (the only basic residues in the 18 aa helix A core) showed K403E to only marginally decrease the flotation efficiency of 18H-GFP, while K406E had no significant effect (), consistent with the NMR observation that these amino acids have weak and no lipid contact, respectively (). A double K to R mutation designed to retain the positive charge at these amino acid positions did not affect membrane association at all (), suggesting that K403 might contribute to membrane association via its positive charge, perhaps by neutralizing negatively charged lipid head groups. Overall, as mutations that change the leucine-rich non-polar face of the helix have more detrimental effects on membrane association than other amino acid substitutions, the results were consistent with the NMR-based structure of helix A and show that amphipathic helix A has a key role in membrane targeting.
Membrane flotation analysis of helix A-GFP fusion protein and its mutant derivatives.
Helix A is required for efficient membrane association of full-length 1a
To extend the results from helix A-GFP fusion proteins, the contribution of helix A to membrane association of full-length 1a was assessed using biochemical and cell imaging approaches. By membrane flotation gradient analyses, the flotation efficiency of wt 1a was ~ 96% (), confirming 1a's previously established high affinity for membranes 
. Deleting the 35 aa or 18 aa helices reduced 1a-membrane association by over two-fold (). The three L to A mutations, either as single mutations or as a triple combination, similarly reduced the flotation efficiency of full length 1a to ~45%. Single alanine insertions immediately downstream from L396 and L400 reduced flotation efficiency to levels similar to full helix A deletions (), confirming the importance of correct spacing to maintain the amphipathic characteristics of helix A. The importance of the charged lysines at positions 403 and 406 at the hydrophilic face of helix A was assessed using alanine or arginine substitutions. Single position substitution mutants and double mutants K403/406A and K403/406R maintained full flotation efficiency (, single mutations not shown). 1a mutants K403E, K406E, and double mutant K403/406E retained intermediate flotation efficiencies showing that although the positive charge at these positions is not required, reversing it to a negative charge destabilizes membrane association (). The K403/406E single and double mutations showed a somewhat greater inhibition of membrane association in the context of full length 1a (~63% for the double mutant in ) than in the context of the 18 aa helix fused to GFP (~77%, ), suggesting the possibility that residues outside of the 18 aa helix core might cooperatively influence membrane association.
Helix A and specific non-polar (L) and polar residues (K) in it are required for efficient 1a membrane binding and ER targeting.
Since none of the deletions and mutations completely abolished 1a-membrane association, we used confocal immunofluorescence microscopy to compare the sub-cellular localization of the 1a mutants with that of wt 1a (). Wildtype 1a localized predominantly to the perinuclear ER membrane, co-localizing almost completely with the distribution of ER marker Sec63p. In contrast, the 1a protein mutants that lacked either the 35aa or 18aa helices no longer co-localized with Sec63p and displayed a mostly diffuse cytoplasmic localization (). Confocal fluorescence images showed similar staining throughout the cytoplasm for 1a triple mutant L396/400/407A and the K403/406E double mutants, although in these cases a minority of 1a retained ER association. By contrast, the K403/406R mutant co-localized with Sec63p throughout, as for wt 1a ().
Combined, the flotation and confocal results demonstrate that 1a has both helix A-dependent and -independent modes of membrane association, but that helix A is crucial for efficient membrane association and normal 1a localization to perinuclear ER membranes. While other aa such as the positively charged lysines contribute, the leucines on the hydrophobic side of helix A are the most important residues for effective association of 1a with ER membranes.
Helix A determines the type and ultrastructure of 1a-induced ER membrane rearrangements
We previously showed that, in the absence of 2aPol
or other viral components, 1a targets itself to perinuclear ER membranes and induces spherular invaginations that by EM are indistinguishable from those that replicate BMV RNA when 1a is expressed together with low 2aPol
levels expressed from the yeast ADH1
. Examples of such spherules are shown in , top left panel. In contrast, 1a plus high 2aPol
levels expressed by the strong yeast GAL1
promoter shift the predominant viral-induced membrane rearrangements from spherules to large, karmellae-like, multilayer stacks of double membrane layers surrounding the nucleus (, top right panel). Although dramatically different in organization, such membrane layers support BMV RNA replication as efficiently as spherules 
Helix A and specific non-polar (L) and polar residues (K) in it are required for 1a-induced ER membrane rearrangements.
