The lipodystrophy protein seipin is found at
endoplasmic reticulum lipid droplet junctions
and is important for droplet morphology
Kimberly M. Szymanski*, Derk Binns*, Rene ´ Bartz†, Nick V. Grishin‡, Wei-Ping Li†, Anil K. Agarwal§, Abhimanyu Garg§,
Richard G. W. Anderson†, and Joel M. Goodman*¶
Departments of *Pharmacology,†Cell Biology,‡Biochemistry, and§Internal Medicine, University of Texas Southwestern Medical School, Dallas, TX 75390
Edited by Jan L. Breslow, The Rockefeller University, New York, NY, and approved September 19, 2007 (received for review May 3, 2007)
often accompanied by severe hypertriglyceridemia, insulin resis-
tance, diabetes, and fatty liver. It can be inherited or acquired. The
most severe inherited form is Berardinelli-Seip Congenital Lipo-
dystrophy Type 2, associated with mutations in the BSCL2 gene.
BSCL2 encodes seipin, the function of which has been entirely
unknown. We now report the identification of yeast BSCL2/seipin
through a screen to detect genes important for lipid droplet
morphology. The absence of yeast seipin results in irregular lipid
droplets often clustered alongside proliferated endoplasmic retic-
ulum (ER); giant lipid droplets are also seen. Many small irregular
lipid droplets are also apparent in fibroblasts from a BSCL2 patient.
Human seipin can functionally replace yeast seipin, but a missense
mutation in human seipin that causes lipodystrophy, or corre-
sponding mutations in the yeast gene, render them unable to
complement. Yeast seipin is localized in the ER, where it forms
puncta. Almost all lipid droplets appear to be on the ER, and seipin
is found at these junctions. Therefore, we hypothesize that seipin
is important for droplet maintenance and perhaps assembly. In
addition to detecting seipin, the screen identified 58 other genes
whose deletions cause aberrant lipid droplets, including 2 genes
encoding proteins known to activate lipin, a lipodystrophy locus
in mice, and 16 other genes that are involved in endosomal–
lysosomal trafficking. The genes identified in our screen should be
of value in understanding the pathway of lipid droplet biogenesis
and maintenance and the cause of some lipodystrophies.
BSCL2 ? lipid bodies
sterol esters (1–3). These conditions can be acquired or inher-
ited, partial or generalized. A notable example of acquired
partial lipodystrophy occurs in HIV-infected patients on highly
active retroviral therapy (HAART) who lose subcutaneous
adipose tissue from the face and limbs but gain it elsewhere, such
as the upper back and neck (4). The most severe lipodystrophies
are the Berardinelli-Seip congenital generalized forms. Patients
affected are born with little or no adipose tissue. As children,
they have ravenous appetites and grow more rapidly than
normal. In general, patients with lipodystrophy have high cir-
culating levels of triglycerides, develop deposits of fat in their
muscles and liver, and often acquire insulin resistance and
diabetes, similar to obese patients.
Two Berardinelli-Seip Congenital Lipodystrophy (BSCL)
genes have thus far been identified. BSCL1 encodes an acylg-
lycerol phosphate acyltransferase, which catalyzes a critical step
in the biosynthesis of triglycerides (5). The more severe form of
the disease, in which adipose tissue is virtually absent, is caused
by mutations in BSCL2. BSCL2 encodes a protein, seipin, the
function of which is entirely unknown (6, 7); a role in the
differentiation of mesenchymal cells into preadipocytes has been
postulated to explain the lack of adipose tissue (6, 8). Human
ipodystrophies are disorders in the development or mainte-
nance of adipose tissue, the storage site for triglycerides and
seipin is predicted from its primary sequence to span a mem-
brane twice, with both termini facing the cytoplasm, and a
glycosylation site in the luminal segment (9); seipin-GFP
partially localizes to the endoplasmic reticulum (ER) (10).
Interestingly, mutations in the glycosylation site cause Silver
syndrome and motor neuropathy (10), a result of a severe ER
stress response (11). Whereas BSCL2 mutations that cause
lipodystrophy are recessive, Silver syndrome is the result of a
We reasoned that defects in the assembly of lipid droplets
(lipid bodies; adiposomes) might cause lipodystrophy, because
several lines of evidence suggest that lipid droplets are derived
from ER (12), there is no agreement on mechanism. Most
models offer a budding pathway, whereas an alternative idea is
that they are ‘‘nursed’’ alongside the ER (13). No ER assembly
factors have been identified. To this end, we screened a yeast
deletion library for aberrant lipid droplets. We report the results
of the screen, including the identification of yeast seipin as a
factor in lipid droplet assembly or maintenance.
