Proteasome activator complex PA28 identified as an
accessible target in prostate cancer by in vivo selection
of human antibodies
David Sánchez-Martína,1, Jorge Martínez-Torrecuadradab, Tambet Teesaluc,d,e, Kazuki N. Sugaharac,d,
Ana Alvarez-Cienfuegosa, Pilar Ximénez-Embúnb, Rodrigo Fernández-Periáñeza, M. Teresa Martínf,
Irene Molina-Privadoa, Isabel Ruppen-Cañásb, Ana Blanco-Toribioa, Marta Cañamerog, Ángel M. Cuestaa,2,
Marta Comptea, Leonor Kremerf, Carmen Bellash, Vanesa Alonso-Caminoa,3, Irene Guijarro-Muñoza, Laura Sanza,
Erkki Ruoslahtic,d, and Luis Alvarez-Vallinaa,4
aMolecular Immunology Unit, Hospital Universitario Puerta de Hierro Majadahonda, 28222 Majadahonda, Madrid, Spain;bProteomics Unit, Centro Nacional
de Investigaciones Oncológicas, 28029 Madrid, Spain;cCancer Center, Sanford-Burnham Medical Research Institute, La Jolla, CA 92037;dCenter for
Nanomedicine, University of California, Santa Barbara, CA 93106;eLaboratory of Cancer Biology, Institute of Biomedicine, Centre of Excellence for
Translational Medicine, University of Tartu, 50090 Tartu, Estonia;fProtein Tools Unit, Centro Nacional de Biotecnología, Consejo Superior de Investigaciones
Científicas, 28049 Madrid, Spain;gHistopathology Unit, Centro Nacional de Investigaciones Oncológicas, 28029 Madrid, Spain; andhPathology Department,
Hospital Universitario Puerta de Hierro Majadahonda, 28222 Majadahonda, Madrid, Spain
Edited by Richard A. Lerner, The Scripps Research Institute, La Jolla, CA, and approved July 9, 2013 (received for review January 3, 2013)
Antibody cancer therapies rely on systemically accessible targets
and suitable antibodies that exert a functional activity or deliver
a payload to the tumor site. Here, we present proof-of-principle of
in vivo selection of human antibodies in tumor-bearing mice that
identified a tumor-specific antibody able to deliver a payload and
unveils the target antigen. By using an ex vivo enrichment process
against freshly disaggregated tumors to purge the repertoire, in
combination with in vivo biopanning at optimized phage circula-
tion time, we have identified a human domain antibody capable
of mediating selective localization of phage to human prostate
cancer xenografts. Affinity chromatography followed by mass
spectrometry analysis showed that the antibody recognizes the
proteasome activator complex PA28. The specificity of soluble an-
tibody was confirmed by demonstrating its binding to the active
human PA28αβ complex. Whereas systemically administered con-
trol phage was confined in the lumen of blood vessels of both
normal tissues and tumors, the selected phage spread from tumor
vessels into the perivascular tumor parenchyma. In these areas,
the selected phage partially colocalized with PA28 complex. Fur-
thermore, we found that the expression of the α subunit of PA28
[proteasome activator complex subunit 1 (PSME1)] is elevated in
primary and metastatic human prostate cancer and used anti-
PSME1 antibodies to show that PSME1 is an accessible marker in
mouse xenograft tumors. These results support the use of PA28 as
a tumor marker and a potential target for therapeutic intervention
in prostate cancer.
phage library|phage display|single domain antibody|
physiological selection|tumor-associated antigen
been growing rapidly in the last decade. Fully human anti-
bodies can be selected either from transgenic animals or from
large phage-displayed antibody libraries (1). There is, however,
a shortage of disease-specific targets for therapeutic antibodies.
In fact, just 5 targets constitute one-third of the 55 human mAb
anticancer candidates against known targets (1). Unbiased func-
tional identification of clinically relevant antibodies and their tar-
gets is an important goal. Developing an mAb without prior
knowledge of the target, following a functional screening, has
a high potential for innovation (2). In this approach, mAbs are
selected based on their ability to bind to complex targets or to elicit
a biological response, and their corresponding targets are charac-
terized afterward, usually by a proteomics strategy. Using func-
tional screens in vitro, however, some of the antibodies selected fail
to fulfill their intended role in the clinic due to the differences with
the in vivo environment, where they might exert their therapeutic
he use of monoclonal antibodies (mAbs) in the clinic has
function. One of the aspects that need to be considered when
selecting antibodies for therapy is the importance of ensuring their
ability to reach the target. Several strategies are used to identify
ligands that are accessible from bloodstream, including functional
genomics analysis (3), subtractive proteomics (4), the in vivo bio-
tinylation of vascular proteins (5), and an in vivo phage display
screening for peptides that home to specific targets in the vascu-
In vivo peptide phage display has been particularly effective
in the identification of markers that distinguish the vessels of
diseased tissues from normal vessels (11). Homing moieties—
mainly peptides and recombinant antibodies—have been used in
targeted delivery of therapeutic compounds to diseased organs
and represent a promising area of pharmaceutical intervention
(12–14). Whereas antibodies have certain advantages over pep-
tides as targeting agents (e.g., higher affinity and longer circu-
lation time) (15), in vivo antibody screening in tumor-bearing
mice has not been accomplished due, among others, to technical
limitations such as unspecific binding of phage clones. Here, we
overcome some of the limitations and report the isolation of
a prostate tumor-homing antibody (011H12) from a human VH
domain antibody library (DAb library), the identification of its
receptor, and the subsequent validation in primary and meta-
static human prostate cancer samples.
