The polyomavirus BK agnoprotein co-localizes with lipid droplets
Gunhild Unterstaba, Rainer Goserta, David Leuenbergera,1, Pascal Lorentzb,
Christine H. Rinaldoc, Hans H. Hirscha,d,⁎
aTransplantation Virology, Institute for Medical Microbiology, Department of Biomedicine, University of Basel, CH-4003 Basel, Switzerland
bBio-Optics Facility, Department of Biomedicine, University of Basel, Basel, Switzerland
cDepartment of Microbiology and Infection Control, University Hospital of North Norway, Tromsø, Norway
dInfectious Diseases and Hospital Epidemiology, University Hospital Basel, Basel, Switzerland
a b s t r a c t a r t i c l ei n f o
Received 20 August 2009
Returned to author for revision
17 December 2009
Accepted 7 January 2010
Agnoprotein encoded by human polyomavirus BK (BKV) is a late cytoplasmic protein of 66 amino acids (aa)
of unknown function. Immunofluorescence microscopy revealed a fine granular and a vesicular distribution
in donut-like structures. Using BKV(Dunlop)-infected or agnoprotein-transfected cells, we investigated
agnoprotein co-localization with subcellular structures. We found that agnoprotein co-localizes with lipid
droplets (LD) in primary human renal tubular epithelial cells as well as in other cells supporting BKV
replication in vitro (UTA, Vero cells). Using agnoprotein-enhanced green fluorescent protein (EGFP) fusion
constructs, we demonstrate that agnoprotein aa 20–42 are required for targeting LD, whereas aa 1–20 or aa
42–66 were not. Agnoprotein aa 22–40 are predicted to form an amphipathic helix, and mutations A25D and
F39E, disrupting its hydrophobic domain, prevented LD targeting. However, changing the phosphorylation
site serine-11 to alanine or aspartic acid did not alter LD co-localization. Our findings provide new clues to
unravel agnoprotein function.
© 2010 Elsevier Inc. All rights reserved.
Polyomaviruses (PyV) are non-enveloped double-stranded DNA
viruses that have been isolated from a number of vertebrates
including birds, rodents, cattle, monkeys, and humans. With few
exceptions, PyV infections are little symptomatic in their natural
hosts. Six PyVs have been detected in human specimens: PyV BK
(BKV) and JC (JCV) in kidney, urine and brain (Chesters, Heritage, and
McCance, 1983; Dorries and ter Meulen, 1983; Gardner et al., 1971;
Padgett et al., 1971), KI virus and WU virus in respiratory secretions,
Merkel cell carcinoma virus in a rare skin cancer, and simian virus
SV40, though less consistently, in various tissues (Jiang et al., 2009).
Epidemiological data in humans indicate seroprevalence rates for BKV
and JCV of 82% and 58% in the general adult population (Egli et al.,
2009; Knowles et al., 2003). In immunocompetent individuals, both
viruses persist for life and urinary shedding is observed in 7% and 19%,
respectively (Egli et al., 2009). In immunosuppressed patients, three
major PyV diseases can arise: (1) PyV-associated nephropathy, caused
mostly by BKV, in 1–10% of kidney transplant patients (Ramos et al.,
2009). (2) PyV-associated hemorrhagic cystitis, attributed mostly to
BKV, in 5–15% of allogenic hematopoietic stem cell transplant patients
(Dropulic and Jones, 2008). (3) PyV-associated multifocal leukoence-
phalopathy caused mostly by JCV in patients with profound and
persistent immunodeficiency due to hematological malignancy,
advanced HIV-AIDS, or exposure to potent anti-lymphocyte agents
(Jiang et al., 2009; Khanna et al., 2009). Despite the known clinical
impact of BKV and JCV replication in immunosuppressed patients,
effective antiviral therapy is lacking (De Clercq, 2004; Rinaldo and
Hirsch, 2007). To identify potential antiviral targets of PyV replication,
we are characterizing the BKV-host interaction.
Agnoprotein is one of the six major BKV proteins and expressed as
a late gene, after the early genes large T- and small T-antigen at
around 36 h post infection, together with the three capsid proteins
VP1, VP2 and VP3 (Rinaldo et al., 1998). Although the agnoprotein is
abundantly expressed in vivo in kidney transplant patients with
nephropathy, the immune system fails to mount significant cellular
and humoral responses (Leuenberger et al., 2007). BKV agnoprotein is
only 66 amino acid (aa) long with an approximate 80% homology to
JCV or SV40 agnoprotein suggesting a conserved function (White and
Khalili, 2005). For JCV agnoprotein, diverse functions have been
Virology 399 (2010) 322–331
⁎ Corresponding author. Transplantation Virology, Institute for Medical Microbiol-
ogy, Department of Biomedicine, University of Basel, CH-4003 Basel, Switzerland. Fax:
+41 61 267 3283.
E-mail address: firstname.lastname@example.org (H.H. Hirsch).
1Present address: Liverpool School of Tropical Medicine, Liverpool, UK.
0042-6822/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/yviro
reported including negatively regulating JCV DNA replication and
transcription (Safak et al., 2001), impairing the host cell response to
DNA damage (Darbinyan et al., 2004), and promoting virus release by
interacting with the microtubule associated protein FEZ1 (Suzuki et
al., 2005). Most likely, BKV agnoprotein shares some or all of these
functions, but the major function is not clear. Recent data indicated
that BKV agnoprotein is phosphorylated by protein kinase C at serine-
11 (Johannessen et al., 2008). Since BKV agnoprotein is highly basic
(pI ≈ 10.0), it is possible that it binds to DNA as reported for SV40
agnoprotein (Jay et al., 1981). We previously observed an inhibitory
effect of agnogene transfection on BKV non-coding control region-
driven gene expression (Gosert et al., 2008). Given the presumably
essential function in the BKV life cycle, we investigated its subcellular
localization by screening cellular markers for co-localization using
immunofluorescence microscopy and confocal laser scanning micros-
copy (CLSM). In this report we provide first evidence that agnoprotein
interacts with lipid droplets via a core domain from aa 20 to 42.
Subcellular distribution of agnoprotein
To analyze the subcellular distribution of agnoprotein, we infected
renal proximal tubular epithelial cells (RPTECs) and Vero cells with
BKV. We then performed immunofluorescence microscopy with a
polyclonal antiserum directed against the agnoprotein and a mono-
clonal antibody directed against the viral early protein large T-antigen
(LTag). BKV agnoprotein was detected in the cytoplasm while LTag
was seen in the nuclei of RPTECs (Fig. 1a) and Vero cells (Figs. 1c, d).