To identify genes in this pathway, we screened 4,936 yeast
deletion clones for lipid droplet abnormalities by growing the
strains overnight and treating them with BODIPY 493/503, a dye
commonly used to stain lipid droplets (14). Most strains showed
a wild-type phenotype, an average of 5–6 brightly stained lipid
droplets of fairly uniform size per cell (Fig. 1). Although there
was significant variability in lipid droplets among cells of any
particular strain, inspection of several fields of cells revealed 59
deletion strains with clearly different phenotypes (Table 1).
Traits were scored for differences in droplet number, size,
dispersion throughout the cytoplasm (droplets in some strains
were aggregated or centralized around the nuclear envelope),
staining intensity, resolution of droplets from the cytoplasm or
from each other (clear or indistinct), or a combination [support-
ing information (SI) Table 2]. To determine the dependence of
phenotype on growth conditions, each strain was observed both
in mid-log and stationary phase. Interestingly, 14 strains showed
a wild-type phenotype in log phase but developed a high number
of lipid droplets in stationary culture (SI Table 3).
R.B., N.V.G., R.G.W.A., and J.M.G. analyzed data; and J.M.G. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
¶To whom correspondence should be addressed at: Department of Pharmacology, Univer-
sity of Texas Southwestern Medical School, 5323 Harry Hines Boulevard, Dallas, TX
75390-9041. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
December 26, 2007 ?
vol. 104 ?
no. 52 www.pnas.org?cgi?doi?10.1073?pnas.0704154104
The largest functional group of proteins represented in Table
1 are those in endosomal/vacuolar (lysosomal) trafficking (16
genes), of which five encode subunits of the vacuolar ATPase,
indicating the direct or indirect importance of this pump to lipid
droplet assembly and/or maintenance. Defects in retrograde
transport from Golgi to ER (in the erd1? and sec22? strains)
suggest that factors are lost from the ER that are important for
droplet morphology. Several genes encoding mitochondrial and
endoplasmic reticular proteins also gave aberrant phenotypes;
these two organelles apparently make tight contacts with lipid
droplets (15). Also in the gene collection were SPO7 and NEM1,
which form a membrane complex that activates PAH1 (16), the
ortholog of lipin, a phosphatidic acid hydrolase (17) that is
defective in a form of mouse lipodystrophy (18). Indeed, a pah1?
strain obtained from Carman (17) showed a similar phenotype-
neutral lipid frequently localized to the ER (data not shown).
We focused on determining the identity of YLR404w, a gene
of unknown function. Combined analysis of primary sequence
and predicted structure of multiple sequences by using PRO-
MALS (19) strongly suggested that YLR404w encodes the
Saccharomyces cerevisiae seipin (Fig. 2A). Seipins share a con-
served structure in animals, plants, and fungi. As is true for
human seipin, all members of the family share a core structure,
?210 aa residues in length, flanked by two predicted transmem-
brane helices, one at each end (red cylinders), with several
additional patches of hydrophobic residues (yellow). The amino
and carboxyl termini of the yeast protein have been predicted to
face the cytosol with the large central domain in a luminal
illustrated in Fig. 2A comprises most of yeast seipin, but other
species have extensions at either or both ends, which display low
In addition to decreasing the number of lipid droplets, dele-
tion of the yeast seipin homolog drastically alters lipid droplet
morphology when cells are grown in glucose medium, as shown
by electron microscopy (Fig. 3 Aa–Af). Lipid droplets in normal
cells are small electron-transparent organelles (Fig. 3Aa). In
contrast, the deletion strain typically contained clusters of
organelles of complex morphology, with both electron-
transparent and -opaque areas (Fig. 3Ab, arrow; higher resolu-
tion in Fig. 3Ad). The opaque regions could sometimes be
resolved as layers of membrane. Membranes continuous with
both cortical and perinuclear ER often penetrated the cluster
and were frequently engorged with lipid (Fig. 3Af). Organelles
within the cluster sometimes appeared as grapes on a stem of ER
(Fig. 3Ae). A less common phenotype for the strain was the
presence of very large lipid droplets (Fig. 3Ac). The images
suggest that lipid droplet assembly is unregulated in the absence
of seipin, yielding immature organelles that bud chaotically from
The aberrant assembly of lipid droplets in YLR404w-deficient
cells is further highlighted when cells are grown on oleic acid.
