Synthesis of a phosphoantigen prodrug that potently activates Vγ9Vδ2 T-lymphocytes

Article abstract of DOI:10.1016/j.chembiol.2014.06.006

Phosphoantigen-sensitive Vγ9Vδ2 T cells are important responders to infections and malignancy. However, the mechanisms by which phosphoantigens stimulate Vγ9Vδ2 T cells are unclear. Here, we synthesized phosphoantigen prodrugs and used them to demonstrate that intracellular delivery of phosphoantigens is required for their activity. The pivaloyloxymethyl prodrug is the most potent phosphoantigen described to date, with stronger stimulation of Vγ9Vδ2 T cells from human peripheral blood and greater ability to induce lysis of Daudi lymphoma cells relative to the previously most potent compound, (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP). We demonstrate high binding affinity between phosphoantigens and the intracellular region of butyrophilin 3A1 (BTN3A1), localized to the PRY/SPRY (B30.2) domain, but also affecting the membrane proximal region. Our findings promote a phosphoantigen prodrug approach for cancer immunotherapy and unravel fundamental aspects of the mechanisms of Vγ9Vδ2 T cell activation.

Full text of DOI:10.1016/j.chembiol.2014.06.006

Chemistry & Biology
Article
Synthesis of a Phosphoantigen Prodrug
that Potently Activates Vg9Vd2 T-Lymphocytes
1
1
2,3
2
1
1
Chia-Hung Christine Hsiao, Xiaochen Lin, Rocky J. Barney, Rebekah R. Shippy, Jin Li, Olga Vinogradova,
2
,4
1,
David F. Wiemer, and Andrew J. Wiemer *
Department of Pharmaceutical Sciences, University of Connecticut, Storrs, CT 06269, USA
2Department of Chemistry, University of Iowa, Iowa City, IA 52242, USA
1
3Department of Chemistry, Western Wyoming Community College, Rock Springs, WY 82901, USA
4Department of Pharmacology, University of Iowa, Iowa City, IA 52242, USA
*Correspondence: andrew.wiemer@uconn.edu
http://dx.doi.org/10.1016/j.chembiol.2014.06.006
SUMMARY
cells also can suppress viruses and malignancies (Gao et al.,
2003; Wang et al., 2006).
Phosphoantigens such as HMBPP are detected by the
Phosphoantigen-sensitive Vg9Vd2 T cells are impor-
tant responders to infections and malignancy. How- Vg9Vd2 T cell receptor after binding to butyrophilin 3A1
(
BTN3A1) (Harly et al., 2012; Wang et al., 2013) through major
ever, the mechanisms by which phosphoantigens
stimulate Vg9Vd2 T cells are unclear. Here, we
synthesized phosphoantigen prodrugs and used
them to demonstrate that intracellular delivery of
phosphoantigens is required for their activity. The
pivaloyloxymethyl prodrug is the most potent phos-
phoantigen described to date, with stronger stimula-
tion of Vg9Vd2 T cells from human peripheral blood
histocompatibility complex-independent mechanisms that are
not yet completely clear. Both intracellular and extracellular
binding of BTN3A1 to phosphoantigens have been recently re-
ported (Sandstrom et al., 2014; Vavassori et al., 2013). Once acti-
vated, Vg9Vd2 T cells eliminate phosphoantigen-loaded target
cells through killer cell lectin-like receptor subfamily K, member
1
recruit other immune cells via chemokine production (Li et al.,
receptor signaling and perforin release (Wrobel et al., 2007) and
and greater ability to induce lysis of Daudi lym- 2005, 2013). Therefore, development of compounds that
phoma cells relative to the previously most potent can bind to BTN3A1 in cancer cells to promote their lysis by
compound, (E)-4-hydroxy-3-methyl-but-2-enyl pyro- Vg9Vd2 T cells may be a promising anticancer strategy. How-
ever, questions remain about how to best target phosphoanti-
phosphate (HMBPP). We demonstrate high binding
affinity between phosphoantigens and the intracel-
lular region of butyrophilin 3A1 (BTN3A1), localized
to the PRY/SPRY (B30.2) domain, but also affecting
the membrane proximal region. Our findings promote
a phosphoantigen prodrug approach for cancer
immunotherapy and unravel fundamental aspects
of the mechanisms of Vg9Vd2 T cell activation.
gens to their cellular binding site.