Repertoire Enrichment Strategy. In preparation for in vivo screen-
ing, we hypothesized that the ideal repertoire should be moder-
ately enriched against the target organ, leaving enough diversity
for a variety of antigens in the target tissue. We compared
enrichment of the phage antibody library in phage that bind to
Author contributions: D.S.-M., J.M.-T., T.T., K.N.S., L.K., E.R., and L.A.-V. designed research;
D.S.-M., J.M.-T., T.T., K.N.S., A.A.-C., P.X.-E., R.F.-P., M.T.M., I.M.-P., I.R.-C., A.B.-T., Á.M.C.,
M. Compte, V.A.-C., and I.G.-M. performed research; D.S.-M., J.M.-T., T.T., K.N.S., P.X.-E.,
M.T.M., I.R.-C., M. Cañamero, L.K., C.B., L.S., E.R., and L.A.-V. analyzed data; and D.S.-M.,
J.M.-T., T.T., K.N.S., E.R., and L.A.-V. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1Present address: Laboratory of Cellular Oncology, Center for Cancer Research, National
Cancer Institute, National Institutes of Health, Bethesda, MD 20892.
2Present address: Buchmann Institute of Molecular Life Science, Goethe University Frank-
furt, 60438 Frankfurt, Germany.
3Present address: Department of Molecular Medicine, Mayo Clinic, Rochester, MN 55905.
4To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| August 20, 2013
| vol. 110
| no. 34
cultured tumor cells (in vitro strategy) or cell suspensions from
freshly excised tumors (ex vivo strategy). To avoid overselection
for antibody clones to tumor antigens inaccessible from the cir-
culation, we monitored not only the increase in the recovery (Fig.
1A) but also the diversity loss during the rounds of enrichment
(Fig. 1B). Sequencing 96 clones from the input and output of each
selection round confirmed the expectation that in vitro selection
resulted in a bias toward a few dominant sequences whereas ex
vivo selection of the phage repertoire enriched a more diverse
repertoire of phage clones.
Optimizing Circulation Time and Selection of Candidates. A circula-
tion time of 15 min is considered to be sufficient to address most
of the targets exposed in the endothelium in nonmalignant tis-
sues but may be too short for dysfunctional tumor vessels (16). A
circulation time of 1 h has been proposed for a filamentous
phage peptide library (17) based on the half-life of the phage
although there is evidence that the displayed exogenous protein
or peptide can make the circulation half-life as short as 1.5 min
and as long as 4.5 h (18, 19). We administered i.v. a phage pool
that had been previously enriched ex vivo and compared the
recovery of phage from different organs at different time points
after injection [ranging from 15 min to 24 h, using 4 h half-life of
wild-type phage as a reference (20)]. The number of phage re-
covered from the tumor at 15 min after the injection was ∼40
times higher than at 24 h; in the nonmalignant organs, this dif-
ference was ∼100-fold (Fig. 1C, Upper). This difference is more
easily visualized using a value normalized to control phage (Fig.
1C, Lower). These results suggest that specific accumulation in the
target organ is nearly four times greater at 24 h than at 15 min.
Having established a 24-h circulation time as well suited for
the in vivo selections, a sample of clones was sequenced from the
input pool, from phage recovered from tumor tissue and from
reference organs. Candidate tumor-homing phage clones were
appraised based on several criteria: abundance in the target out-
put, similarity between self and consensus complementarity de-
termining regions (CDR) sequences obtained from hierarchical
clustering, presence and composition of the amber stop codons,
and presence and composition of “common known homing
motifs” [such as integrin-binding RGD, neuropilin-1 binding
monoclonal phage. The recovery of phage increases
as the ex vivo selections progress. The recovery is
greater if the repertoire was previously enriched
in vitro (A). However, this preenrichment greatly
reduces the available diversity (B). (C) Influence of
the circulation time. Using an ex vivo enriched rep-
ertoire as input, the total recovery of infective par-
ticles is maximum at 15 min and drops gradually
until 24 h (Upper). This recovery is observed for all of
the organs analyzed. However, as the time of circu-
lation increases, there is a higher specific recovery from
the tumor (Lower). T, tumor; K, kidney; L, liver; H,
heart. (D) Competition experiments using mixtures of
fresh monoclonal phage. After 24 h of circulation, the
phage were recovered from the organs or the tumor,
and the percentage of each was compared with that
present in the input (Upper) or with the average of the
organs (Lower). (E) Normalized values showing the
differential accumulation of phage clones (n = 5 mice
per group). (i) α–βGal phage used as a reference: the
expected retention for a phage whose target is not
present in the tissue. Different clones that show re-
tention in the tumor (iii–v), and retention in all of the
organs (ii). Significant differences (P value) U Mann–
Enrichment, diversity loss, and homing of
| www.pnas.org/cgi/doi/10.1073/pnas.1300013110 Sánchez-Martín et al.
tissue penetrating C-end Rule, CD13 binding NGR, etc. (21–23)
(SI Appendix, Tables S1 and S2 and Fig. S1)]. We also explored
a computer-based approach (24) by analyzing the frequency of all
possible combinations of three amino acids among the CDRs for
the in vivo output of a single preenriched repertoire. Despite the
fact that the number of sequences analyzed in this way was limited,
we found (SI Appendix, Fig. S2) significant differences in the
sequences obtained from the tumor (TCs) and from the organs
(OCs), suggesting that deep sequencing (25) would be helpful in
identifying new candidate binders and consensus sequences and in
validating previously selected candidates. The similarity heat map
for the selected sequences and this computed group of consensus
sequences (SI Appendix, Fig. S2, Bottom) show that TCs have more
triplets of amino acids also present in other TCs than in OCs. OCs
have a similar number of triplets also present in TCs and OCs, and
the selected candidates mostly resemble the distribution pattern of
TCs. The selected candidate clones were finally evaluated in an in
vivo play-off (competitive homing) experiment and compared with
a reference irrelevant clone that harbors the anti-β-galactosidase
(α−βGal) specificity. Comparison of the recovery to the phage
representation in the input pool (Fig. 1D, Upper) or to the average
from the normal organs (Fig. 1D, Lower) yielded four promising
phage clones (Fig. 1D, arrows) for further analysis.