We observed that agnoprotein displayed a fine granular, almost
reticular cytoplasmic distribution as well as a localization to
prominent dense donut-like structures (Fig. 1b). To investigate
whether or not the cytoplasmic distribution of agnoprotein was
dependent on BKV infection and other viral proteins, we transiently
transfected Vero cells with an agnoprotein expression vector driven
by a CMV-promoter. A similar cytoplasmic distribution of agnoprotein
was observed as in infected cells, with a fine reticular pattern and a
prominent donut-like pattern (Figs. 1e, f). To broaden the range of
immunofluorescence co-staining, we generated His- and GST-tagged
agnoprotein (pCMV-agno-His, pCMV-GST-agno) and confirmed that
both fusion proteins showed a distribution similar to native
agnoprotein after transfection of Vero cells (data not shown). We
also generated stably transfected UTA cell lines in which agnoprotein
expression was under the control of a tetracycline-repressible tet-off
promoter where a similar staining pattern was seen (Figs. 1g, h). The
data indicate that the cytoplasmic distribution of agnoprotein to fine
granular and condensed donut-like structures occurs independently
of other viral proteins. Given their prominent size, we noted that the
donut-like structures could be readily detected by the transmission
channel of a confocal microscope (Figs. 1i–k).
Co-localization studies of agnoprotein with cellular marker proteins
The fine granular and the donut-like patterns of agnoprotein
appeared to differ in number and intensity between neighboring cells,
suggesting an association with a dynamic subcellular structure. To
identifythedonut-like structures, wetested antibodies directed against
marker proteins of various cell organelles and structures for co-
localization with agnoprotein by confocal microscopy and subsequent
deconvolution. Because some antibodies against cellular antigens were
from rabbit, and no anti-agnoprotein antibody from mouse was
available, we expressed tagged agnoprotein to allow co-staining for
agnoprotein with tag-specific mouse monoclonal antibodies.
To test for association of agnoprotein with secretory structures, we
first examined co-localization of tagged agnoprotein with calnexin
and p63, both ER resident proteins. As judged from the staining
pattern and, for p63 also from the deconvolution providing a
maximum intensity projection and a corresponding three-dimen-
sional view (isosurface), no co-localization of agnoprotein with either
calnexin or p63 was apparent (Supplementary Fig. 1a, b). Next, we
investigated co-localization of agnoprotein with the Golgi complex,
using antibodies against beta-COP and giantin, with COPII transport
vesicles, using Sar1 as a marker, and with early endosomes, by
detecting the EEA1 marker protein (Supplementary Fig. 1c–f). We
found no evidence for co-localization of tagged agnoprotein with any
marker proteins tested. We also addressed cytosolic structures for co-
localization with agnoprotein by choosing polyA-binding protein
(PABP) as a marker for mRNP complexes, and the pUB-R2 protein as a
Fig. 1. The subcellular distribution of BKV agnoprotein. Confocal images of agnoprotein
(green) and LTag (red) in infected RPTECs (a), with enlargement of agnoprotein specific
donut-like structures (b), and Vero cells (c, d) at 72 h post infection. Detection of
agnoprotein after transfection of agnoprotein gene in Vero cells (e, f) and the UTA-agno
cell line (g, h). Bar 20 μm. Enlargement of agnoprotein specific donut-like structures in
UTA-agno cells: (i) confocal image, (j) transmission channel and (k) merge of images
shown in i and j. Bar 5 μm.
G. Unterstab et al. / Virology 399 (2010) 322–331
marker for the 26S proteasome but failed to detect any co-localization
(Supplementary Fig. 2a, b). For JCV agnoprotein, an interaction with
the heterochromatin binding protein (HP1α) has been described,
which leads to an indirect interaction with the nuclear envelope
(Suzuki et al., 2005). However, we could not detect co-localization
with HP1α or with Lamin-A (Supplementary Fig. 2c, d).
Fig. 2. Co-localization studies for BKV agnoprotein and cytoskeletal proteins. (rows a–d) Detection of agnoprotein (green) and tubulin (red) in UTA-agno-cells. (a) Cells were left
(isosurface). (e) Detection of agnoprotein (green) in BKV-infected RPTECs at 72 h post infection together with actin microfilaments (red). (f) Detection of agnoprotein (green) in
pCMV-agno transfected Vero cells together with intermediate filament vimentin (red). Bar 20 μm.
G. Unterstab et al. / Virology 399 (2010) 322–331
To analyze co-localization of agnoprotein with components of the
cytoskeleton, we examined microtubules, the intermediate filament
vimentin, and the microfilament actin. Staining for tubulin and
agnoprotein in stably transfected UTA cells indicated that agnoprotein
was not co-localizing with microtubules (Fig. 2a). Treatment of UTA
cells with paclitaxel, stabilizing the microtubules, did not appear to
alter the distribution of agnoprotein (Fig. 2b). Treatment of cells with
vincristine depolymerized microtubules and lead to homogeneously
distributed tubulin dimers throughout the cytoplasm (Fig. 2c). In such
cells, the fine granular distributed agnoprotein seemed to partially co-
localize with the tubulin signals as evidenced by the yellow signals
(Fig. 2d, left panel), but deconvolution studies indicated that this was
not due to shared surfaces but due to an overlap of the green and red
staining located at different levels of the stack (Fig. 2d, middle and
right panel). Interestingly, the areas of agnoprotein-donuts appeared
retained, spared from the tubulin signals (see arrows Fig. 2c, middle
panel). Finally, we investigated the potential interaction of agnopro-
tein with microfilamentsand intermediate filaments, using antibodies
against actin and vimentin, respectively, but obtained no evidence for
co-localization (Figs. 2e, f). We further extended our studies to several
other inducible cellular target structures, namely P-Bodies, aggre-
somes,ribosomesandlysosomes.Noneofthese structures werefound
to co-localize with agnoprotein (data not shown). Table 1 summarizes
all marker proteins tested.
BKV agnoprotein co-localizes with lipid droplets
In order to test co-localization of agnoprotein with lipid droplets,
we stained BKV-infected or agnoprotein transfected cells with
LipidTOX™ and simultaneously detected agnoprotein. A significant
co-localization of lipid droplets and agnoprotein became apparent.