Growth of wild-type yeast on fatty acids causes the production
of large lipid droplets of fairly uniform size (21) (Fig. 3Ag). In
contrast, seipin-deficient cells produced a large number of lipid
droplets of widely varying sizes and often of irregular shapes
(Fig. 3Ah), including giant ones (Fig. 3Ai), which might represent
products of fusion of smaller organelles.
To determine whether protein targeting to lipid droplets is
disrupted in the YLR404w knockout (KO), we monitored the
localization of fluorescently tagged Erg6p, a protein abundant in
yeast lipid droplets (22). Erg6p localized to individual lipid
droplets in the wild-type strain and to the ER/lipid droplet
clusters in the seipin KO strain, indicating that protein targeting,
(Scale bar, 2 ?m.)
Examples of strains with aberrant lipid body morphology. Cells were harvested in log or stationary phase, and lipid droplets were stained with BODIPY.
Table 1. Genes involved in lipid body morphology
Endosome/vacuole FAB1, PEP5, PEP7, PPA1, TFP1, VAC14, VMA2, VMA4,
VMA6, VMA7, VMA22, VPS4, VPS16, VPS21, VPS24,
VPS35, VPS51, VPS66
ERD1, OST4, SCP160, SEC22, SPO7, YIL039W
ATP3 MDM20, MRM2, TOM5
ANP1, CHC1, MOG1
BEM2, CNM67, SLA2, SRV2
ADE8, ADE12, ADK1
BUR2, DOA1, HPR1, KEM1, MED2, MSN1, PAF1, ROX3,
APN1, ECM1, EST3, NEM1
BUD32, DRS2, PLC1, SSD1, TPD3
Other trafficking pathways
Actin or tubulin related
Base and nucleotide metabolism
Other nuclear activities
Genes whose deletions produce aberrant lipid bodies are grouped by localization or function, based on
Szymanski et al.
December 26, 2007 ?
vol. 104 ?
no. 52 ?
at least to the vicinity of lipid droplets, does not require seipin
(Fig. 3C). To confirm the identity of ER membranes in the
cluster, we expressed GFP fused to a secretion signal sequence
and the four amino acid ER retention signal HDEL. Interest-
ingly, ?90% of lipid droplets in wild-type cells appeared in close
proximity to the ER, suggesting that they may remain attached.
In the mutant, a bright patch of labeled ER corresponded to the
lipid droplet clusters (arrow), suggesting a proliferation of ER
within this structure (Fig. 3D).
We studied a BSCL2 fibroblast line [taken from a patient with
a nonsense mutation (23)] to see whether human cells also
showed aberrant droplets. Compared with normal human fibro-
blasts, the seipin-deficient cells had many smaller lipid droplets,
often not resolved from each other when stained either with Oil
ADRP (24) (Fig. 3B). Oleate caused accumulation of lipid
droplets in normal fibroblasts and a more intense staining of the
small lipid droplets in the mutant. Electron microscopy con-
firmed that the cytoplasm of the mutant cells was filled with
many tiny clear organelles, which we assume to be lipid droplets,
instead of the few distinct ones seen in normal cells (Fig. 3E).
To test the functional relationship between human and yeast
seipin, human seipin was expressed in the ylr404w? deletion strain.
This protein can be generated from two putative alternative trans-
lational start sites from mRNA of different lengths, yielding pro-
teins of 398 and 462 aa (6, 9). Both forms complemented the yeast
the yeast protein with respect to morphology (Fig. 2B) and number
of lipid droplets (Fig. 2C). In contrast, the missense mutation
A212P in human seipin that causes lipodystrophy (7) failed to
complement. Two analogous mutations in the yeast protein (S224P
human A212P) only weakly complemented the strain. Thus, we
conclude that YLR404w encodes yeast seipin and that the human
and yeast proteins have similar functions in lipid droplet assembly
To gain information on the role of seipin in lipid droplet
function, we determined its localization in yeast. When tagged at
either the amino or carboxyl terminus with mCherry and ex-
pressed from the strong PGK promoter, the protein was clearly
localized to the ER based on colocalization with GFP-HDEL
(Fig. 4A). Human seipin also resides in the ER (10). However,
the pattern of seipin in the ER was not uniform; patches of seipin
(PSIPRED) (29) is shown below the sequences. Predicted transmembrane ?-helices are colored red. (B and C), Complementation of lipid droplet phenotype in
ylr404w? by human and yeast seipins. Wild-type or the seipin-deleted strain was transformed with empty plasmid (pRS315) or plasmid-containing sequences
encoding the long or short form of human seipin (HSeipinL or HSeipinS), A212P human seipin, yeast seipin (YSeipin), or yeast seipin with the mutations S224P
or G225P. BODIPY-stained images are shown in B (scale bar, 5 ?m); the histogram in C shows the number of lipid droplets per cell.