Homing Assays. The ability of the four selected clones to home to
the tumor was compared with the reference α−βGal phage. A
fresh preparation of each phage clone was i.v. injected into
a group of mice (n = 5 per group), with the circulation time set at
24 h (Fig. 2). To validate the reproducibility of the normalized
value used, two additional mice were injected with a 1:100 di-
lution of one of the selected clones to compare the total amount
of phage recovered and the normalized value for a given speci-
ficity and different dosage (SI Appendix, Fig. S3). The reference
α−βGal phage (Fig. 1E, i) showed a baseline accumulation in the
tumor, similar to the accumulation in the kidney, and a lower
accumulation in both heart and liver. Three of the four selected
clones showed a higher uptake in the tumor (010H01, 011H12,
and 010D6; Fig. 1E, iii–v) whereas the remaining clone, which
has one of the lowest scores in the heat map (SI Appendix, Fig.
S2, Bottom), accumulated in all of the organs (011A10; Fig. 1E,
ii). We focused on the clone designated as 011H12 for further
characterization based on the distribution observed.
011H12 Phage Accumulates in Tumors. Phage may be prone to un-
specific aggregation (26, 27) and/or interaction with tissue com-
ponents accessible from the bloodstream. Therefore, we decided
to investigate whether the retention observed for the different
phage was specific to the antibody harbored, or merely the result
of entrapment of phage aggregates in the tumor vascular tree. If
the latter were true, (i) a colocalization of endothelial markers
with the i.v. administered phage should be expected; and (ii) the
differences in the amount of phage recovered should be related to
differences in the number of vessels.
We compared the endothelial staining of the tumors from
mice dosed with the reference phage, with tumors of mice that
had received the 011H12 clone. Analysis of equally vascularized
tumors from mice injected with the 011H12 phage or the ref-
erence phage (Fig. 2A) showed greater accumulation of the
011H12 phage in the tumors both when the total area positive for
the phage staining (Fig. 2B), or the amount of phage over en-
dothelium area (Fig. 2C), was compared (P < 0.0001, Fig. 2D).
The high dispersion of the values reflects the heterogeneous ac-
cumulation of the phage (Fig. 2E); whereas the reference phage
was confined to the lumen of the vessels (Fig. 2F, Top), the 011H12
phage was also found spreading from the vessels into wide areas of
surrounding perivascular stroma (Fig. 2F, Middle and Bottom).
This specific accumulation pattern was not observed in the refer-
ence organs where both phage clones were confined to the vessels
(SI Appendix, Fig. S4). We suggest that the progressive accumu-
lation of the 011H12 phage in the tumor is due to the combination
of the interaction with a tumor-associated antigen with leaky tumor
vessels and long phage circulation time.
PA28 Is the Target of the 011H12 DAb. To identify the target of the
011H12 DAb, we subjected tumor extracts to affinity chroma-
tography. We used columns coupled with the anti−βGal DAb (SI
Appendix, Fig. S5) or the 011H12 DAb in tandem for the sepa-
ration. Affinity-purified proteins were digested by using the filter-
aided sample preparation (FASP) method and then identified by
Eluted proteins from the anti−βGal and 011H12 columns were
compared, and those present in both eluates (more than one
peptide in the control) were disregarded (SI Appendix, Table
S3). The remaining proteins were ranked according to the
number of peptides found in the eluate from the 011H12 col-
umn (SI Appendix, Table S4). To prioritize the potential can-
didates, the number of spectra associated with each protein
[peptide spectrum match (PSM) number] was also considered.
Notably, two of the selected proteins are subunits of the 11S
proteasome activator complex [PA28 (28, 29)]: subunit α [pro-
teasome activator complex subunit 1 (PSME1)] and subunit β
(PSME2). SDS/PAGE revealed two closely spaced bands at
about 29 kDa in the 011H12 DAb eluate that were not present
in the control DAb eluate (Fig. 3A). These two bands were
further identified as PSME1 (Q06323; 28.723 kDa) and PSME2
(Q9UL46; 27.402 kDa) by mass spectrometry (SI Appendix,
Table S5), in agreement with the results obtained by the FASP
method (SI Appendix, Table S4). The identity of the antigen
recognized by the 011H12 DAb was confirmed by ELISA. Sol-
uble purified 011H12 DAb showed reactivity with the active
PA28αβ complex (30) and to a lesser extent with individual
PA28 subunits (Fig. 3B). No reactivity was observed with two
lower score candidates (HPRT and NDKB) or βGal. Further-
more, 011H12 bound specifically to immobilized PA28αβ in
a concentration-dependent manner (Fig. 3 C and D).
have received the phage (either the reference α–βGal or 011H12) were ana-
lyzed by immunofluorescence, and the areas positive for CD31 (red) and the
phage (green) were quantified. No difference was found in the total area
positive for endothelial markers (A). There was a higher phage staining in the
011H12 clone (B). The V parameter (C, total phage/total endothelium) was six
times higher in the 011H12 clone (D). **Significant differences (P < 0.0001) U
Mann–Whitney. (E) Heterogeneous distribution of the phage. The 011H12
phage accumulates in the interior of the tumors, compared with the refer-
ence phage. In both cases, there is unspecific retention of the phage in the
rim of the tumor. (F) The distribution of the reference phage (α–βGal) is re-
stricted to the inner side of some vessels, as well as the edge of the tumor
(Top). The distribution of the 011H12 phage is heterogeneous: there are small
deposits in wide stromal areas (Middle) as well as major deposits around some
vessels identified as such by CD31 staining (Bottom). (Scale bar: 50 μm.)