Although all cells are per se able to form lipid droplets, the number of
lipid droplets varied in cell culture from scarce to numerous (see also
Fig. 1). We therefore induced lipid droplet formation by exposing cells
to oleate (Brasaemle and Wolins, 2006) (Fig. 3). A substantially higher
number of agnoprotein-stained donut-like structures was seen
together with an increasing number of lipid droplets compared to
oleate-unexposed cells. Of note, agnoprotein-specific signals showed
the characteristic donut-like appearance, whereas lipid droplets
appeared like small filled vesicles. The overlay demonstrates that
agnoprotein envelops each lipid droplet in agnoprotein-expressing
infected or transfected cells, resulting in the typical donut-like
appearance. To independently confirm our results in the absence of
antibody staining, we transfected cells with an expression construct
encoding an agnoprotein fused in frame to the N-terminus of
enhanced green-fluorescent protein (EGFP). Following lipid droplet
induction with oleate and staining with LipidTOX™, we obtained
serial z-slices, resulting in a 3D image, a so called z-stack. Subsequent
deconvolution was done using the Huygens professional software.
This methodis usedto eliminate blurring andnoisein order to recover
the original object. Thereafter, restored images were visualized with
Imaris (see Materials and methods). Two different regions of a
cropped stack from a typical cell are shown (Figs. 4a–d and e–h). Fig.
4a depicts an optical slice of the deconvoluted volume. A yellow signal
is visible around the lipid droplets as the result of co-localization of
agnoprotein with the surface of lipid droplets. This signal became
more pronounced after defining a third channel displaying areas of
Target structures tested for co-localization with BKV agnoprotein using CLSM.
Target structureTarget protein Co-Localization
Fig. 3. BKV agnoprotein co-localizes with lipid droplets. To load cells with lipid droplets, cells were incubated with 300 μM oleate, bound to defatted BSA for 16 h. Left set: Confocal
image of Vero cells, transfected with pCMV-agno prior to oleate treatment. Cells were stained for agnoprotein (green) and lipid droplets using LipidTox™ (red). Because cells were
transfected and not infected, no LTag specific signal was obtained. Right set: Confocal image of BKV-infected RPTECs. Cells were treated with oleate at 24 h post infection. At 72 h post
infection, the early viral protein LTag (magenta) was detected together with agnoprotein (green) and lipid droplets (red). The lower right panel of each set represents a merged
image of each channel together with Hoechst stained nuclei (blue). Bar 20 μm.
G. Unterstab et al. / Virology 399 (2010) 322–331
co-localization in yellow, resulting in Pearson's coefficients of 0.9305
and0.8966,respectively(Figs. 4b andf). Figs. 4c ande providea three-
dimensional view (isosurface) of the corresponding areas shown in
Figs. 4a and f, demonstrating that agnoprotein almost entirely
envelops the lipid droplet surface. Fig. 4d demonstrates the donut-
like structure of agnoprotein distributed around lipid droplets by
depicting a section of the co-localization channel. Figs. 4g and h show
a cut along one plane of the isosurface model of Fig. 4e, allowing a
view inside the lipid droplet. It is evident that agnoprotein interacts
with the lipid droplet surface, resulting in its donut-like appearance.
The agnoprotein core of aa 20–42 mediates lipid droplet interaction
Secondary structure analysis of agnoprotein predicts aa 22–40 to
forman alpha helixwithamphipathic character.The hydrophobicface
is composed of seven amino acids, shown in yellow, and the charged
amino acids at the opposite side, shown in pink and blue (Fig. 5a). To
identify the domain involved in lipid droplet targeting we constructed
a set of N- and C-terminal truncated agnoprotein mutants fused to
EGFP (Fig. 5a). We transfected Vero cells with the different expression
constructs and verified the expression of the indicated fusion proteins
by Western blot (Fig. 5a). Co-localization with lipid droplets was
studied with or without lipid droplet induction (Fig. 5b). As shown,
agno(1-42)- and agno(20-66)-EGFP fusion proteins, both containing
the predicted helix co-localized with lipid droplets as observed for the
full-length agno-EGFP. In contrast, agno(1-20)- and agno(42-66)-
EGFP, both lacking the predicted helix failed to target lipid droplets.
We therefore concluded that aa 20–42 are necessary for directing
agnoprotein to lipid droplets. To further validate our hypothesis, we
disrupted the amphipathic character of the helix by specific point
mutations previously described for RNaseE in Escherichia coli
(Khemici et al., 2008). Accordingly, we generated a full-length
agnoprotein mutant in which alanine at position 25 was replaced
with aspartic acid (25D) and phenylalanine at position 39 with
glutamic acid (39E). As shown, agno(25D39E)-EGFP did not target
lipid droplets (Fig. 5c) but instead displayed a cellular distribution in
cytoplasm and nucleus indistinguishable from EGFP alone (Fig. 5b). To
independently investigate the distribution of mutant 25D39E agno-
protein, we used the agno specific antiserum for IF and confirmed co-
localization with the EGFP signal in the cytoplasm and in the nucleus
(Fig. 5c, lower row). From these data, we concluded that the
agnoprotein core of aa 20–42 is not only necessary to mediate co-
localization with lipid droplets, which required the hydrophobic
domain of the amphipathic helix, but also to mediate the preferential
retention in the cytoplasm. As phosphorylation of agnoprotein at
serine-11 by PKC has been suggested to modulate its effect on virus
replication (Johannessen et al., 2008), we constructed an agno-EGFP
fusion protein replacing serine at position 11 with aspartic acid (11D).
We found that the mutant agno(11D)-EGFP targeted lipid droplets
(Fig. 5d). Similarly, changing serine-11 into alanine, which prevents
phosphorylation at this site, also did not alter co-localization with
lipid droplets (data not shown).
In this study, we conducted a detailed analysis of the subcellular
distribution of the BKV agnoprotein using immunofluorescence
microscopy and CLSM. We observed that agnoprotein is distributed
in two patterns in the cytoplasm, one being fine granular, the other
being large and donut-like. Screening several cellular markers for co-
localization, we were able to exclude association with a number of
cellular structures (Table 1). However, we found that agnoprotein co-
localizes with lipid droplets presenting as donut-like structures. This
Fig. 4. BKV agnoprotein co-localizes with lipid droplets as verified by deconvolution
and visualization of 3D images. Vero cells were transfected with pAgno-EGFP and
loaded with lipid droplets after transfection. Cells were stained for lipid droplets and a
stack of 81 slices was acquired along the z-axis. Agno-EGFP signals are shown in green
and the lipid droplets are shown in red. Two different regions of a cropped stack are
visualized (a–d, bar 3 μm and e–h, bar 2 μm). (a) z-slice of the deconvolved volume.