shows that deleting helix A (1aΔ35H, 1aΔ18H) abrogated 1a's ability to induce either type of ER membrane rearrangement. Likewise, mutating the hydrophobic face of helix A in triple mutant 1a L396/400/407A or reversing the positive charge of the two lysines in double mutant 1a K403/406E abolished 1a's ability to form either membrane rearrangement, whether expressed alone or together with GAL1-promoter -driven 2aPol (). Arginine substitution of the single negatively charged amino acid E405 maintained a wt phenotype, although alanine substitution at this position resulted in a >30-fold reduction in the number of spherules, showing the importance of a charged, hydrophilic amino acid at this position (data not shown).
In surprising contrast, double mutant 1a K403/406R showed an entirely different phenotype. This mutant, which as described earlier maintained full flotation efficiency (), was revealed by EM analysis to form dramatically more, and somewhat smaller, membrane-bound spherules than wt 1a (). To specify which of the two amino acid changes contributed to this phenotype, single mutants 1a K403R and 1a K406R were generated and expressed in yeast cells. In keeping with the DSA/membrane interaction of K403 but not K406 (), shows that 1a K403R maintained this mutant phenotype while 1aK406R induced spherules with the frequency and size of wt 1a. Moreover, 1a K403R induced high frequency, smaller spherules even in the presence of high levels of GAL1-promoter-driven 2aPol expression (), conditions under which wt 1a preferentially induces ER membrane layers rather than spherules (). These results show both that 1a-ER membrane association through helix A is crucial for 1a-induced membrane rearrangements, and that additional characteristics of helix A have important roles in determining the type of membrane rearrangement and the extent of membrane curvature.
Hereafter, we will refer to helix A mutants that have lost all membrane-rearranging capacity, like triple mutant L396/400/407A, as Class I mutants, and to mutants with the hyper- abundant, smaller spherule phenotype, like K403R, as Class II mutants. To evaluate the possible role of other helix A amino acids in Class I or Class II phenotypes, we first made alanine substitutions at the other residues besides L396/400/407 in the major membrane interacting face of helix A, i.e., F394, T397, Y401, Y404 and T408 (, bottom view; see also ). Strikingly, EM analysis showed that 1a T397A, 1a Y401A and 1a Y404A all were Class II mutants, inducing a plethora of small spherules like 1a K403R (). Flotation analyses showed that all four of these Class II mutants also maintained wt 1a levels of membrane association (). The F394A substitution, positioned on the same side of helix A as the above Class II mutants but at the N-terminal end of helix A ( and ), had a partial Class II phenotype of producing spherules at normal frequency but slightly smaller diameter than wt 1a. An alanine substitution at K403 resulted in a similar phenotype. By contrast, spherules of wt frequency and size were produced by 1a T408A, at the C-terminal end of helix A, by 1a bearing alanine substitutions at residues on the upper face of helix A (), i.e., V392, L398, N399, and Q402. and by 1a A395S (results not shown).
Class II mutations in 1a helix A induce hyper-abundant membrane invaginations.
To more accurately and precisely describe the Class II mutant phenotypes, we measured the abundance and diameter of spherules in the subset of cells that were sectioned through their nuclei among a total of 200 cells for each mutant. As shown in , spherule abundance in yeast cells expressing Class II mutants was 5- to 7- fold higher than in cells expressing wt 1a. Moreover, the average spherule diameter in cells expressing wt 1a was ~66 nm, but was only ~40–55 nm in cells expressing Class II mutants.
Effects of two classes of helix A mutations on 1a-induced ER membrane rearrangements, BMV RNA3 accumulation, RNA replication, 1a and 2aPol accumulation, and localization.