Tumor-specific homing of 011H12 phage. Sections from tumors that
Sánchez-Martín et al.PNAS
| August 20, 2013
| vol. 110
| no. 34
Localization of PSME1 Expression in PPC-1 Human Prostate Cancer
Xenografts. To investigate the expression and subcellular localiza-
tion of PA28 subunits in tumors, we focused on PSME1 as a sur-
rogate for the complex. The localization of PSME1 was compared
with CD31 in sections of PPC-1 tumor xenografts. PSME1 was
heterogeneously distributed: most of the blood vessels showed
a positive staining, and there were also broad areas that were
negative for CD31, but PSME1 positive (SI Appendix, Fig. S6).
Tumors from mice injected i.v. with 011H12 phage showed areas
of colocalization of the antibody with PSME1 (Fig. 4). The
partial colocalization of the antibody phage and PSME1 is
compatible with PSME1 being the target of the 011H12 DAb.
Control phage remained confined to the blood vessels.
To study the accessibility of PSME1 expressed in prostate
tumors, a rabbit anti-PSME1 antibody and a rabbit control IgG
antibody were labeled with Cy7 and i.v. injected into PPC-1 tumor-
bearing mice (n = 3 per group). Both antibodies showed a very
early bladder signal spike upon injection, which could be attrib-
uted to a small percentage of trapped free Cy7 with rapid disap-
pearance (no detectable bladder signal at 24 h postinjection). In
both cases, a predominant liver signal was observed (SI Appendix,
Fig. S7A), a pattern of distribution expected when IgG rabbit
antibodies are used for in vivo imaging (31). The anti-PSME1
antibody localized in the tumors whereas the control IgG was not
detected in the tumors (SI Appendix, Fig. S7B). Maximum resolu-
tion was achieved at ∼24 h. Ex vivo imaging of the organs further
confirmed the specific accumulation of the anti-PSME1 antibody in
the tumor. Whereas the uptake by the kidney was similar for both
antibodies, the liver accumulated some more anti-PSME1 antibody
than control IgG (SI Appendix, Fig. S7 C and D). This accumulation
is likely due to differences in the IgG composition and the degree of
labeling, between anti-PSME1 and control IgG (31). In fact, the
expression of PSME1 in liver was confined to Kupffer cells (nuclear
pattern), absent in hepatocytes, and sporadically expressed in ductal
and endothelial cells (SI Appendix, Fig. S8), showing that the liver
uptake of anti-PSME1 was not antigen-specific.
Validation of PSME1 as Tumor Marker in Primary and Metastatic
Human Prostate Cancer. Validation of a candidate protein bio-
marker in human normal and tumor tissue samples is an essential
step in moving from the initial discovery to possible applications.
Therefore, we studied PSME1 expression in human prostate
sections by immunohistochemical staining on two tissue micro-
arrays. In normal prostate, the anti-PSME1 antibody staining
was restricted to basal cells, with faint cytoplasmic expression in
the luminal cells. Most blood vessels were positive, with occa-
sional reactivity in the surrounding stromal cells (Fig. 5 A and B).
In contrast, prostate cancer samples showed much higher PSME1
expression levels than the corresponding normal counterparts.
Remarkably, a strong staining was observed in cancerous epithe-
lium and in the stromal compartment, where extracellular PSME1-
positive deposits were evident. Blood vessels remained positive for
PSME1 staining (Fig. 5 B and C). We next investigated PSME1
expression in advanced prostate cancer with metastases to bone
and surrounding soft-tissue (Fig. 5E). Most of the tumor cells ex-
pressed PSME1. Endothelial cells and some stromal cells were also
highly positive for PSME1. The prostatic origin of metastases was
confirmed by PSA immunostaining (Fig. 5F).
Phage display is one of the preferred methods to obtain candidate
therapeutic human antibodies with one of the highest transition
rates between phase I and phase III studies (>65%) and with
good perspectives to raise the transition rate between phase III
and approval from 12.5% to near 30% (1). However, despite the
wide use of the phage display technology to obtain antibodies
against known and available targets (32), we believe that the
potential of this technology goes further and that it can also be
used to isolate antibodies against unknown but relevant targets.
This procedure could, at the same time, provide an antibody and
unveil a novel target (that could be either a truly unknown target
or an epitope not previously identified).
We hypothesized that using a repertoire to select antibodies
directly in vivo in a tumor mouse model could provide antibodies
able to effectively target the tumor in vivo. Because cells culti-
vated in vitro can modify the pattern of markers expressed on
their surface and because not all these markers would be acces-
sible from the blood vessels, antibodies selected against a purified
tumor lysates were purified using affinity chromatography with columns
coupled with the α–βGal and 011H12 DAbs. The eluates were subjected to
SDS/PAGE and stained using SYPRO dye, and the bands (arrows) were ana-
lyzed by mass spectrometry. (B) Purified soluble α–βGal or 011H12 DAbs (20
μg/mL) were tested against a panel of plastic-immobilized candidates (1 μg
per well). The interaction between 011H12 and the PA28αβ complex in vitro
is dose-dependent. Microtiter plates were coated (1 μg per well) with βGal
(C) or PA28αβ complex (D) and incubated with increasing concentration of
anti-βGal or 011H12 DAb. Data shown are means ± SD from triplicates and
are representative of three independent experiments.
Specific interaction of 011H12 DAb with PA28 subunits. (A) Fresh
PSME1 in PPC-1 tumor xenografts. PSME1 is heterogeneously distributed:
most of the blood vessels seem to be positive, and there are wide areas that
seem to be unrelated to CD31 structures (SI Appendix, Fig. S6). The reference
phage (α−βGal, first row; third row, magnification of the boxed area) pri-
marily localizes inside PSME1-positive vessel structures but is not seen
extravascularly. The 011H12 phage (second row; fourth row, magnification
of the boxed area) localizes inside PSME1-positive vessels (white arrowheads)
but is also present in extravascular tumor area where 011H12 colocalizes with
PSME1 (white arrows). (Scale bar: 50 μm.) Red, PSME1; green, phage.