(b) z-slice of the deconvolved volume with co-localizing pixels selected (Pearson
coefficient 0.9305) and shown in a new channel (yellow). (c) Isosurface rendering with
automatic threshold settings. (d) Visualization of a slice with manual defined thickness
from the isosurface rendering, omitting the red and green channel. (e) Isosurface
rendering with automatic threshold settings, omitting the co-localization channel.
(f) z-slice of a deconvolved stack with co-localizing pixels selected (Pearson
coefficient 0.8966) shown in a new channel (yellow). (g) Volume rendering of the
deconvolved stack that was cut to allow a view inside the lipid droplet showing three
channels or only the red and yellow channel (h).
Fig. 5. BKV agnoprotein secondary structure prediction and EGFP-fusion constructs. (a) Secondary structure prediction of the BKV agnoprotein was performed on the NPSA website
using the HNN tool (http://pbil.ibcp.fr/htm/index.php). The server predicts formation of a helix within aa 22–40. Illustration of the predicted amphipathic structure as a helical
wheel, drawn using the Helical Wheel Viewer on the website http://cti.itc.virginia.edu/∼cmg/Demo/wheel/wheelApp.html. aa 22–39 are shown (highlighted in yellow); nonpolar
aa (orange), polar, uncharged aa (green), acidic aa (pink), and basic aa (blue). Schematic presentation of agnoprotein-EGFP fusion constructs indicating the aa in parenthesis and
deleted parts as dotted line. The helix from aa 22–40 is shown as a blue barrel. Expression of the fusion proteins was verified by Western blot using an anti-GFP antibody. (b, c, d)
Confocal images of Vero cells transfected with EGFP and different pAgno-EGFP fusion constructs as indicated in (a) either without (−) or after 0.3 mM oleate exposure for 16 hours.
Agno-EGFP is shown in green, lipid droplets in red (LipidTOX™) and nuclei in blue (Hoechst 33342). Regions of co-localization appear yellow in the merge panel. (c, lower row)
Confocal images of Vero cells transfected with pAgno25D39E-EGFP expression construct to confirm proper expression of the agno25D39E-EGFP fusion protein. Agnoprotein was
detected using an agno-specific antiserum (magenta), EGFP specific signals are shown in green. Co-localization of both signals is shown in the merge panel. Bar 20 μm.
G. Unterstab et al. / Virology 399 (2010) 322–331
G. Unterstab et al. / Virology 399 (2010) 322–331
Fig. 5 (continued).
G. Unterstab et al. / Virology 399 (2010) 322–331
interaction was observed in different cell types including RPTECs,
Vero and UTA cells and became significantly more prominent when
lipid droplets were induced by exposure to oleate. We independently
verified this interaction by using agnoprotein-EGFP fusion proteins.
Moreover, by constructing N-terminal and C-terminal truncations of
agnoprotein, we provide evidence that the agnoprotein core from aa
20–42 is necessary to mediate lipid droplet targeting. The secondary
structure of this core region is predicted to form an amphipathic helix,
a structure known to mediate associations to lipid surfaces (van
Kuppeveld et al., 1997). Accordingly, two point mutations at positions
25 and 39, which disrupt the hydrophobic domain of the amphipathic
helix by introducing negatively charged aa, were sufficient to prevent
agnoprotein-EGFP association with lipid droplets. Conversely, mutant
agno-EGFP fusion proteins lacking the helix or having a disturbed
amphipathic helix were no longer exclusively retained in the
cytoplasm but were also present in the nucleus of the cells in both
untreated and oleate-treated cells. This observation suggests that the
amphipathic helix also is a major determinant of the cytoplasmic
retention of the agnoprotein. Changing phosphorylation site at serine-
11 to either aspartic acid to simulate phosphorylation, or to alanine to
prevent phosphorylation, did not affect lipid droplet targeting of
Lipid droplets are highly dynamic organelles that consist of a
hydrophobic core composed of neutral lipids, mostly triacylglycerol
and sterolesters, surrounded by a polar phospholipid monolayer.
Numerous proteins have been identified that attach to this membrane
and their number is growing. Recent work from different groups has
aimed at elucidating the lipid droplet proteome (Brasaemle et al.,
2004). Most interestingly, some proteins only temporarily localize to
lipid droplets and seem not primarily related to lipid metabolism (for
reviews, see Welte, 2007; Zehmer et al., 2009). Some proteins
detected in lipid droplets are also involved in vesicle transport, RNA
splicing, transcription as well as translation (Guo et al., 2008). It is
clear today, that virtually all eukaryotic cells are able to form lipid
droplets, in addition to adipocytes specialized in fat metabolism. Lipid
droplets appear to travel rapidly along microtubules and transiently
interact with various cellular organelles including endosomes (Liu et
al., 2007). Moreover, it has been proposed that lipid droplets are also
communicating with peroxisomes, mitochondria, lysosomes and the
ER via so called “transient inter-compartmental contact sites”
(Zehmer et al., 2009).
Our observation that the BKV agnoprotein targets lipid droplets
raises the question of the biological relevance of this interaction,
particularly in the late phase of the polyomavirus replication cycle. At
least three intracellular pathogens are taking advantage of lipid
droplets. The core proteins of HCV and the closely related GB virus-B
localize to lipid droplets (Barba et al., 1997; Hope et al., 2002) to
recruit nonstructural proteins as well as the viral RNA replication
complexes to lipid droplet-associated membranes (Miyanari et al.,
2007). On the other hand, the intracellular bacterium Chlamydia
trachomatis secretes a protein that targets lipid droplets to redirect
and exploit this source of lipids (Kumar et al., 2006). We noted that
association of BKV agnoprotein with lipid droplets occurs even in the
presence of little agnoprotein expression. Also, this interaction is seen
spontaneously, in the absence of deliberate lipid droplet induction
with oleate. These observations suggest that the presence of lipid
droplets in the cell is a strong dynamic determinant for the subcellular
localization of agnoprotein. In fact, we observed that JCV agnoprotein
also co-localizes with lipid droplets (unpublished data). Re-appreci-
ation of the work from Shishido-Hara et al. independently supports
that JCV agnoprotein distributes to fine granular as well as to donut-
like structures (Shishido-Hara et al., 2004).