YLR404w is yeast BSCL2/seipin. (A) Alignment of seipins by using PROMALS (19). NCBI gene identification numbers (gi numbers) are shown after the
www.pnas.org?cgi?doi?10.1073?pnas.0704154104Szymanski et al.
were observed with overexpressed protein. We looked in more
detail at the colocalization of droplets and ER (Fig. 4B). When
of lipid droplets (values obtained from two observers) could not
remain tethered to the ER. When droplets, ER and seipin were
simultaneously imaged by using BODIPY, CFP-HDEL, and
chromosomally expressed seipin-mCherry (which was active and
punctate), most lipid droplets colocalized or overlapped with
as indicated. Lipid droplets are in clusters in the KO (arrow in b, and in higher resolution in d–f). In oleate medium, lipid droplets in wild type are of uniform size
(g) but are irregular in size, shape, and number in the KO strain (h and i). (All scale bars, 200 nm.) ER, endoplasmic reticulum; LB, lipid droplet; N, nucleus; V,
vacuole. (B) Lipid droplets are aberrant in seipin-deficient human fibroblasts. Cells were stained with Hoechst and either Oil Red O (with or without culturing
with oleate) or antibodies against the lipid droplet marker protein ADRP (without oleate). (Scale bars, 10 ?m.) (C) Erg6p is targeted to the aberrant clusters
(arrows), suggesting normal protein targeting. The merge image also incorporates brightfield to outline the cells. (Scale bar, 5 ?m.) (D) The ER is concentrated
in the aberrant clusters (arrows). Arrowhead indicates a giant lipid droplet. Note that lipid droplets appear to be attached to ER in the wild-type strain. (Scale
bar ? 5 ?m.) (E) Proliferation of small lipid droplets in seipin-deficient fibroblasts, grown without oleate. Note the presence of many small clear organelles in
the mutant cells. (Scale bars, 2 ?m.)
Abnormal lipid droplets in seipin-deficient cells. (A) ultrastructure of wild-type or the seipin KO yeast strains. Cells were cultured in glucose or oleate
Both seipin and GFP-HDEL were expressed on plasmids and driven by the PGK promoter. (B) Most lipid droplets appear bound to the ER. Droplets that were
chromosomally tagged with Erg6p-mCherry also expressed the GFP-HDEL ER marker. (C) Seipin marks docking sites of lipid bodies. Seipin-mCherry, or inactive
seipin-G225P-mCherry was chromosomally expressed at the seipin locus. CFP-HDEL, driven by the PGK promoter, was plasmid expressed. Lipid droplets were
stained with BODIPY. Yellow boxes frame two cells showing clear seipin droplet proximity. Droplet clusters in cells expressing the mutant protein (null
background) also colocalize seipin. (Scale bars, 5 ?m.)
Yeast seipin localizes in the ER and marks sites of contact with lipid droplets. (A) Seipin, tagged at either terminus with mCherry, localizes to the ER.
Szymanski et al.
December 26, 2007 ?
vol. 104 ?
no. 52 ?
4C). Colocalization was even clearer with inactive seipin G225P,
where lipid droplet clusters were intensely stained with seipin.
These results strongly suggest that lipid droplets normally reside
on the ER and that seipin is found at the junctions.