Partial colocalization of in vivo administered 011H12 phage with
| www.pnas.org/cgi/doi/10.1073/pnas.1300013110Sánchez-Martín et al.
protein or an in vitro cultured cell may fail to access to the tumor
core effectively. The in vivo strategy, which is unbiased by any
prior knowledge of overexpressed markers, circumvents these
limitations as the only antibodies that can be selected are those
that effectively target the tumor. An ex vivo enrichment strategy,
before the in vivo selection, facilitates the selection for highly
diverse repertoires as the number of mice is drastically reduced.
The bias toward the more abundant tumor cells is compensated in
the ensuing in vivo selection that will target only the systemically
accessible antigens. This procedure has already been carried
out successfully using peptide phage display in tumor-bearing
mice and in other disease models (9, 33, 34); however, there are
very few examples of human antibodies selected with a similar
procedure, and always against vascular targets: thymic endothe-
lium (35) and atherosclerotic endothelial and subendothelial
In vivo peptide phage display was originally designed to reveal
features specific for the blood vessels of the tissue of interest (6).
This design was partially a consequence of the mechanism of
selection as the repertoire contacts mainly with the vascular wall
and the short circulation time intended not to allow time for the
extravasation of the phage particles. However, although target-
ing of the endothelium has repeatedly been reported as a tumor-
targeting strategy (4, 37–39), there is growing evidence that other
antigens (40–42) not necessarily restricted to the endothelium
may also be valuable targets. Nonetheless, in vivo phage display
still faces technical challenges such as the unspecific binding and
lack of enrichment. We have studied the dramatic effect of the
dosage and the circulation time in both challenges. The circulation
time is one of the key parameters in the in vivo selection, which,
combined with the size of the phage particles used and the fen-
estration size [highly heterogeneous in tumor endothelium, rang-
ing from 200 to 780 nm (43)], may influence the outcome of the
selection with a greater probability of discovering nonvascular
targets in tumors. Increasing the circulation time, although it may
compromise the infectivity of the bound phage if they are in-
ternalized or exposed to proteases, may allow identification of
antigens present in the extravascular compartment of solid tumors.
We explored several combinations of in vitro and ex vivo en-
richment before the in vivo selection (SI Appendix, Fig. S9) and,
based on extensive thorough sequencing, contrasted the perfor-
mance of a preenriched repertoire at different circulation times.
The results generally agree with previous work (35) for reference
organs although we focused only on viable phage (whereas in
previous work the quantification has been done by phage stain-
ing, which does not differentiate between infective and non-
infective particles; therefore the greater accumulation found in
others may be a consequence of the mononuclear phagocytic
system). However, by using a previously enriched library for the
sole purpose of establishing the appropriate circulation time, we
were able to detect differences in the target organ that favored
the circulation time of 24 h that would have been too subtle if
using a naïve library. The differences in half-life due to different
fusion proteins displayed (17–19) also suggest that the circula-
tion time is a parameter that should be carefully evaluated for
every experimental model.
The selection procedure outlined here facilitates the identifi-
cation of ligands not restricted to the endothelium. The tumor
distribution pattern of 011H12 clone is distinct from the vascular
staining pattern. Furthermore, a staining for an endothelial marker
and the phage not only can distinguish between an unspecific re-
tention of the phage inside the vessels and a specific retention of
the phage (that could show a different staining pattern), but
also is useful to quantify these differences. The total positive
area of tissue sections positive for CD31 immunoreactivity is
similar for the mice that received the α−βGal reference phage
and for those that received the 011H12 phage. The six times
higher accumulation of 011H12 phage (compared with α−βGal
phage) is therefore not a consequence of differences in tumor
We believe that the strategy described here addresses several
issues considered as crucial for the developing of new therapies
(44). It allows for the identification and validation of new tar-
gets, which could be related to the host–host microenvironment
interactions. Furthermore, the antibodies selected from this li-
brary share a high thermal stability and are less prone to ag-
gregation than scFv selected from other commonly used libraries
(45, 46). Moreover, they belong to the class of minimal binding
proteins, which are the preferred targeting elements for some
The 011H12 DAb was found to recognize the human PA28αβ
complex. We are still investigating the precise epitope. We hy-
pothesize that it recognizes a composite conformational epitope
formed by at least two PA28 subunits (PSME1 and PSME2)
based on the following considerations: the inability of 011H12
to detect the denatured antigens in Western blotting, the ability
to bind to PSME 1, the ability to bind—albeit less strongly—to
PSME 2 despite low sequence homology between these two
proteins, and the fact that the ELISA signal increased when the
PA28αβ complex was used as the target. PA28 is a ring-shaped
multimeric complex that binds to the two ends of the standard
(20S) proteasome and stimulates its capacity to hydrolyze small
peptides. In mammals, PA28 expression is modulated by IFNγ
(47), and it is a potent stimulator of MHC class I antigen pre-
sentation (48). However, it has been proposed that the involvement
of PA28 in antigen presentation may be only a secondary effect of
a physiological function (49). PSME1 has recently been proposed as
a potential tumor marker in human esophageal squamous cell
carcinoma, based on studies using proteomics approaches (50).
Moreover, the C terminus of the protein has been identified (51) as
human prostate tissues. (A) Panoramic view (15×) of normal prostate
showing moderate staining for PSME1. (B) Higher magnification (63×) of the
indicated region from A illustrating the staining of basal epithelial cells,
blood vessels, and surrounding stroma. (C) Low-magnification (15×) of pros-
tate carcinoma exhibiting a strong widespread PSME1 immunoreactivity. (D)
Higher magnification (63×) of the indicated region from C showing strong
staining in tumor cells, blood vessels, and stroma. (E) PSME1 and (F) PSA
immunostainings (magnification, 15×) in bone metastases with extension to
surrounding soft tissues. Insets show higher magnifications (×63) of the boxed
areas. (Scale bars: A, C, D, and E, 100 μm; B, D, and Insets, 20 μm.)