Protein targeting to lipid droplets has been studied for several
proteins that interact with lipid droplets, including perilipin A, the
HCV core protein, the closely related GBV-B core protein (Hope and
McLauchlan, 2000; Hope et al., 2002; Subramanian et al., 2004) and
caveolin-1 (Ostermeyer et al., 2004). From these data, no consensus
sequence for targeting lipid droplets can be defined. Rather, different
modes of interaction seem to exist which include amphipathic helices
as predicted for the agnoprotein core aa 20–42. At present, the role of
agnoprotein in the context of lipid droplets remains speculative.
Cultured RPTECs, the natural target cells for BKV replication in BKV-
associated nephropathy show only a few lipid droplets when
maintained under tissue culture conditions. In the kidney, however,
the situation might be different (Nast and Cohen, 1985). For RPTECs
and other cells, it has been shown that fatty acids induce oxidative
stress and apoptosis and that one way to protect cells, is the
esterification to form triacylglycerol and cholesterol esters that are
stored in lipid droplets (Ishola et al., 2006; Urahama et al., 2008).
Thus, it is possible that lipid droplets might be disadvantageous for
viral replication, which is counteracted by agnoprotein. Also,
agnoprotein might facilitate lipid droplet degradation or promote
viral egress (Myhre et al., 2010). Recent work observed that
pravastatin, a cholesterol-lowering agent represses BKV replication
(Moriyama and Sorokin, 2008). Our observation that the BKV
agnoprotein co-localizes to lipid droplets is surprising and will
provide new insights not only into its role in the viral life cycle but
may prove to be another example of how small DNA viruses may be
useful probes for the study of as yet incompletely defined cellular
Materials and methods
Cell culture and reagents
Vero cells (ATCC CRL1587) were grown in DMEM (6046, Sigma, St.
Louis, MO), supplemented with 10% fetal bovine serum and 2 mM L-
glutamine (S0113 and K0302, respectively, Biochrome AG, Berlin,
Germany). Human renal proximal tubular epithelial cells (RPTECs)
(Sc-4100)werepurchasedfromScienCellLaboratories, SanDiego, and
maintained in epithelial cell medium (EpiCM, Sc-4101). For passaging
of cells, passage kit 2 (2040002) was used. The UTA-agno-11 cell line,
which expresses the BKV agnoprotein in a tetracycline-dependent
manner, was derived from the osteosarcoma cell line U2OS. Briefly,
U2OS cells were stably transfected with the regulatory plasmid
encoding for the transactivator and selected with G418 to yield clone
UTA 6 (Englert et al., 1995). These cells were then transfected withthe
agnoprotein-encoding construct pTRE-agno that confers puromycin
resistance. Puromycin-resistant clones were selected to yield the cell
line UTA-agno-11. This cell line was maintained in DMEM (6046,
Sigma), supplemented with 10% fetal bovine serum and 2 mM L-
glutamine (S0113 andK0302, respectively, BiochromeAG), 500 μg/ml
G418, 1.5 μg/ml puromycin, and 1 μg/ml tetracycline. To induce
agnoprotein expression, cells were cultivated in the absence of
tetracycline for 48 h. For manipulation of the microtubule network,
cells were treated with 1 μM vincristine (V8388) or paclitaxel (T7402)
(Sigma) for 4 h. To induce lipid droplets, culture medium was
supplemented with 300 μM oleate (O1383, Sigma) bound to
essentially fatty acid free bovine albumin (A6003, Sigma) as described
(Brasaemle and Wolins, 2006).
Expression plasmids and reagents
The pCMV-agno (pRC-agno) and pTRE-agno constructs have been
described (Leuenberger et al., 2007; Rinaldo et al., 1998). pCMV-GST-
agno, encoding for agnoprotein fused to the C-terminus of GST, was
generated by subcloning the GST-BKVagno fragment (BspEI–NotI)
from pFastBacGST-BKVagno (Leuenberger et al., 2007) into XmaI–NotI
cut pEGFP-N1 replacing the EGFP-N1 coding sequence. To construct
the pcDNA-agno-HIS expression plasmid, encoding for a C-terminal
HIS tagged agnoprotein, the agno-encoding sequence was amplified
using the forward primer 5'-GGTTGGGCTAGCATGGTTCTGCGCCAG-3′
G. Unterstab et al. / Virology 399 (2010) 322–331
and the reverse primer 5'-TTTTTTCTCGAGTCAATGATGATGATGAT-
GATGACCGGAGTCTTTTACAGAGTCT-3′ with pCMV-agno as template.
The NheI–XhoIinsert wasligated intothepcDNA3.1plasmid. To obtain
an expression plasmid for an agno-EGFP fusion protein, with EGFP
fused on the C-terminus of the agnoprotein, the agno-encoding
sequence was amplified using the forward primer 5'-GCGAATTCA-
CCATGGTTCTGCGC-3′ and reverse primer 5'-GCGGATCCCGCAAG-
GAGTCTTTTAC-3′, thereby introducing EcoRI and BamHI restriction
sites and eliminating the stop codon of the agno-coding sequence. For
C-terminal truncation mutants agno 1–20 and agno 1–42, the reverse
primers 5'-GCGGATCCCGTCCAGTCCAGGTTTTACC-3′ and 5'-GCGG-
ATCCCGACCTCTACAAAATTCCAGC-3′ were used, respectively, togeth-
er with the forward primer described for full-length agno-EGFP.
For the N-terminal truncation mutants agno20-66 and agno42-66,
we used the forward primers 5'-GCGAATTCACCATGGGAACAAAAA-
AAAGAGC-3' and 5'-GCGAATTCACCATGGGTGAAGACAGTGTAGACG-3'
together with the reverse primer described for full-length agno-EGFP.