Prodrug approaches increasingly are applied to compounds
employed in antiviral and antitumor therapy (Hecker and Erion,
2008). The majority of prodrug applications of phosph(on)ates
seek to facilitate passive diffusion through the cell membrane
by masking negative charge until the compound is within the
cell, eliminating the need for endocytic uptake. Because small,
highly charged molecules can exhibit low membrane perme-
ability (Kornberg et al., 1972), synthesis of protected compounds
such as pivaloyloxymethyl (POM) derivatives (e.g., Adefovir dipi-
voxil) (Cundy, 1999; De Clercq, 2004; Marcellin et al., 2003) has
been used to advance such molecules to clinical use. We hy-
INTRODUCTION
The cells of the immune system recognize diverse foreign- and pothesized that an analogous strategy could be utilized to deter-
self-antigens, generating responses that remove a variety of mine the role of cellular entry in phosphoantigen-induced T cell
foreign invaders with great specificity. As such, generation of activation. Upon diffusion into the cell, the protecting groups
drug-like molecules that can regulate immune cell function from a phosphoantigen prodrug are thought to be released
is of wide interest. The gd T cells, usually only comprising 1%– by cellular esterases, leaving an anionic phosphoantigen inside
1
0% of circulating T cells, have received attention recently the cells. If intracellular/cytoplasmic binding to BTN3A1 is
due to their function in cancer immunosurveillance (Bonneville required for phosphoantigen activity, one might expect this strat-
et al., 2010). Unlike the ab T cells that respond to peptides, a sub- egy to increase potency relative to the free phosphoantigen
set of gd T cells (Vg9Vd2 T cells) detects small phosphorous- anion, because the negatively charged phosphoantigen should
containing compounds (phosphoantigens) (Wiemer and Wiemer, not transverse the cell membrane as readily.
2
014). The Vg9Vd2 T cells most likely gained sensitivity to phos-
With this in mind, we developed a synthetic route to make
phoantigens to fight bacterial infections, given that the most POM-protected phosphoantigens and examined the activity
potent natural phosphoantigen, (E)-4-hydroxy-3-methyl-but-2- of these compounds in Vg9Vd2 T cell activation. Here, we
enyl diphosphate (HMBPP) is produced by bacteria, but not hu- demonstrate that a POM-protected phosphoantigen is a highly
mans (Morita et al., 2007). However, phosphoantigen-activated potent Vg9Vd2 T cell stimulus, which effectively delivers the
Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved 945

Chemistry & Biology
PA Prodrug
Figure 1. Chemical Structures of Phos-
phoantigens
Compounds evaluated in this study include
the control phosphoantigens (HMBPP, IPP, and
risedronate) and the phosphonate (1a), and its
protected counterparts (1b, 1c, and 1d).
yield. Unfortunately, use of compound 1d
to prepare compound 1b proved prob-
lematic. Because reaction at the hydroxyl
group proved to be competitive, a slightly
modified sequence was employed to
reach compound 1b. Phosphonate 4 first
phosphoantigen to the intracellular binding site of BTN3A1. The was converted to the bis-POM analog 6. While some of the
compound, while nontoxic on its own, induces potent killing of mono-POM mono-methyl ester 7 also was formed, this was
malignant cells by Vg9Vd2 T cells, which could ultimately lead not a significant complication because the two compounds
to a new class of cancer immunotherapy agents.
were readily separable by column chromatography. Reaction
of compound 6 with selenium dioxide gave the POM-protected
target 1b in modest yield. Compound 7 then was treated inde-
pendently with selenium dioxide to obtain the closely related
compound 1c. Reduction with borohydride was avoided to mini-
mize potential decomposition of the POM groups.