Immunohistochemical detection of PSME1 in normal and tumor
Sánchez-Martín et al.PNAS
| August 20, 2013
| vol. 110
| no. 34
a cytoplasmic marker (52). Our data indicate that PA28 can be
a tumor marker for human prostate cancer and that it is accessible
to targeting with i.v. injected anti-PSME1 antibodies, providing an
opportunity to specifically deliver therapeutic agents into primary
and metastatic prostate carcinoma. Antibody staining of tumor
xenograft sections outlined two main PSME1-positive areas within
the tumor (blood vessels and some areas in the perivascular region)
that could correspond to tissue disorganized as a result of cellular
activation and movement, or intermittent hypoxia and reperfusion.
We hypothesize that, in those areas, PA28 subunits may be local-
ized in compartments different from those they usually occupy,
becoming available for antibody targeting independently of varia-
tions in the level of expression.
The PA28 complex has been observed in the nucleus (mainly
PA28γ) and the cytoplasm [mainly PA28αβ (30, 53)]. How PA28
subunits might be targeted to the cell surface is unclear; however,
several other intracellular proteins are aberrantly expressed at
the cell surface in malignant tumors of various types (4, 42). It is
also possible that a fraction of the PA28 subunits are expressed
at the cell surface even in normal cells and that the overexpression
we demonstrated in prostate cancers makes this protein a selective
target in tumors. These possibilities remain to be fully explored,
but the results we report here strongly suggest the possibility that
PA28 subunits are useful biomarkers and potential drug targets
for prostate carcinoma.
Materials and Methods
DAb repertoires were purified by poly(ethylene glycol) precipitation (32). The
repertoire was circulated in vivo in tumor-bearing mice, and phage were re-
covered by trypsin release and infection (32). The target of the DAb was purified
in column from a tumor lysate and identified by mass spectrometry. A de-
scription of reagents, procedures, and protocols can be found in SI Appendix.
ACKNOWLEDGMENTS. We thank A. Gómez-Robles for advice, discussion,
and statistical guidance; M. J. Coronado for help with the confocal micro-
scope; and O. Coux (Centre de Recherche en Biochimie Macromoléculaire-
Centre National de la Recherche Scientifique, Universités Montpellier,
France) for providing the PA28αβ complexes. We thank the Centro Nacional
de Investigaciones Oncológicas Tumor Bank and the H. U. Marqués de Val-
decilla Tumor Bank (Santander, Spain) for kindly providing the cases in-
cluded in this study. This study was supported by Ministerio de Economía y
Competitividad Grant BIO2011-22738 and Comunidad de Madrid Grant
S2010/BMD-2312 (to L.A.-V.), and by US Department of Defense Prostate
Cancer Program Grant W81XWH-10-1-0199 (to E.R.). D.S.-M. was supported
by Comunidad de Madrid/Fondo Social Europeo Training Grant FPI-000531,
by a Fundación de Investigación del H. U. Puerta de Hierro travel grant, and by
a Boehringer Ingelheim Fonds travel grant. P.X.-E. was partially supported
by Instituto de Salud Carlos III Grant CA10/01231. V.A.-C. was supported by
Predoctoral Fellowship BFI07.132 from the Gobierno Vasco.
1. Nelson AL, Dhimolea E, Reichert JM (2010) Development trends for human mono-
clonal antibody therapeutics. Nat Rev Drug Discov 9(10):767–774.
2. Beck A, Wurch T, Bailly C, Corvaia N (2010) Strategies and challenges for the next
generation of therapeutic antibodies. Nat Rev Immunol 10(5):345–352.
3. St Croix B, et al. (2000) Genes expressed in human tumor endothelium. Science
4. Oh P, et al. (2004) Subtractive proteomic mapping of the endothelial surface in lung
and solid tumours for tissue-specific therapy. Nature 429(6992):629–635.
5. Rybak JN, et al. (2005) In vivo protein biotinylation for identification of organ-specific
antigens accessible from the vasculature. Nat Methods 2(4):291–298.
6. Pasqualini R, Ruoslahti E (1996) Organ targeting in vivo using phage display peptide
libraries. Nature 380(6572):364–366.
7. Laakkonen P, Zhang L, Ruoslahti E (2008) Peptide targeting of tumor lymph vessels.
Ann N Y Acad Sci 1131:37–43.
8. Lee SM, et al. (2007) Targeting bladder tumor cells in vivo and in the urine with
a peptide identified by phage display. Mol Cancer Res 5(1):11–19.
9. Zhang L, Giraudo E, Hoffman JA, Hanahan D, Ruoslahti E (2006) Lymphatic zip codes
in premalignant lesions and tumors. Cancer Res 66(11):5696–5706.
10. Laakkonen P, Porkka K, Hoffman JA, Ruoslahti E (2002) A tumor-homing peptide with
a targeting specificity related to lymphatic vessels. Nat Med 8(7):751–755.
11. Ruoslahti E (2004) Vascular zip codes in angiogenesis and metastasis. Biochem Soc
12. Alley SC, Okeley NM, Senter PD (2010) Antibody-drug conjugates: Targeted drug
delivery for cancer. Curr Opin Chem Biol 14(4):529–537.
13. Caraglia M, et al. (2012) Tumour-specific uptake of anti-cancer drugs: The future is
here. Curr Drug Metab 13(1):4–21.
14. Ruoslahti E, Bhatia SN, Sailor MJ (2010) Targeting of drugs and nanoparticles to tu-
mors. J Cell Biol 188(6):759–768.
15. Cuesta AM, Sainz-Pastor N, Bonet J, Oliva B, Alvarez-Vallina L (2010) Multivalent
antibodies: When design surpasses evolution. Trends Biotechnol 28(7):355–362.
16. Molek P, Strukelj B, Bratkovic T (2011) Peptide phage display as a tool for drug dis-
covery: Targeting membrane receptors. Molecules 16(1):857–887.