The EcoRI–BamHI digested PCR product was ligated into accord-
ingly digested pEGFP-N1, resulting in pAgno-EGFP. The point
mutation agnoproteins A25D F39E, and S11D were obtained by
synthesizing the EcoRI–BamHI insert (Eurogentec, Liège, Belgium) and
ligating it into pEGFP-N1 as described above. Aa exchange of Ala→Asp
at position 25 was introduced by changing the codon GCT into GAT
and aa exchange of Phe→Glu at position 39 by changing the codon
TTT into GAA. Ser→Asp exchange at position 11 was introduced by
changing the codon TCT into GAT. All constructs were verified by
Transfection and infection
Transfection was done using Lipofectamine2000 (Invitrogen,
Carlsbad, CA) according to manufacturers instructions. RPTEC and
Vero cells were infected with BKV (Dunlop) containing Vero cell
supernatants as described previously for human umbilical vein
epithelial cells (Grinde et al., 2007).
Immunofluorescence (IF) and antibodies
Cells were fixed with 4% paraformaldehyde at room temperature
10 min. Fortheuse oftheanti-actinantibodycellswerepermeabilized
with ice-cold methanol at room temperature for 5 min. For the
detection of lipid droplets, permeabilization was done with 0.05%
0.05% Saponin. After incubation withprimary antibodies, diluted in 3%
BSA/PBS, at room temperature for 1 h, cells were washed three times
with PBS and incubated with secondary fluorescently labeled
antibodies, diluted in 3% BSA/PBS, at room temperature for 45 min.
After three washes, IF specimens were mounted in 90% Glycerol
gallate (P-3130, Sigma) as antifading agent.
Primary antibodies were used as followed: anti-agnoprotein
antiserum 1:800 (Rinaldo et al., 1998); anti-p63 (G1/296) 1:1000,
anti-Giantin (G1/133)1:400, anti-LAMP1(G1/139)1:1000 were kind
gifts from HP Hauri (Biozentrum, University of Basel) (Linstedt and
Hauri, 1993; Schweizer et al., 1993; Schweizer et al., 1988); anti-Sar1
1:500 kind gift from FT Wieland (Biochemistry Center, Heidelberg
University), anti-beta-COP 1:30 kind gift from BL Tang (University of
Singapore)(Rustet al.,2001). Commercially availableantibodieswere
used as followed: anti-actin1:200 (ab40864, Abcam, Cambridge, MA),
anti-calnexin 1:200 (SPA-860, StressGen, San Diego, CA), anti-EEA1
1:100 (610456, BD Biosciences, San Jose, CA), anti-His 1:300 (DIA900,
Dianova, Hamburg, Germany), anti-HP1α 1:50 (05-689, Upstate
Biotechnology, Billerica, MA), anti-LaminA 1:200 (ab8980, Abcam),
anti-PABP 1:50 (P6246, Sigma), anti-tubulin 1:200 (A-11126, Molec-
ular Probes, Carlsbad, CA), anti-vimentin 1:40 (V6389, Sigma); anti-
eIFη 1:200 (sc-16378), anti-GRP781:100 (sc-1051), anti-pUB-R2 1:25
(sc-13725), anti-p70s6k 1:500 (sc-8416) and anti-RPS6 1:25 (sc-
13007) wereall fromSanta Cruz Biotechnology (Santa Cruz, CA).Anti-
SV40Tag 1:50 (DP02, Calbiochem, San Diego, CA) cross-reacts with
BKV-LTag. Lipid droplets were visualized using LipidTOX™ (34476,
Invitrogen) at a 1:1000 dilution in the secondary antibody mixture.
Secondary antibodies were chosen depending on the primary
antibodies used: for triple staining, the following combinations were
used: LipidTOX™, anti-mouse-Alexa 647 1:200 (A-21463, Molecular
Probes), anti-rabbit-Alexa 488 1:1000 (A-21441, MolecularProbes) or
LipidTOX™, EGFP, anti-rabbit-Alexa 633 1:400 (A-21071, Molecular
Probes). For double labeling, anti-mouse-Alexa 488 1:800 (A-11029,
Molecular Probes) and anti-rabbit-Cy3 1:2000 (111-165-144, Jackson
Immunoresearch, West Grove, PA) or anti-goat-Cy3 1:600 (705-165-
003, Jackson Immunoresearch) and anti-rabbit-Alexa 488 were
combined. DNA was stained by Hoechst 33342 dye (0.5 μg/ml,
H21492, Invitrogen). In order to show agno-specific signals consis-
tently in green, channels were switched if necessary.
Confocal laser scanning microscopy (CLSM)
Digital optical sections were taken with a confocal laser scanning
microscope (Zeiss LSM 510 Meta, Carl Zeiss AG, Oberkochen,
Germany), using a 40 Plan-Neofluar/NA1.3 oil or a 63 Plan-
Apochromat/NA1.4 oil objective. Depending on the different fluoro-
chrome-tagged secondary antibodies following lasers were used:
Enterprise 405 nm; laser line 488 nm of the Argon laser, He–Ne
images were acquired sequentially using the multi-track mode.
For z-stacks acquisition of EGFP and LipidTox™ fluorescent
specimen, the requirements of the Nyquist theorem were fulfilled,
with voxel xyz size of 50, 50, 160 nm, respectively. The stack
contained 81 optical slices. Deconvolution and visualization was done
essentially as described (Rust et al., 2001). For deconvolution, the
Huygens professional software (Scientific Volume Imaging, Hilver-
sum, the Netherlands) was used, applying the Classic Maximum
Likelihood Estimation Mode (CMLE) using a theoretical point spread
function (PSF). The signal to noise ratio was set at 15 and 20 for green
(channel 1) and red (channel 2), respectively, and the background
was set to a value of 100. For 3D-rendering and data analysis, the
Imaris software (Bitplane AG, Zürich, Switzerland) was used. Co-
localizing pixels were visualized by defining a third channel (yellow),
resulting in Pearson's coefficients of 0.9305 and 0.8966 for pictures
6a–d and 6e–h, respectively.
Western blot was performed as described (Gosert et al., 2008).
Briefly, transfected cells were lysed in RIPA buffer (150 mM NaCl,
50 mM Tris–HCl, pH 8.0, 1% NP-40, 0.5% deoxycholate, 0.1% SDS, and
protease inhibitors [Roche]) at 24 h post transfection. Cell lysates
were separated by SDS-PAGE and electrotransferred onto 0.2 μm
nitrocellulose membrane (Whatman). The primary anti-GFP (ab1218,
Abcam) was diluted 1:1000. Secondary anti-mouse HRP conjugated
antibody (P0447, DAKO, Glostrup, Denmark) was used at a 1:3000
dilution. For detection the SuperSignal West Dura Kit was used
(Thermo Fisher Scientific, Rockford, IL).
