HIV2 genomic RNA contains a novel type of IRES located downstream of its initiation codon

1
HIV-2 genomic RNA contains a novel type of
IRES located downstream of its initiation codon
Nature Structural and molecular biology
Presented by Junzhi Shi
2
Abstract

• Eukaryotic translation initiation begins with
  assembly of a 48S ribosomal complex at the 5 cap
  structure or at an internal ribosomal entry
  segment (IRES). In both cases, ribosomal
  positioning at the AUG codon requires a 5
  untranslated region upstream from the initiation
  site.
• Here, we report that translation of the genomic
  RNA of human immunodeficiency virus type 2
  takes place by attachment of the 48S
  ribosomal preinitiation complex to the coding
  region, with no need for an upstream 5
  untranslated RNA sequence.

3

• This unusual mechanism is mediated by an RNA
  sequence that has features of an IRES with the
  unique ability to recruit ribosomes upstream from
  its core domain.
• A combination of translation assays and
  structural studies reveal that sequences located
  50 nucleotides downstream of the AUG codon are
  crucial for IRES activity.

4
Introduction

• HIV-2 genome
• IRESs in HIV-1
• IRESs in HIV-2

5
HIV-2 genome

• Human immunodeficiency virus type 2 (HIV-2) is a
  member of the lentivirus group of retroviruses
  and is one of the two etiological agents of AIDS
  in humans1. HIV-2 and HIV-1 share a common
  genomic organization, and both are closely
  related to simian immunodeficiency viruses
  (SIVs).

6

• The full-length genomic RNA of HIV-2 serves both
  as genome for the production of new virions
  and as messenger RNA for the synthesis of Gag
  and Gag-Pol protein precursors during the late
  step of viral replication. The HIV-2 genomic RNA
  has a 548-nucleotide (nt) 5 untranslated region
  (UTR) that harbors several RNA motifs such as the
  TAR stem-loop, the primer binding site (PBS) and
  a series of hairpin motifs necessary for genome
  dimerization and packaging (Psi motifs).

7

• Protein synthesis is key to the regulation of
  eukaryotic gene expression and is mediated by a
  number of proteins called initiation factors,
  which promote efficient binding of the ribosome
  to the 5 end of the mRNA, scanning of the 5 UTR
  and selection of the correct initiation codon.

8
IRESs in HIV-1

• A number of viral and cellular mRNAs initiate
  translation by direct ribosome binding to the 5
  UTR. during this process, the 40S small ribosomal
  subunit binds an IRES located upstream from the
  AUG initiation site. In many instances, this
  mechanism has been shown to be mediated by the
  three-dimensional structure of the IRES.
• Two distinct IRESs have been identified in the
  HIV-1 genomic RNA. The first lies within the 5
  UTR and is active during the G2/Mphase of the
  cell cycle and the second is located entirely
  within the gag coding region and drives
  translation of a novel 40-kDa N-terminally
  truncated Gag isoform by initiation at an
  internal AUG codon.

9
IRESs in HIV-2

• The present study reveals that the genomic RNA of
  HIV-2 drives synthesis of p57 Gag and of two
  additional Gag isoforms having apparent molecular
  weights of 50 and 44 kDa, all three of which are
  produced by a cap-independent mechanism.
• To our surprise, delineation of the RNA sequence
  required for internal ribosome binding revealed
  that the IRES was located entirely downstream of
  the first AUG initiation site, with no
  involvement of the upstream 5 untranslated
  region.

10

• We further supported this by constructing a
  leaderless HIV-2 genomic RNA starting directly at
  the gag AUG codon. Not only was this RNA
  faithfully translated in rabbit reticulocyte
  lysate (RRL), but protein production was
  sevenfold more efficient than that from the
  wild-type HIV-2 RNA.
• Site-directed mutagenesis revealed that assembly
  of the 48S preinitiation complex occurred at the
  authentic AUG initiation site despite the lack of
  an upstream UTR.
• Finally, a structural model of this unusual IRES
  was constructed from chemical and enzymatic
  probing data and revealed a highly structured
  region downstream of the AUG codon featuring a
  central long-distance interaction.

11

• Truncation of the first 50 nt of the IRES element
  almost abolished translation at the proximal
  site.
• Together, our results describe a novel type of
  IRES sequence that has the ability to recruit
  a ribosomal subunit upstream from its core
  sequence.