RESULTS
Synthesis of a Phosphoantigen and Its
Pivaloyloxymethyl-Protected Prodrug
Vg9Vd2 T cells can be activated by direct phosphoantigens such
as HMBPP, which can occur at low nanomolar concentrations, Pivaloyloxymethyl-Protected Phosphoantigens Potently
and indirect phosphoantigens such as the nitrogenous bis- Stimulate Human Peripheral Blood Vg9Vd2 T Cells
phosphonates, for which activation occurs at micromolar con- We evaluated compounds 1a and 1b for their ability to stimulate
centrations and is dependent upon their ability to increase the Vg9Vd2
T cell proliferation (Figure 3A). Peripheral blood
concentration of the direct phosphoantigen isopentenyl diphos- mononuclear cells (PBMCs) derived from healthy donors were
phate (IPP) (Sicard et al., 2005) (Figure 1). However, diphos- incubated with 1 mM of compound 1a or 1b for 3 days in the
phates such as HMBPP have poor pharmacokinetic properties, presence of interleukin 2 (IL-2). Cells were examined for the pro-
most notably rapid clearance from the plasma due to phospha- portion of cluster of differentiation 3 (CD3)+ Vg9Vd2 T cells in
tase-mediated degradation and low membrane permeability due live cultures 2 weeks post stimulation by flow cytometry. As
to the negative charges of the acidic oxygen atoms at physiolog- expected, treatment with HMBPP (58.5% ± 1.8%) significantly
ical pH (Sicard et al., 2005). Therefore, we chose to generate a expanded the population of Vg9Vd2 T cells as compared to
series of prodrug compounds based on the more stable phos- cells treated with IL-2 alone (14.6% ± 3.7%). Compound 1a
phonate substructure (i.e., compound 1a).
was active at the reported concentrations (Table 1) (Bo e¨ dec
We have previously developed synthetic routes to make et al., 2008). Interestingly, the POM prodrug (compound 1b)
POM-protected bisphosphonates, which dramatically enhance was much more potent (57.7% ± 9.3%, at 1 mM) than com-
cellular activity (Wiemer et al., 2008). To initiate the current syn- pound 1a and the activity approached the potency of HMBPP
thetic pathway, we prepared the key phosphonate salt (com- (Figure 3A). This is a key finding because it demonstrates a dra-
pound 1a) and the corresponding POM-modified compounds matic increase in activity through use of a phosphoantigen
1
b and 1c (Figure 2). Phosphonate 1a was chosen as our initial prodrug and supports an important role for intracellular delivery
target because it has good chemical stability (decreasing the of phosphoantigens.
possibility of phosphatase metabolism) and is a modest phos-
To examine further the activity of these compounds in stimu-
phoantigen (Bo e¨ dec et al., 2008). However, to secure an efficient lating human Vg9Vd2 T cells, compounds 1a (unprotected), 1b
preparation of the parent structure as both the disodium salt 1a (two POM groups), 1c (one POM group; one methyl group),
and as the POM-protected compounds 1b and 1c, we devel- and 1d (two methyl groups) along with the known phosphoanti-
oped the synthetic sequence shown in Figure 2. Thus after prep- gens HMBPP, IPP, and risedronate (Figure 3B and Figure S1
aration of prenyl bromide 3 from alcohol 2, this allylic halide was available online) were tested in a dose response manner. As
allowed to react with the anion of dimethyl methylphosphonate expected, HMBPP potently expanded Vg9Vd2 T cells when
to afford the phosphonate 4. Compound 4 was oxidized upon stimulated for 3 days. At all of the tested concentrations
treatment with selenium dioxide, and then the E alcohol 1d (0.001–10 mM), HMBPP significantly increased Vg9Vd2 T cell
was isolated after treatment with NaBH
4
to reduce any aldehyde, population (Figure 3B), where HMBPP at 0.01 mM reached the
using procedures we had applied earlier to preparation of ger- peak of Vg9Vd2 T cell expansion to 63.3% ± 7.0%, which was
anyl (Kim et al., 2004) and then prenyl (Maalouf et al., 2007) de- more than a 4-fold increase compared to IL-2 alone treatment
rivatives. Hydrolysis of compound 1d under standard conditions (14.6% ± 3.8%). The half maximal effective concentration
(
McKenna et al., 1977) gave the expected sodium salt 1a in good (EC50) of HMBPP in our hands was 0.00051 mM (Table 1).