We thank F.T. Wieland for the anti-Sar1 antiserum, B.L. Tang for
providing us with the anti-beta-COP antibody, and H.P. Hauri for the
antibodies against p63, giantin and LAMP1. We are grateful to J.
Hagmann for help with deconvolution and M. Wernli for excellent
technical assistance. This study was supported in part by Swiss
National Fonds Grant 3200B0-110040/1 to H.H.H.
G. Unterstab et al. / Virology 399 (2010) 322–331
Appendix A. Supplementary data Download full-text
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.virol.2010.01.011.
Barba, G., Harper, F., Harada, T., Kohara, M., Goulinet, S., Matsuura, Y., Eder, G., Schaff, Z.,
Chapman, M.J., Miyamura, T., Brechot, C., 1997. Hepatitis C virus core protein shows
a cytoplasmic localization and associates to cellular lipid storage droplets. Proc.
Natl. Acad. Sci. U.S.A. 94 (4), 1200–1205.
Brasaemle, D.L., Wolins, N.E., 2006. Isolation of lipid droplets from cells by density
gradient centrifugation. Curr. Protoc. Cell Biol. Chapter 3, Unit 3 15.
Brasaemle, D.L., Dolios, G., Shapiro, L., Wang, R., 2004. Proteomic analysis of proteins
associated with lipid droplets of basal and lipolytically stimulated 3T3-L1
adipocytes. J. Biol. Chem. 279 (45), 46835–46842.
Chesters, P.M., Heritage, J., McCance, D.J., 1983. Persistence of DNA sequences of BK
virus and JC virus in normal human tissues and in diseased tissues. J. Infect. Dis. 147
Darbinyan, A., Siddiqui, K.M., Slonina, D., Darbinian, N., Amini, S., White, M.K., Khalili, K.,
2004. Role of JC virus agnoprotein in DNA repair. J. Virol. 78 (16), 8593–8600.
De Clercq, E., 2004. Antivirals and antiviral strategies. Nat. Rev. Microbiol. 2 (9),
Dorries, K., ter Meulen, V., 1983. Progressive multifocal leucoencephalopathy: detection
of papovavirus JC in kidney tissue. J. Med. Virol. 11 (4), 307–317.
Dropulic, L.K., Jones, R.J., 2008. Polyomavirus BK infection in blood and marrow
transplant recipients. Bone Marrow Transplant. 41 (1), 11–18.
2009. Prevalence of Polyomavirus BK and JC Infection and Replication in 400
Healthy Blood Donors. J. Infect. Dis. 199, 837–846.
Englert, C., Hou, X., Maheswaran, S., Bennett, P., Ngwu, C., Re, G.G., Garvin, A.J., Rosner,
M.R., Haber, D.A., 1995. WT1 suppresses synthesis of the epidermal growth factor
receptor and induces apoptosis. EMBO J. 14 (19), 4662–4675.
isolated from urine after renal transplantation. Lancet 1 (7712), 1253–1257.
Gosert, R., Rinaldo, C.H., Funk, G.A., Egli, A., Ramos, E., Drachenberg, C.B., Hirsch, H.H.,
2008. Polyomavirus BK with rearranged noncoding control region emerge in vivo
in renal transplant patients and increase viral replication and cytopathology. J. Exp.
Med. 205 (4), 841–852.
Grinde, B., Gayorfar, M., Rinaldo, C.H., 2007. Impact of a polyomavirus (BKV) infection
on mRNA expression in human endothelial cells. Virus Res. 123 (1), 86–94.
Guo, Y., Walther, T.C., Rao, M., Stuurman, N., Goshima, G., Terayama, K., Wong, J.S., Vale,
R.D., Walter, P., Farese, R.V., 2008. Functional genomic screen reveals genes
involved in lipid-droplet formation and utilization. Nature 453 (7195), 657–661.
Hope, R.G., McLauchlan, J., 2000. Sequence motifs required for lipid droplet association
and protein stability are unique to the hepatitis C virus core protein. J. Gen. Virol. 81
(Pt. 8), 1913–1925.
Hope, R.G., Murphy, D.J., McLauchlan, J., 2002. The domains required to direct core
proteins of hepatitis C virus and GB virus-B to lipid droplets share common features
with plant oleosin proteins. J. Biol. Chem. 277 (6), 4261–4270.
Ishola Jr., D.A., Post, J.A., van Timmeren, M.M., Bakker, S.J., Goldschmeding, R., Koomans,
H.A., Braam, B., Joles, J.A., 2006. Albumin-bound fatty acids induce mitochondrial
oxidant stress and impair antioxidant responses in proximal tubular cells. Kidney
Int. 70 (4), 724–731.
Jay, G., Nomura, S., Anderson, C.W., Khoury, G., 1981. Identification of the SV40
agnogene product: a DNA binding protein. Nature 291 (5813), 346–349.
Jiang, M., Abend, J.R., Johnson, S.F., Imperiale, M.J., 2009. The role of polyomaviruses in
human disease. Virology 384 (2), 266–273.
Johannessen, M., Myhre, M.R., Dragset, M., Tummler, C., Moens, U., 2008. Phosphor-
ylation of human polyomavirus BK agnoprotein at Ser-11 is mediated by PKC and
has an important regulative function. Virology 379 (1), 97–109.
Khanna, N., Elzi, L., Mueller, N.J., Garzoni, C., Cavassini, M., Fux, C.A., Vernazza, P.,
Bernasconi, E., Battegay, M., Hirsch, H.H., for the Swiss HIV Cohort Study, 2009.
Incidence and outcome of progressive multifocal leukoencephalopathy in 20 years
of the Swiss HIV Cohort Study. Clin. Infect. Dis. 48, 1459–1466.
Khemici, V., Poljak, L., Luisi, B.F., Carpousis, A.J., 2008. The RNase E of Escherichia coli is a
membrane-binding protein. Mol. Microbiol. 70 (4), 799–813.
Knowles, W.A., Pipkin, P., Andrews, N., Vyse, A., Minor, P., Brown, D.W., Miller, E., 2003.
Population-based study of antibody to the human polyomaviruses BKV and JCV and
the simian polyomavirus SV40. J. Med. Virol. 71 (1), 115–123.