As mentioned earlier, when wt 1a and high levels of 2aPol
are co-expressed, only 15–25% of cell sections with BMV-induced, perinuclear membrane rearrangements show spherules, while 75–85% bear double-membrane layers that support efficient RNA replication 
. Even under such conditions of high 2aPol
expression, the four Class II mutants induced ~6- to 8-fold more spherules than wt 1a and reduced the frequency of cells with double membrane layers by >3- to 10-fold (). Thus, helix A mutations not only alter 1a's intrinsic functions for ER membrane rearrangement, but also the ability of 2aPol
to modulate the type of 1a-induced ER membrane rearrangements.
1a-stimulation of 2aPol accumulation does not require formation of new membrane compartments
In addition to mediating its own membrane association, wt 1a also recruits 2aPol
to the RNA replication complex, mediated at least in part by a direct interaction between 1a's C-terminus and the N-terminus of 2aPol 
. In conjunction with such recruitment in this and previous studies 
, co-expressing wt 1a increased 2aPol
accumulation by approximately two-fold (). Accordingly, we measured 2aPol
accumulation in the presence of the various 1a mutants to determine to what extent this 1a function depended on sequences in helix A. All mutant 1a proteins accumulated to levels similar to wt 1a, but Class I and Class II mutants showed directly opposite effects on 2aPol
accumulation (). Class I 1a mutants that lack the ability to induce ER invaginations not only retained the ability to stimulate 2aPol
accumulation, but did so to nearly double the level of wt 1a (). In contrast, Class II 1a mutants that form more numerous, smaller spherules, lost the ability to stimulate 2aPol
levels over those in cells expressing 2aPol
alone (). Thus, 1a-mediated stimulation of 2aPol
accumulation was inversely correlated with the capacity of 1a to induce ER membrane invaginations.
Class I and Class II mutations in 1a helix A have opposite effects on 2aPol accumulation.
The localization of wt 1a and selected representatives of the Class I and II 1a derivatives in cells co-expressing 2aPol
is shown in (see also for localization of the 1a derivatives without 2aPol
). For each 1a mutant class, similar results were obtained with all members, and representative results are shown in . In these studies we used a replication-competent GFP-2aPol
fusion protein to allow direct fluorescence microscopy detection rather than immunofluorescence, which is often compromised by low 2aPol
detection sensitivity 
As seen previously 
fluorescence in the absence of 1a was mostly faint and diffusely cytoplasmic with a few punctate dots. When co-expressed with wt 1a, GFP-2aPol
co-localized with 1a in typical partial to almost complete ring-like perinuclear ER structures (), consistent with prior observations 
. Although Class II 1a mutants failed to significantly stimulate 2aPol
accumulation, the GFP-2aPol
that accumulated in cells expressing Class II mutants co-localized with the mutant 1a in perinuclear rings similar to wt 1a (, right two columns). By contrast, in the presence of the reduced membrane affinity Class I 1a mutants, GFP-2aPol
accumulated in large cytoplasmic clusters also containing a significant fraction of the mutant 1a, while the remaining 1a was distributed diffusely over the cytoplasm (, third and fourth columns), as when these Class I mutants were expressed without 2aPol
As another assessment of membrane association, flotation efficiency of 1a or its helix A mutants remained unaffected when co-expressed with 2aPol (compare with ). In the presence of wt 1a, 2aPol accumulation was stimulated and essentially all 2aPol became membrane-associated (). Likewise, 2aPol was recruited to membranes by the Class II 1a mutants with ~98% efficiency, but without any increased accumulation (). When co-expressed with any Class I 1a mutants, the efficiency of 2aPol flotation with membranes was only 50–60% (), slightly higher than without 1a and similar to the reduced membrane-association of Class I mutant 1a proteins themselves, with or without 2aPol ( and ).
RNA3 requires 1a-induced membrane invaginations to acquire a membrane-associated, nuclease-resistant state
In yeast cells, the half-life of RNA3 increases from 5–10 min in the absence of 1a to more than 3 hours in the presence of 1a, which is reflected in a marked increase in RNA3 accumulation 
. Accordingly, as shown in , lanes 1 and 2, GAL1
promoter-driven wt 1a increased RNA3 accumulation ~20-fold. Strikingly, the effects of the Class I and Class II mutations on 1a stimulation of RNA3 accumulation were opposite to each other and opposite to the effects of each mutant on 2aPol
. Co-expressing class II 1a mutants stimulated RNA3 accumulation ~40-fold, or double the stimulation by wt 1a (), in parallel with the increased frequency of spherule formation by these mutants (). In contrast, Class I 1a mutants showed no ability to stimulate RNA3 accumulation, so that RNA3 levels in cells expressing Class I 1a mutants were similar to those in cells lacking 1a ().