17. Newton JR, Kelly KA, Mahmood U, Weissleder R, Deutscher SL (2006) In vivo selection
of phage for the optical imaging of PC-3 human prostate carcinoma in mice. Neo-
18. Molenaar TJ, et al. (2002) Uptake and processing of modified bacteriophage M13 in
mice: Implications for phage display. Virology 293(1):182–191.
19. Zou J, Dickerson MT, Owen NK, Landon LA, Deutscher SL (2004) Biodistribution of
filamentous phage peptide libraries in mice. Mol Biol Rep 31(2):121–129.
20. Yip YL, Hawkins NJ, Smith G, Ward RL (1999) Biodistribution of filamentous phage-
Fab in nude mice. J Immunol Methods 225(1-2):171–178.
21. Teesalu T, Sugahara KN, Kotamraju VR, Ruoslahti E (2009) C-end rule peptides me-
diate neuropilin-1-dependent cell, vascular, and tissue penetration. Proc Natl Acad Sci
22. Pasqualini R, et al. (2000) Aminopeptidase N is a receptor for tumor-homing peptides
and a target for inhibiting angiogenesis. Cancer Res 60(3):722–727.
23. Koivunen E, Gay DA, Ruoslahti E (1993) Selection of peptides binding to the alpha
5 beta 1 integrin from phage display library. J Biol Chem 268(27):20205–20210.
24. Kolonin MG, et al. (2006) Synchronous selection of homing peptides for multiple
tissues by in vivo phage display. FASEB J 20(7):979-981.
25. Dias-Neto E, et al. (2009) Next-generation phage display: Integrating and comparing
available molecular tools to enable cost-effective high-throughput analysis. PLoS ONE
26. Wen Q, Tang JX (2006) Temperature effects on threshold counterion concentration to
induce aggregation of fd virus. Phys Rev Lett 97(4):048101.
27. Tang JX, Janmey PA, Lyubartsev A, Nordenskiöld L (2002) Metal ion-induced lateral
aggregation of filamentous viruses fd and M13. Biophys J 83(1):566–581.
28. McCusker D, Wilson M, Trowsdale J (1999) Organization of the genes encoding the
human proteasome activators PA28alpha and beta. Immunogenetics 49(5):438–445.
29. Knowlton JR, et al. (1997) Structure of the proteasome activator REGalpha (PA28al-
pha). Nature 390(6660):639–643.
30. Le Feuvre AY, Dantas-Barbosa C, Baldin V, Coux O (2009) High yield bacterial ex-
pression and purification of active recombinant PA28alphabeta complex. Protein Expr
31. Vasquez KO, Casavant C, Peterson JD (2011) Quantitative whole body biodistribution of
fluorescent-labeled agents by non-invasive tomographic imaging. PLoS ONE 6(6):e20594.
32. Sánchez-Martín D, Cuesta AM, Fogal V, Ruoslahti E, Alvarez-Vallina L (2011) The
multicompartmental p32/gClqR as a new target for antibody-based tumor targeting
strategies. J Biol Chem 286(7):5197–5203.
33. Staquicini FI, et al. (2011) Vascular ligand-receptor mapping by direct combinatorial
selection in cancer patients. Proc Natl Acad Sci USA 108(46):18637–18642.
34. Teesalu T, Sugahara KN, Ruoslahti E (2012) Mapping of vascular ZIP codes by phage
display. Methods Enzymol 503:35–56.
35. Johns M, George AJ, Ritter MA (2000) In vivo selection of sFv from phage display li-
braries. J Immunol Methods 239(1-2):137–151.
36. Deramchia K, et al. (2012) New human antibody fragments homing to atherosclerotic
endothelial and subendothelial tissues: An in vivo phage display targeting human
antibodies homing to atherosclerotic tissues. Am J Pathol 180(6):2576–2589.
37. Carmeliet P (2005) Angiogenesis in life, disease and medicine. Nature 438(7070):932–936.
38. Ferrara N, Kerbel RS (2005) Angiogenesis as a therapeutic target. Nature 438(7070):967–974.
39. Neri D, Bicknell R (2005) Tumour vascular targeting. Nat Rev Cancer 5(6):436–446.
40. Fogal V, Zhang L, Krajewski S, Ruoslahti E (2008) Mitochondrial/cell-surface protein p32/
gC1qR as a molecular target in tumor cells and tumor stroma. Cancer Res 68(17):7210–7218.
41. Kelly KA, et al. (2008) Targeted nanoparticles for imaging incipient pancreatic ductal
adenocarcinoma. PLoS Med 5(4):e85.
42. Christian S, et al. (2003) Nucleolin expressed at the cell surface is a marker of endo-
thelial cells in angiogenic blood vessels. J Cell Biol 163(4):871–878.
43. Hobbs SK, et al. (1998) Regulation of transport pathways in tumor vessels: Role of
tumor type and microenvironment. Proc Natl Acad Sci USA 95(8):4607–4612.
44. Weiner LM, Surana R, Wang S (2010) Monoclonal antibodies: Versatile platforms for
cancer immunotherapy. Nat Rev Immunol 10(5):317–327.
45. Dudgeon K, Famm K, Christ D (2009) Sequence determinants of protein aggregation
in human VH domains. Protein Eng Des Sel 22(3):217–220.
46. Famm K, Hansen L, Christ D, Winter G (2008) Thermodynamically stable aggregation-
resistant antibody domains through directed evolution. J Mol Biol 376(4):926–931.
47. Groettrup M, et al. (1996) A role for the proteasome regulator PA28alpha in antigen
presentation. Nature 381(6578):166–168.
48. Sijts A, et al. (2002) The role of the proteasome activator PA28 in MHC class I antigen
processing. Mol Immunol 39(3-4):165–169.