946 Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved

Chemistry & Biology
PA Prodrug
Figure 2. Synthesis of a Bis-Pivaloyloxymethyl Phosphoantigen Prodrug with Potent Expansion of Vg9Vd2 T Cells from Human Peripheral
Blood
Synthetic pathways to compounds 1a–1d.
Much higher levels of IPP are needed to trigger an IPP-specific (63.3 ± 7.0%; EC50 = 0.00051 mM) (Figure 3B; Table 1). The
Vg9Vd2 T cell response (Sireci et al., 2001). As expected, EC50 of compound 1b for T cell expansion was ꢀ740x lower
1
,000 mM IPP significantly expanded the Vg9Vd2 T cell popula- than that of compound 1a, illustrating the vast improvement in
tion (Figure S2). The EC50 of IPP was 36 mM, approximately efficacy seen by intracellular phosphoantigen delivery.
1,000-fold less potent than HMBPP (Table 1). The nitrogen-
7
containing bisphosphonate risedronate indirectly activates Pivaloyloxymethyl-Containing Compound 1b Is a Direct-
Vg9Vd2 T cell through inhibition of farnesyl diphosphate syn- Acting Phosphoantigen
thase and intracellular IPP accumulation (Thompson and Risedronate is known to stimulate Vg9Vd2 T cell proliferation by
Rogers, 2004). Consistently, risedronate increased the Vg9Vd2 inhibition of farnesyl diphosphate (FPP) synthase in the mevalo-
T cell population in total PBMC cultures compared to treatment nate pathway, which results in accumulation of IPP. To determine
with IL-2 alone. Risedronate reached maximal stimulation at whether the stimulatory effect of compound 1b on Vg9Vd2 T cell
1
0 mM (Figure S2). Risedronate at 1,000 mM, however, resulted proliferation could be caused by inhibition of FPP synthase and/
in a significant reduction of the activity, most likely due to or other enzymes of the mevalonate pathway, expansion of
toxicity, similar to the effects observed with other nitrogen- Vg9Vd2 T cells by risedronate and compound 1b was compared
containing bisphosphonates at concentrations above 100 mM inthepresenceorabsenceoflovastatin,which inhibits3-hydroxy-
(
Miyagawa et al., 2001). To assess this toxicity, we measured 3-methylglutaryl-coenzyme A reductase upstream of FPP syn-
cell viability at the day 14 time point and calculated half maximal thase and prevents the synthesis of IPP. Treatment with 10 mM
inhibitory concentration (IC50) values for each compound lovastatin alone did not alter the proliferation of Vg9Vd2 T cells.
(Table 1). The selectivity indexes of HMBPP and risedronate In the presence of 10 mM lovastatin, the stimulatory effect of ri-
were 980 and 38, respectively. sedronate at 10 mM on Vg9Vd2 T cellproliferationwassignificantly
Consistent with a previous study (Bo e¨ dec et al., 2008), the so- reduced from 48.9 ± 20.9% to 10.6 ± 4.0%. However, the stimu-
dium salt 1a at 10 mM significantly increased the proportion of latory effect of compound 1b on Vg9Vd2 T cell proliferation was
Vg9Vd2 T cells in the presence of IL-2, with an EC50 of 4 mM (Fig- not significantly affected in the presence of lovastatin (Figure 3C).