Class I and Class II mutations in 1a helix A have opposite effects on recruiting genomic RNA3 to a membrane-associated, nuclease-resistant state.
Wild type 1a recruits RNA3 into a membrane-associated, nuclease-resistant state 
. To define the state of RNA3 in the presence of the Class I and Class II 1a mutants, we assayed RNA3's membrane flotation efficiency, sedimentation, and nuclease sensitivity when co-expressed with these mutants (). Without wt 1a, RNA3 remained at the bottom of flotation gradients, indicative of a complete lack of membrane-association. In sedimentation assays, RNA3 from cells lacking 1a was mainly detected in the membrane-depleted supernatant and readily degraded with micrococcal nuclease (). In the presence of wt 1a or its Class II mutants, at least 80% of RNA3 segregated with the membrane fraction in the top gradient fractions or the membrane-enriched pellet fraction in sedimentation assays, and became highly nuclease-resistant, while Class I mutants failed to induce RNA 3 membrane association or nuclease resistance (). Thus, the loss or enhancement by Class I or II 1a helix A mutants of wt 1a's ability to stimulate RNA3 accumulation in vivo was closely linked with RNA3's acquisition of a membrane-associated, nuclease-resistant state, and with the capacity of each 1a mutant's ability to induce membrane invaginations.
Helix A mutations abolish BMV RNA replication in yeast and a natural plant host
In cells expressing 1a and 2aPol
, RNA3 transcripts are recruited into 1a- and 2aPol
-containing replication complexes to serve as templates for synthesis of negative-strand RNA3, which in turn becomes the template for synthesis of progeny positive-strand RNA3 and subgenomic positive-strand RNA4 
. Since the Class I and Class II mutants in 1a helix A show opposite effects on 1a-associated intracellular localization, ER membrane rearrangements and stimulation of 2aPol
and RNA3 accumulation () we compared how these mutants affect BMV RNA replication in yeast and in a natural plant host of BMV, barley.
Yeast cells expressing wt 1a, 2aPol and RNA3 supported efficient viral RNA replication (). In contrast, in cells expressing Class I 1a mutants, 2aPol and RNA3, only weak RNA3 signals were detected, similar to the levels of RNA3 derived entirely from plasmid-based transcription in cells lacking 1a (). In cells expressing Class II 1a mutants, 2aPol and RNA3, positive-strand RNA3 accumulated to levels intermediate between those in cells with and without wt 1a (), consistent with the ability of class II 1a mutants to mediate RNA3 recruitment to a membrane-protected state (). However, positive-strand RNA4 and negative-strand strand RNA3, which are only synthesized as products of viral RNA replication, were undetectable in cells expressing any of the Class I 1a mutants, and reached only 5–10% of wt levels in cells expressing most Class II 1a mutants (). The only exception was Class II mutant 1aT397A, which weakly stimulated 2aPol accumulation () and retained ~25% of wt 1a replication levels (). Thus, BMV RNA replication was severely inhibited by the helix A mutations in both classes.
1a helix A mutations abolish BMV RNA replication in yeast cells and barley.
To compare the replication competence of the 1a helix A mutants in yeast to that in BMV's natural plant host, 7-day old leaves of barley plants were inoculated with in vitro transcribed wt or mutant RNA1 transcripts and equal amounts of RNA2 and RNA3 transcripts. Seven to nine days post inoculation with wt BMV RNAs, even leaves that were not inoculated but rather depended on systemic viral spread for infection contained abundant levels of RNA1, 2, 3 and RNA4 (, lane 1). However, none of the RNA 1 mutants, including 1aT397A, supported detectable systemic infection (). Thus, 1a mutations in helix A that abolish or severely inhibit BMV RNA replication in yeast also render the virus severely replication-deficient in its natural host.