We report the identification of 59 genes that are important for
normal lipid droplet morphology in yeast, which should help to
disparate classes of protein and organellar functions apparently
collaborate to maintain droplets. Some of these, for example
transcription factors, would alter droplets simply by modifying
basal metabolic rates leading to changes in droplet volume. The
appearance of several genes that control vesicle transport
through the central vacuolar system in the gene set, however,
suggests a major role for these pathways. Although much work
is required to understand the action of these genes in droplet
59 validates the screen as a means to approach the study of these
The data presented here indicate that lipid droplets are aberrant
proximity to the ER and that seipin is in these apparent junctions
organelles. For example, the ER may serve as a transient reservoir
acid esterification, ER droplet junctions may provide a direct path
for enlarging the droplet without a requirement for de novo
biogenesis of more organelles. Although this concept appears not
to be consistent with the proliferation of droplets in yeast and
human fibroblasts in the absence of seipin, this phenotype could
reflect a cellular response to a deficiency in droplet function rather
than a direct effect of the absence of seipin itself. Determining the
kinetics of the appearance of the phenotype vs. the disappearance
of seipin expressed on a regulated promoter should address this
issue. Although the precise role of seipin is not clear, the seipin-
deficient phenotype in yeast, where the ER is often found wrapped
of the two organelles to each other.
Our data strongly suggest that failure to form normal lipid
droplets in adipocytes or adipocyte precursors is the primary
cause of BSCL2 disease. Patients deficient in BSCL2/seipin lack
essentially all adipose tissue even though BSCL2 mutant fibro-
blasts can still produce lipid droplets, albeit aberrant ones. In
contrast, patients develop fatty liver and muscles, which our data
suggest is caused at least in part to a lack of seipin in these cells
rather than high circulating triglyceride levels. Why is there no
apparent adipose tissue in BSCL2 patients? It is possible that
adipocytes are more dependent on seipin compared with other
cells for generation of lipid droplets and that identifiable adi-
pocytes fail to appear in the absence of seipin. Alternatively,
formation of structurally and functionally normal lipid droplets
of seipin, preadipocytes either revert or are shunted to an
apoptotic pathway. Such a mechanism would clarify the previous
idea that seipin plays some role in differentiation of mesenchy-
mal cells into adipocytes (6, 8). An answer will require a detailed
study of adipocyte differentiation in BSCL2-deficient animals.
Materials and Methods
deletion library (Open Biosystems) were cultured overnight in yeast extract/
peptone/dextrose (YPD) in 96-well plates, stained for 10 min with 1.25 ?g/ml
BODIPY493/503 (Invitrogen), and observed by fluorescence microscopy. At least
five fields of cells from each strain were inspected. Strains with aberrant lipid
droplets were recultured and checked at least two more times by using two lots
of the library. Sixty-three strains were identified with altered lipid droplet mor-
phology (by K.M.S.). For assignment into phenotypic groups, log phase and
stationary phase cells were obtained by culturing four dilutions of each strain
8–12) of each culture were stained with BODIPY and imaged by obtaining
projections of ?20 z-sections through the cell (3 ?m spacing between planes).
methods described in ref. 15. Cells were fixed with 1% glutaraldehyde and
coded, and three of us (K.M.S., D.B., and J.M.G.) collaborated to group strains
with similar morphology, considering images from log and stationary cells inde-
pendently. Four of the original 63 strains were indistinguishable from wild type
in this secondary analysis and were not considered further. The pah1? strain (in
the W303 background) was a kind gift from Gil-Soo Han and George Carman
(Rutgers University, New Brunswick, NJ).
Yeast Complementation Studies. Cells were transformed with pRS315 (25) that
contained yeast seipin or the long or short forms of human seipin (IMAGE
clones 3533654 and 3939021; Open Biosystems). Clone 3939021 contains a
nine-base insertion in the coding region; its presence or absence did not alter
complementation. Expression was driven by the yeast PGK1 promoter (15),
and cells were cultured in synthetic dextrose (2%) medium. Site-directed
mutagenesis to generate missense mutations was performed by using Pfu
Colocalization. For studies with the lipid droplet protein marker Erg6p, mCherry
expressed a form of GFP-HDEL secreted into the ER lumen from pDN330, a kind
PGK (15) to drive expression from the PGK promoter. For some experiments
enhanced cyan fluorescent protein (ECFP) (kindly provided by Scott Gibson,
in pDN330 before insertion of the HDEL-containing fragment into pRS315-PGK.
terminus of yeast seipin and expressed in pRS315-PGK.
Electron Microscopy. Yeast cells were grown in glucose or oleate medium and
processed for electron microscopy as described in ref. 15. Human cells were
processed as described in ref. 28.