12
results

• Immature HIV-2 particles contain three isoforms
  of Gag
• Cap-independent expression of p57, p50 and
  p44
• Cap-independent translation in HeLa cells
• Translation of gag does not require the 5 UTR
• Toward a structural model for the HIV-2 IRES

13
Immature HIV-2 particles contain three
isoforms of Gag

• HIV-2 particles were generated by HIV-2 pRod10
  DNA transfection in cells treated with protease
  inhibitors. Immature virions were collected 48 h
  after DNA transfection and analyzed by SDS-PAGE
  followed by immunodetection using antibodies to
  capsid p27 (CAp17) and matrix p17 (MAp17) (Fig.
  1a).

14
These experiments were carried out in the
presence of highly specific HIV protease
inhibitors to inhibit proteolytic processing of
Gag immature precursor by the HIV virally encoded
protease. Under these conditions, the canonical
Pr57Gag protein (hereafter referred to as Gag
p57) together with two additional proteins having
apparent molecular weights of 50 and 44 kDa
(referred to as p50 andp44) were detected in
immature virions (Fig. 1b, lane 2).
Notably, incubation of the membrane with an
antibody directed against the extreme N-terminal
region of the matrix (MAp17) did not detect p50
and p44 in the immature virus particles (lane 4),
suggesting that they are N-terminally truncated
isoforms of Gag.
15
The presence of three Gag isoforms was confirmed
by in vitro translation of a construct containing
the gag gene (Fig. 1c).
16

• Analysis of the HIV-2 sequence revealed two in
  frame internal AUG codons at nt 746 and 899 of
  the gag coding region (named AUG2 and AUG3, AUG1
  being the authentic initiation site).
• Therefore, sitedirected mutagenesis was
  performed to change AUG2, AUG3 or both to CUC
  (Fig. 1d). This resulted in the loss of p50, p44
  or both, respectively,

17

• showing that these AUGs are bona fide initiation
  codons.
• To ensure that these AUGs are used in cells,
  plasmids carrying the wildtype gag (pGagHIV-2)
  or double-mutant gag with AUG2 and AUG3 changed
  to CUC (pGagHIV-2 CUC2/3) were transfected into
  293T cells. The double mutant did not express p50
  or p44, confirming results obtained in vitro
  (Fig. 1e).
• It is noteworthy that expression from AUG2 and
  AUG3 was weaker from the Gag-encoding plasmid
  than from the entire proviral clone (compare p50
  and p44 in Fig. 1b,e).

18
Cap-independent expression of p57, p50 and p44

• As AUG1 is surrounded by sequences favorable for
  translation initiation,it seemed unlikely that
  initiation at AUG2 and AUG3 could be the result
  of leaky ribsomal scanning. However, we
  investigated this possibility directly using an
  antisense 2-O-methyloligoribonucleotide
  complementary to the AUG1 codon and its
  downstream sequence.
• 2-O-methyloligoribonucleotides were annealed
  to the HIV-2 RNA (see Methods) and the resulting
  oligonucleotide-mRNA duplex was translated Under
  these conditions, production of p57 was severely
  compromised (8090 inhibition), showing that
  virtually no initiation took place at AUG1.

19
In contrast, translation at the downstream
initiation sites (AUG2 and AUG3) was not
substantially affected, showing that such
translation does not involve leaky ribosomal
scanning.
20

• We further addressed this by translating a series
  of mono- (Mo-) and bicistronic (Bi-) RNAs(Fig.
  2b). These showed inefficient expression of a
  reporter gene (LacZ) when the 5 UTR alone was
  driving translation (constructs
• Mo- and Bi-AUG1 lanes 1 and 4) the level of
  expression was barely above the negative control
  (lane 3). In contrast, when the 5 UTR was
  followed by the coding region spanning from AUG1
  to AUG3 (constructs Mo- and Bi-AUG3), three
  isoforms of LacZ were detected that were
  expressed very efficiently (lanes 2 and 5).
  However, insertion of a longer fragment of the
  gag coding region up to the fifth AUG (position
  1089) did not change either the relative ratios
  or the amounts of proteins made (data not shown).

21
(No Transcript)
22

• To map the IRES boundaries, deletions were
  generated in Bi-AUG3 and the resulting RNAs were
  translated in RRL (Fig. 2c). To oursurprise,
  deletion of the entire 5 UTR (construct
  Bi-AUG13) had virtually no impact on
  b-galactosidase (b-gal) expression (Fig. 2c, lane
  4), suggesting that the coding region alone is
  sufficient for internal initiation. Moreover,
  the largest isoform of b-gal was still expressed
  from Bi-AUG13 RNA.