49. Sijts EJ, Kloetzel PM (2011) The role of the proteasome in the generation of MHC class
I ligands and immune responses. Cell Mol Life Sci 68(9):1491–1502.
50. Zhang J, et al. (2011) Using proteomic approach to identify tumor-associated proteins as
51. El Ayed M, et al. (2010) MALDI imaging mass spectrometry in ovarian cancer for
tracking, identifying, and validating biomarkers. Med Sci Monit 16(8):BR233–BR245.
52. Lemaire R, et al. (2007) Specific MALDI imaging and profiling for biomarker hunting
and validation: Fragment of the 11S proteasome activator complex, Reg alpha frag-
ment, is a new potential ovary cancer biomarker. J Proteome Res 6(11):4127–4134.
53. Brooks P, et al. (2000) Subcellular localization of proteasomes and their regulatory
complexes in mammalian cells. Biochem J 346(Pt 1):155–161.
| www.pnas.org/cgi/doi/10.1073/pnas.1300013110Sánchez-Martín et al.
Supporting Information Appendix
Sánchez-Martín et al. Proteasome activator complex PA28 identified as an accessible target in
prostate cancer by in vivo selection of human antibodies
Figure S1. Decision tree. A
database was constructed containing
the information of the CDR’s and
organ for each sequence retrieved.
Sequences were aligned using
ClustalW2 (1) and grouped and
visualized using Jalview (2). The
weight of each
calculated as the percentage of the
sequences in it. Each sequence
within a family was tested for a
parameters of number (if present) of
amber stop codons and the first
nucleotide after each; as well as
(number of cysteines, instability
index, etc.), if the sequence was not
valid, an alternative valid sequence
was searched inside the family [the
information was discarded if no
found]). All valid sequences were
comparison with the consensus
sequence inside their family and
with known motifs present in the
literature (RGD, CendR, NGR, etc.). The frequency of pre-selected families and sequences
was compared in the input, output in reference organs and output in tumor. Selected
individual sequences were further analyzed in competition assays.
Figure S2. Computational analysis
of the sequences.
Resampling-based evaluation assessing
organs) of the combination of three
residues and their
position within the CDR’s after an in
vivo selection of a pre-enriched
repertoire (round 0.2.0). The plot of
the experimental smallest p-values (red
line) determined by Fisher’s exact test
was compared with the corresponding
simulated datasets (light gray). Plotted
are the 50 smallest p-values (index
number of p-values 1 through 50,
ascending order) generated in each of
the 1000 permutations. (upper right)
For each of the 50 smallest p-values, a
new p-value (psim) was calculated.
Each psim represents the probability of
the experimental p-values to lay within
the simulated distribution, and they
were calculated as the proportion of
simulated p-values smaller than the
experimental p-values. The 50 psimwere
plotted in a box plot, with median
0.024. (mid) A similar analysis was
performed for the combination of all in
vivo rounds of
combination of residues and their
significantly (p-value < 0.05) different between the tumor and the combined organs let us calculate
different consensus sequences with preferential accumulation in the tumor (TC) and consensus
sequences absent from the tumor (OC). (lower) Heat map based on the evaluation of the selected
sequences (Table S1) a posteriori for common three residues with the consensus sequences.
Figure S3. Usefulness of the normalized value. Using as input either 1 × 1011 TU of
monoclonal phage (n = 5) or 1 × 109 TU (n = 2), the total recovery of infective particles is
highest for the highest dose, from all organs (a). However, the normalized value follows
approximately the same pattern (b), and can be even pooled into a new variable for the same
monoclonal phage (c).
Figure S4. Immunofluorescence in non-tumor organs. The reference phage (Gal) and the
011H12 phage shared a similar localization in non-tumor organs, mostly confined in vessel
structures in the kidney (a) and heart (c), or scattered throughout the liver (b) (red: CD31,
Figure S5. Purification of the antigen using immobilized antibodies. The anti-β-
galactosidase DAb was purified from E. coli supernatant using protein A (a, left panel; MW:
Molecular weight, BG1, BG2: eluted fractions, W: washing fraction, SN: raw supernatant)
and tested against the immobilized antigen (β-Galactosidase) in ELISA (a, right panel).
Different quantities of commercial β-Galactosidase were diluted in PBS (100 g in 1 ml PBS)
or in tumor lysate (or 40 g in tumor lysate) and purified using a column with immobilized
anti-β-Galactosidase DAb as a proof of concept. The recovered fractions were analyzed by
silver staining (b; Pre: sample, W: washing fraction, E: eluted fraction, MW: molecular
weight, 100: 100 ng of commercial -Galactosidase, 400: 400 ng of commercial β-
Galactosidase) and were tested in western blotting with an anti-β-Galactosidase polyclonal
antibody (c). It is possible to purify the antigen using this strategy, even if it is degraded (b, c).
Figure S6. Distribution of PSME1 and CD31 in PPC-1 tumor xenografts. There are two
distribution patterns of PSME-1 (red: CD31, green: PSME-1): either associated to
endothelium (white arrows) or in clusters in the surrounding perivascular tumor parenchyma
(white arrowheads). Scale bar is 50 µm.
Figure S7. Tumor homing of anti-PSME1 antibodies. Near-infrared fluorescence imaging
of nude mice bearing PPC-1 tumor xenografts. (a) Ventral view shows similar accumulation
of both rabbit polyclonal anti-PSME1 antibody and control rabbit IgG in the bladder. (b)
Control IgG did not localize in the tumor, whereas anti-PSME1 antibody showed a specific
signal. Ex vivo imaging (c) and quantification of fluorescence (d) in different organs (liver,
kidney, spleen) and tumor.
Figure S8. Immunohistochemical detection of PSME1 in mouse tissues. Representative
images of normal tissues sections (x10, left panels) stained with anti-PSME1. Higher
magnification (x40, right panels) view of the indicated region from left panel. Scale bars:
(left) 200 µm; (right) 50 µm.