Fluorescence Microscopy of Human Fibroblasts. Primary human fibroblasts
were obtained from a patient with a BSCL2 mutation (n659delGTATC;
pF105fsx 111) (23) or from a normal healthy volunteer (written informed
consent and approval by the institutional review board were obtained) and
cultured in DMEM with 4.5 g/liter glucose and 10% FBS. For immunofluores-
were then either stained for neutral lipids with Oil Red O or permeabilized
with 0.1% saponin and then stained with a monoclonal antidroplet to ADRP
(Research Diagnostics). DNA was stained by using Hoechst 34580. Cells were
observed with a Zeiss Axioplan 2E microscope using a Plan-Neofluar 40?/1.3
oil differential interference contrast objective.
ACKNOWLEDGMENTS. We thank Tom Januszewski for expert technical assis-
tance with yeast electron microscopy; Elliott Ross for helpful suggestions
throughout this project; and John Zehmer, Pingsheng Liu, and Joe Albanesi
of Health Grants HL20948 and GM52016, The Perot Foundation, a Cecil H.
Green Endowed Chair (to R.G.W.A.), National Institutes of Health Grant
DK54387 (to A.G.), The Welch Foundation I-1085, American Heart Association
Texas Affiliate 0555043Y, and National Science Foundation Grant MCB-
0455329 (to J.M.G.).
1. Agarwal AK, Barnes RI, Garg A (2004) Int J Obes Relat Metab Disord 28:336–339.
2. Garg A (2004) N Engl J Med 350:1220–1234.
3. Simha V, Garg A (2006) Curr Opin Lipidol 17:162–169.
A (2002) Nat Genet 31:21–23.
6. Agarwal AK, Garg A (2004) Trends Mol Med 10:440–444.
7. Magre J, Delepine M, Khallouf E, Gedde-Dahl T, Jr, Van Maldergem L, Sobel E, Papp J,
Meier M, Megarbane A, Bachy A, et al. (2001) Nat Genet 28:365–370.
8. Agarwal AK, Garg A (2006) Annu Rev Genomics Hum Genet 7:175–199.
9. Lundin C, Nordstrom R, Wagner K, Windpassinger C, Andersson H, von Heijne G,
Nilsson I (2006) FEBS Lett 580:2281–2284.
www.pnas.org?cgi?doi?10.1073?pnas.0704154104Szymanski et al.
10. Windpassinger C, Auer-Grumbach M, Irobi J, Patel H, Petek E, Horl G, Malli R, Reed JA, Download full-text
Dierick I, Verpoorten N (2004) Nat Genet 36:271–276.
11. Ito D, Suzuki N (2007) Ann Neurol 61:237–250.
12. Wolins NE, Brasaemle DL, Bickel PE (2006) FEBS Lett 580:5484–5491.
13. Robenek H, Hofnagel O, Buers I, Robenek MJ, Troyer D, Severs NJ (2006) J Cell Sci
15. Binns D, Januszewski T, Chen Y, Hill J, Markin VS, Zhao Y, Gilpin C, Chapman KD,
Anderson RG, Goodman JM (2006) J Cell Biol 173:719–731.
17. Han GS, Wu WI, Carman GM (2006) J Biol Chem 281:9210–9218.
18. Peterfy M, Phan J, Xu P, Reue K (2001) Nat Genet 27:121–124.
19. Pei J, Grishin NV (2007) Bioinformatics 23:802–808.
20. Kim H, Melen K, von Heijne G (2003) J Biol Chem 278:10208–10213.
21. Veenhuis M, Mateblowski M, Kunau WH, Harder W (1987) Yeast 3:77–84.
22. Athenstaedt K, Zweytick D, Jandrositz A, Kohlwein SD, Daum G (1999) J Bacteriol
23. Agarwal AK, Simha V, Oral EA, Moran SA, Gorden P, O’Rahilly S, Zaidi Z, Gurakan
F, Arslanian SA, Klar A, et al. (2003) J Clin Endocrinol Metab 88:4840–
24. Londos C, Brasaemle D, Schultz C, Segrest J, Kimmel A (1999) Semin Cell Dev Biol
25. Sikorski RS, Hieter P (1989) Genetics 122:19–27.
27. Gibson SK, Gilman AG (2006) Proc Natl Acad Sci USA 103:212–217.
28. del Pozo MA, Balasubramanian N, Alderson NB, Kiosses WB, Grande-Garcia A, Ander-
son RG, Schwartz MA (2005) Nat Cell Biol 7:901–908.
29. Jones DT (1999) J Mol Biol 292:195–202.
Szymanski et al.
December 26, 2007 ?
vol. 104 ?
no. 52 ?