23

• To ensure that LacZ was expressed independently
  of the other reporter gene (neo) in the
  bicistronic contructs, a stable hairpin loop was
  inserted at the 5 end. This essentially
  abolished synthesis of the neomycin-resistance
  gene product neo without affecting b-gal
  production (Fig. 2d, lane 2).

24

• These results show that expression of Gag is
  driven by an IRES element located entirely in the
  coding region between AUG1 and AUG3.

25
Cap-independent translation in HeLa cells

• To substantiate the in vitro data, HeLa cells
  were transfected with one of the following
  plasmids pBi-AUG1 (encoding the Bi-AUG1
  construct), pBi-AUG3, pBi-AUG13 or pBi-PV (a
  control construct that has the poliovirus genome
  sequence inserted between the two reporter genes)
  and then selected for neomycin resistance.

26

• RT-PCR analysis was carried out to test for
  aberrant splicing products, but did not detect
  any shorter isoforms of the bicistronic RNAs,
  indicating that no major post-transcriptional
  processing had taken place (Fig. 3a).

27

• As an additional test, CMV promotercontaining
  and promoterless (DCMV) plasmids were transiently
  transfected into HeLa cells and b-gal activity
  was measured 48 h after transfection (Fig. 3b).
  The lack of b-gal expression from the DCMV
  constructs indicates that no cryptic promoter
  sites were used to generate shorter isoforms of
  the bicistronic constructs used in this study.

28

• Expression of b-gal from pBi-AUG1 was weak (Fig.
  3c), confirming that the 5 UTR is very
  inefficient at driving translation, as
• noted above (Fig. 2b). Insertion of the
  AUG1AUG3 segment of

29

• the coding region (pBi-AUG3) had only a small
  effect on translational efficiency. However, the
  additional removal of the 5 UTR (pBi-AUG13)
  resulted in strong enhancement of LacZ expression
  (pBi-AUG13). This suggests that the AUG1AUG3
  RNA segment has strong IRES activity in HeLa
  cells and that this activity can somehow be
  modulated by the presence of the upstream HIV-2
  5 UTR.

30
Translation of gag does not require the 5
UTR

• We then reasoned that if the AUG1AUG3 RNA
  segment were both necessary and sufficient for
  IRES activity in a bicistronic context, this
  sequence could also drive translation of a
  monocistronic construct without a 5 UTR.
  Therefore, leaderless monocistronic constructs
  starting at the first or third AUG codon (T7-AUG1
  and T7-AUG3) were designed. Because of technical
  constraints resulting from the requirements of
  the T7 RNA polymerase, all leaderless RNAs start
  with a single guanine preceding the AUG codon
  (Fig. 4a, top).

31
(No Transcript)
32

• Translation of T7-AUG1 RNA gave rise to p57, p50
  and p44 (Fig. 4a, lane 1), whereas protein
  production from T7-AUG3 RNA was barely detectable
  (lane 3), confirming that the AUG1AUG3 region
  is crucial for initiation at the proximal site.
  Mutating AUG1 to a CUC codon (construct T7-CUC1
  lane 2) resulted in the loss of Gag p57
  production with no change in the synthesis of Gag
  p50 and p44, confirming that AUG1 was indeed the
  initiation site on this leaderless RNA.

33

• The next step was to compare the translational
  efficiency of the wild-type HIV-2 gag RNA with
  that of its leaderless counterpart.

34

• To our surprise, expression of the HIV-2 gag RNA
  was much more efficient in the absence of the 5
  UTR at both low and high concentrations of RNA
  (compare lanes 15 with 610).
• This confirms that the 5 UTR is dispensable for
  gag expression but modulates the IRES activity of
  the coding region. Notably, the unchanged
  relative proportions of the three Gag isoforms
  upon deletion of the 5 UTR imply that the leader
  does not affect start codon selection.

35
Toward a structural model for the HIV-2 IRES

• We modeled the secondary structure of the first
  420 nt of the gag coding region, which
  constitutes the core of the IRES.
• Three synthetic RNAs containing the first 420 nt
  of the coding sequence and differing by their 5
  termini were synthesized in vitro. The first
  includes
• the whole 5 UTR and the coding region between
  AUG1 and AUG3 (nt 1968), the second includes
  the first 420 nt of the gag coding region between
  AUG1 and AUG3 (548968) and the third,
  intermediate construct starts 44 nt upstream of
  the first AUG (504968).