ure 3B; Table 1). Because dimethyl phosphonates are metabol- The data demonstrate that prodrug 1b does not achieve its
ically stable (Serafinowska et al., 1995), the dimethyl ester was stimulatory effect by altering the mevalonate pathway, but rather
predicted to be inactive and indeed compound 1d did not functions by releasing a direct phosphoantigen.
show any stimulatory effect (Figure 3B; Table 1). Replacing one
of the methyl groups with a labile POM group decreased the Only Compound 1b Significantly Induces Vg9Vd2 T Cell
EC50 value to 0.5 mM (Table 1), and this compound (1c) at Proliferation after 2-hr Stimulation
1
6
mM was able to expand Vg9Vd2 T cell population to 51.1 ± Next, we assessed whether duration of the stimulation plays a
.7% compared to 18.4 ± 5.4% from treatment with the dimethyl role in the Vg9Vd2 T cell proliferation between different com-
ester 1d at 1 mM. Compound 1b, which contains two POM pounds. We hypothesized that enhanced uptake of compound
groups, was much more potent than compounds 1a, 1c, and 1b would require a shorter stimulation time for maximal effects.
1
d, providing an EC50 value of 0.0054 mM. Compound 1b at Cells were treated with 0.1 mM HMBPP, 0.1 mM compound 1b,
0
.01 mM significantly expanded the Vg9Vd2 T cell population 100 mM compound 1a, or 100 mM risedronate for either 2 hr or
compared to IL-2 alone treatment and at 0.1 mM reached the the standard 3 days, followed by IL-2 supplementation every
peak of stimulation up to 62.9 ± 1.6% of Vg9Vd2 T cells in 3 days until day 14. Flow cytometry analysis showed that all
PBMC cultures, which was similar to the peak of Vg9Vd2 T cell compounds significantly expanded the Vg9Vd2 T cell population
expansion with 0.01 mM of the potent stimulator, HMBPP when the cells were stimulated for 3 days. In contrast, only
Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved 947

Chemistry & Biology
PA Prodrug
Figure 3. Bis-POM Prodrug 1b Functions as a Direct Phosphoantigen with Elevated Potency Relative to Free Acid 1a and HMBPP
A) POM modification of a phosphoantigen increases cellular potency for activation of Vg9Vd2 T cells. Compounds 1a and 1b were used to stimulate proliferation
for 3 days, and then removed. Cells were allowed to proliferate for 11 more days followed by staining for CD3 and the Vg9Vd2 TCR.
(
(
(
B) Dose response curves for expansion of human peripheral blood Vg9Vd2 T cells by HMBPP and compounds 1a–1d following a 3-day stimulation.
C) Expansion of Vg9Vd2 T cells by compound 1b is not blocked by cotreatment with lovastatin. Cells were stimulated as above with indicated concentrations of
compound 1b or risedronate in the presence or absence of 10 mM lovastatin.
D) Cells were treated for either 2 hr or 3 days with HMBPP, risedronate, compound 1a, or 1b and expansion of Vg9Vd2 T cells was examined by flow cytometry.
(
All data in (B)–(D) are represented as mean ± SD, n = 3, *p < 0.05 by ANOVA with Tukey’s post-hoc analysis. See also Figure S1.
compound 1b was able to significantly stimulate Vg9Vd2 T cell binant proteins are well folded and remain in a monomeric state.
proliferation following a 2 hr-pulse treatment (Figure 3D). These Most of the amide resonance frequencies for PRY/SPRY domain
observations indicate that compound 1b is more effective in acti- overlapped with the peaks in the spectra for BFI (presented over-
vating the Vg9Vd2 T cell population compared to all of the other laid in Figure S4). This suggests the structure of PRY/SPRY
compounds tested, including HMBPP.