Kumar, Y., Cocchiaro, J., Valdivia, R.H., 2006. The obligate intracellular pathogen
Chlamydia trachomatis targets host lipid droplets. Curr. Biol. 16 (16),
Leuenberger, D., Andresen, P.A., Gosert, R., Binggeli, S., Strom, E.H., Bodaghi, S., Rinaldo,
C.H., Hirsch, H.H., 2007. Human polyomavirus type 1 (BK virus) agnoprotein is
abundantly expressed but immunologically ignored. Clin. Vaccine Immunol. 14 (8),
Linstedt, A.D., Hauri, H.P., 1993. Giantin, a novel conserved Golgi membrane protein
containing a cytoplasmic domain of at least 350 kDa. Mol. Biol. Cell 4 (7), 679–693.
Liu, P., Bartz, R., Zehmer, J.K., Ying, Y.S., Zhu, M., Serrero, G., Anderson, R.G., 2007. Rab-
regulated interaction of early endosomes with lipid droplets. Biochim. Biophys.
Acta 1773 (6), 784–793.
Miyanari, Y., Atsuzawa, K., Usuda, N., Watashi, K., Hishiki, T., Zayas, M., Bartenschlager,
R., Wakita, T., Hijikata, M., Shimotohno, K., 2007. The lipid droplet is an important
organelle for hepatitis C virus production. Nat. Cell Biol. 9 (9), 1089–1097.
Moriyama, T., Sorokin, A., 2008. Repression of BK virus infection of human renal
proximal tubular epithelial cells by pravastatin. Transplantation 85 (9),
Myhre, M.R., Olsen, G.H., Gosert, R., Hirsch, H.H., Rinaldo, C.H., 2010. Clinical
polyomavirus BK variants with agnogene deletion are non-functional but rescued
by trans-complementation. Virology (Electronic publication ahead of print).
Nast, C.C., Cohen, A.H., 1985. Renal cholesterol granulomas: identification and
morphological pattern of development. Histopathology 9 (11), 1195–1204.
Ostermeyer, A.G., Ramcharan, L.T., Zeng, Y., Lublin, D.M., Brown, D.A., 2004. Role of the
hydrophobic domain in targeting caveolin-1 to lipid droplets. J. Cell Biol. 164 (1),
Padgett, B.L., Walker, D.L., ZuRhein, G.M., Eckroade, R.J., Dessel, B.H., 1971. Cultivation of
papova-like virus from human brain with progressive multifocal leucoencephalo-
pathy. Lancet 1 (7712), 1257–1260.
Ramos, E., Drachenberg, C.B., Wali, R., Hirsch, H.H., 2009. The decade of polyomavirus
BK-associated nephropathy: state of affairs. Transplantation 87 (5), 621–630.
Rinaldo, C.H., Hirsch, H.H., 2007. Antivirals for the treatment of polyomavirus BK
replication. Expert Rev. Anti. Infect. Ther. 5 (1), 105–115.
Rinaldo, C.H., Traavik, T., Hey, A., 1998. The agnogene of the human polyomavirus BK is
expressed. J. Virol. 72 (7), 6233–6236.
Rust, R.C., Landmann, L., Gosert, R., Tang, B.L., Hong, W., Hauri, H.P., Egger, D., Bienz, K.,
2001. Cellular COPII proteins are involved in production of the vesicles that form
the poliovirus replication complex. J. Virol. 75 (20), 9808–9818.
Safak, M., Barrucco, R., Darbinyan, A., Okada, Y., Nagashima, K., Khalili, K., 2001.
Interaction of JC virus agno protein with T antigen modulates transcription and
replication of the viral genome in glial cells. J. Virol. 75 (3), 1476–1486.
Schweizer, A., Fransen, J.A., Bachi, T., Ginsel, L., Hauri, H.P., 1988. Identification, by a
monoclonal antibody, of a 53-kD protein associated with a tubulo-vesicular
compartment at the cis-side of the Golgi apparatus. J. Cell Biol. 107 (5),
Schweizer, A., Ericsson, M., Bachi, T., Griffiths, G., Hauri, H.P., 1993. Characterization of a
novel 63 kDa membrane protein. Implications for the organization of the ER-to-
Golgi pathway. J. Cell Sci. 104, 671–683.
Shishido-Hara, Y., Ichinose, S., Higuchi, K., Hara, Y., Yasui, K., 2004. Major and minor
capsid proteins of human polyomavirus JC cooperatively accumulate to nuclear
domain 10 for assembly into virions. J. Virol. 78 (18), 9890–9903.
Subramanian, V., Garcia, A., Sekowski, A., Brasaemle, D.L., 2004. Hydrophobic sequences
target and anchor perilipin A to lipid droplets. J. Lipid Res. 45 (11), 1983–1991.
Suzuki, T., Okada, Y., Semba, S., Orba, Y., Yamanouchi, S., Endo, S., Tanaka, S., Fujita, T.,
Kuroda, S., Nagashima, K., Sawa, H., 2005. Identification of FEZ1 as a protein that
interacts with JC virus agnoprotein and microtubules: role of agnoprotein-induced
dissociation of FEZ1 from microtubules in viral propagation. J. Biol. Chem. 280 (26),
Urahama, Y., Ohsaki, Y., Fujita, Y., Maruyama, S., Yuzawa, Y., Matsuo, S., Fujimoto, T.,
2008. Lipid droplet-associated proteins protect renal tubular cells from fatty acid-
induced apoptosis. Am. J. Pathol. 173 (5), 1286–1294.
van Kuppeveld, F.J., Hoenderop, J.G., Smeets, R.L., Willems, P.H., Dijkman, H.B.,
Galama, J.M., Melchers, W.J., 1997. Coxsackievirus protein 2B modifies endo-
plasmic reticulum membrane and plasma membrane permeability and facilitates
virus release. EMBO J. 16 (12), 3519–3532.
Welte, M.A., 2007. Proteins under new management: lipid droplets deliver. Trends Cell
Biol. 17 (8), 363–369.
White, M.K., Khalili, K., 2005. Expression of JC virus regulatory proteins in human
cancer: potential mechanisms for tumourigenesis. Eur. J. Cancer 41 (16),
Zehmer, J.K., Huang, Y., Peng, G., Pu, J., Anderson, R.G., Liu, P., 2009. A role for lipid
droplets in inter-membrane lipid traffic. Proteomics 9 (4), 914–921.
G. Unterstab et al. / Virology 399 (2010) 322–331