36

• When analyzed on a nondenaturing polyacrylamide
  gel in the presence of Mg2 (ref. 17), each of
  the three transcripts primarily runs as a single
  band (data not shown), indicating a lack of
  aggregation, dimerization or stable alternative
  structures.
• Similar ionic conditions and RNA concentrations
  were used for the structure-probing experiments.
  We measured the accessibility of the transcripts
• to the single strandspecific probes dimethyl
  sulfate (DMS) and 1-cyclohexyl-3-(2-morpholinoethy
  l) carbodiimide metho-p-toluene sulphate (CMCT)
  in the presence and absence of Mg2 in an attempt
  to obtain insights into long-range interactions.

37

• In a first round of modeling, only the positions
  modified by DMS and CMCT under all conditions
  were constrained to a single-stranded
  conformation for the RNA-folding prediction
  software Mfold19. The model was then gradually
  refined to account for more of the probing data.

38
(No Transcript)
39

• Most of the nucleotides modified by DMS or CMCT
  are located in single-stranded regions or at the
  edges of helices, positions that are susceptible
  to breathing (for example, A733 or C888), and
  can be protected upon stabilization by Mg2 (for
  example, U608, G842 and G843).
• Two stretches of nucleotides, A592A601 and
  U778U787, show an ambivalent pattern, being
  weakly modified by both single and double
  strandspecific probes and partially protected
  from DMS or CMCT upon Mg2 addition (Fig. 5 and
  Supplementary Fig. 4 online). These two sequences
  can base pair over 10 nt we therefore model them
  as a helical segment extending helix P3.

40

• This interaction is marginally stable and is
  mostly not formed in the absence of magnesium
  ions. This could indicate that P3 is a
  determinant of the
• three-dimensional structure that forms
  cooperatively upon magnesium addition.
  Alternatively, the relative weakness of those
  pairings and their stabilization in the presence
  of magnesium could reflect their high A-U and GU
  contents.

41

• Few positions modified by V1 nuclease were left
  unpaired this could reflect stacking of these
  bases or their involvement in putative
  long-distance interactions. The presence of the
  5 UTR did not alter the modification pattern
  observed for the gag coding region.
• However, the 5 UTR influences the IRES activity
  of the AUG1AUG3 sequence, and two long-distance
  interactions between the 5 UTR and the coding
  region proposed for HIV-1 could be extrapolated
  to HIV-2.
• The first, between the U5 region (nt 186199) and
  sequences around the first AUG (556543), could
  occlude the initiation codon21,22. Among these
  nucleotides, A193 is modified by DMS and nt
  553556 have been modeled in a local interaction
  on the basis of their susceptibility to

42

• V1 nuclease in the absence of the 5 UTR.
  Therefore our data do not support this
  long-distance interaction. Among nucleotides
  involved in the second proposed long-distance
  interaction23, the sequence within the coding
  region (655661) is not reactive to any probe
  used. However, five positions within its
  potential 5 UTRcounterpart (158163) are cleaved
  by RNase V1. Our data could advocate for this
  long-distance interaction, the formation of which
  would not markedly disrupt the structural model
  we propose. Notably, analogous interactions
  within HIV-1 RNA were not detected in vivo24.
• Finally, analysis of 20 HIV-2 and 17 SIV isolates
  revealed a similar secondary structure model,
  suggesting that the structure is conserved(data
  not shown).

43

• Examining the IRES structural model, we
  hypothesized that the presence of the central
  long-distance pairing P3 could be crucial for
  leaderless translation. To test this hypothesis,
  we constructed two leaderless RNAs in which the
  AUG start codon was located a few nucleotides
  before P3 (at position 573 Fig. 5, arrow) or
  within P3 (at position 597 Fig. 5, arrow). The
  T7-AUG597 leaderless RNA was crippled to initiate
  translation at the proximal
• site (Supplementary Fig. 5 online), confirming
  that sequences downstream of the authentic AUG
  gag codon are crucial for translation initiation.

44
Discussion

• Here, we report that HIV-2 viral particles
• contain Gag p57 and two additional shorter
  isoforms of 50 and 44 kDa that are produced from
  translation initiation at two internal AUG codons
  (named AUG2 and AUG3) located within the coding
  region (Fig. 1).
• Using an RRL system, we show that these three
  Gag isoforms are produced by an IRES element
  that lies downstream from the authentic AUG1
  initiation site and spans to AUG3 (Fig. 2).
  Moreover, this RNA sequence is involved in the
  delivery of a preinitiation complex upstream
  from its core sequence to produce the Gag p57
  polyprotein from the authentic AUG codon (Fig.
  2).

45

• Translation of bicistronic constructs in HeLa
  cells confirmed that strong IRES activity mapped
  exclusively to the gag coding region spanning
  from
• AUG1 to AUG3 (Fig. 3). We reasoned that such a
  property should allow Gag production from
  monocistronic leaderless HIV-2 RNAs that start
  with the AUG1 codon. To our surprise, not only
  was such a construct, lacking the entire HIV-2 5?
  UTR, faithfully translated, but Gag production
  was much
• higher than from the 5 UTRcontaining
• HIV-2 genomic RNA (Fig. 4).
• As a first step toward complete structural
  characterization, chemical probing of the 420-nt
  AUG1AUG3 coding region was carried out on
  transcripts
• with or without the 5 UTR (Fig. 5).

46

• We found that the 5 UTR and the coding region
• fold as two independent domains. In the
  modeled structure, the three AUG codons are
  within a single-stranded region, which is
  probably favorable for formation of the
  preinitiation complex. Notably, these codons are
  always located in close proximity to a polypurine
  region that is also single stranded.
• Examination of the structure prompted us to focus
  on the role of the region located downstream of
  AUG1, helix P3, which has been modeled as a
  long-distance interaction. A leaderless RNA
  starting with an AUG codon at position 597 was
  generated and was found to be very inefficient at
  initiating translation at the proximal AUG site
  (Supplementary Fig. 5).This confirmed that RNA
  motifs downstream of the AUG initiation site are
  crucial for ribosomal recruitment upstream of the
  core IRES sequence. Notably, these data also
  showed that the level of p50 and p44 expression
  remained unaffected under conditions where p57
  synthesis was severely impaired (Fig. 2a and
  Supplementary Fig. 5), indicating that expression
  at AUG1 occurs independently from that at AUG2
  and AUG3.

47

• Although the results presented herein show that
  HIV-2 translation can occur by the binding of the
  ribosomal initiation complex to the gag coding
  region in the absence of a 5UTR, data obtained
  in cultured cells (Fig. 3) and in RRL (Figs. 2
  and 4) suggest that the 5UTR can modulate this
  process .In agreement with this, it should be
  noted that capping of wild-type HIV-2 gag
  transcripts increases the relative production of
  p57 compared to p50 and p44 (Fig. 1).
  Furthermore, addition of leader protease results
  in a change of the ratio of p57 to p50 and p44
  (Supplementary Fig. 1). Such an effect can be
  attributed to the strong enhancement of
  cap-independent translation after eIF4GI
  cleavage2529.

48

• This suggests that translation initiation on the
  HIV-2 genomic RNA is likely to occur both by a 5
• UTRdependent mechanism and by internal
  ribosome entry from the coding region. Therefore,
  the interplay between these two mechanisms is
  likely to be a key regulator of the HIV-2 viral
  cycle.
• It is noteworthy that the lentiviral 5 UTR is
  the preferential Gag binding site, which is
  absolutely required for viral assembly and
  encapsidation of the genomic RNA. Such an event
  takes place by initial attachment of the newly
  synthesized Gag polyproteins to RNA stem-loops
  involved in RNA packaging, located within the 5
  UTR. This occurs as the genomic RNA is being
  translated, creating a scaffold of RNAGag
  complex that occludes access to the 5 UTR for
  ribosomes .

49

• Thus, it is tempting to speculate that occlusion
  of the 5UTR by Gag molecules may promote the use
  of the IRES in the coding region. This way, viral
  protein production could continue during the
  early steps of viral assembly at times when the
  5 UTR is exclusively used for viral packaging.
• In agreement with this hypothesis, it should be
  noted is tempting to speculate that occlusion of
  the 5UTR by Gag molecules may promote the use of
  the IRES in the coding region. This way, viral
  protein production could continue during the
  early steps of viral assembly at times when the
  5 UTR is exclusively used for viral packaging.
  In agreement with this hypothesis, it should be
  noted that mutation of the two internal AUG
  codons responsible for synthesis of p50 and p44
  almost abrogates viral replication, suggesting
  that they have a key function in viral
  propagation .

50

• Such a proposed mechanism has important
  implications for our understanding of the viral
  life cycle, in that it suggests that translation
  and packaging are not mutually exclusive events
  as has always been assumed.

51

• That is all !