domain in isolated form and in the context of the full-length
protein remains the same and that the 68-residue membrane
proximal region does not interact with the PRY/SPRY domain
intramolecularly. We performed ITC to measure the interactions
between these three BTN3A1 constructs and HMBPP (Fig-
The Butyrophilin 3A1 SPRY Associated Domain/sp1A/
Ryanodine Receptor Domain, B30.2 Domain, Directly
Binds Phosphoantigens
It recently was reported that BTN3A1 is required for phosphoan- ure 4A). While no binding was observed with the extracellular
tigen detection by Vg9Vd2 T cells (Harly et al., 2012; Vavassori BTN3A1 IgV domain, both the intracellular PRY/SPRY (B30.2)
et al., 2013; Wang et al., 2013). However, the exact region of domain and the BFI domain showed strong binding with dissoci-
d
BTN3A1 where phosphoantigens bind is a matter of debate (Va- ation constants (K ) of 1.61 mM and 1.54 mM, respectively. The
vassori et al., 2013; Wang et al., 2013). Based on the 740x in- interaction of HMBPP with BTN3A1 was enthalpy-driven with
crease in activity of the bis-POM prodrug, we hypothesized an unfavorable entropy change. HMBPP binding to the full intra-
that intracellular binding occurs, which is in contrast to the model cellular domain (DH = À51.9 kJ/mol and TDS = À18.7 kJ/mol)
of extracellular binding to the immunoglobulin-like V (IgV) (Va- and the PRY/SPRY domain (DH = À49.4 kJ/mol and TDS =
vassori et al., 2013). To determine whether the BTN3A1-antigen À16.4 kJ/mol) shares very similar enthalpy and entropy terms,
complex forms intracellularly or extracellularly, we expressed the which reveals the binding site on BTN3A1 for phosphoantigens
1
15
recombinant extracellular IgV domain, the intracellular sp1A/rya- is enclosed in the PRY/SPRY domain (Table S1). H- N HSQC
nodine receptor domain/ SPRY associated domain (PRY/SPRY) titrations of HMBPP agreed with the ITC results. While there
(
(
B30.2) domain, and the butyrophilin 3A1 full intracellular domain were no chemical shift perturbations in the spectrum for the
BFI) and purified the soluble proteins. First, we verified the con- IgV domain upon HMBPP titration (Figure S5), the effects from
formations of the constructs by nuclear magnetic resonance HMBPP on the spectra of BFI and PRY/SPRY were very pro-
(
clear Single Quantum Correlation (HSQC) spectra, all the recom- turbations were observed in the spectra for both BFI (Figure 5A)
1 15
NMR), and, according to the nicely dispersed H- N-Heteronu- nounced. HMBPP-concentration-dependent chemical shift per-
948 Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved

Chemistry & Biology
PA Prodrug
Because compound 1b is expected to undergo metabolic
cleavage, resulting in accumulation of compound 1a, we exam-
ined the binding affinity between compound 1a and the BTN3A1
PRY/SPRY domain. Like IPP, the binding affinity of 1a was unde-
tectable in ITC when using compound 1a as the sole ligand, as
we had purposely chosen a weak phosphoantigen for the pro-
drug studies. However, as expected, compound 1a had a mark-
edly stronger effect on the HMBPP binding to the PRY/SPRY
domain relative to IPP (Figure 4B). Like IPP, compound 1a
dose-dependently decreased the favorable enthalpy of HMBPP
binding to BTN3A1. These effects also were measured at
Table 1. Activity of Compounds for Expansion of Vg9Vd2 T Cells
from Human Peripheral Blood Mononuclear Cells and Inhibition of
Cell Viability
Fold
Fold
difference
difference IC50 Selectivity
a
Compound EC50 (mM) from HMBPP from 1a
(mM) index
0.50 980
ND ND
HMBPP
IPP
0.00051
36
1.0
71,000
NA
NA
NA
1
Risedronate
1.6
3,100
7,800
10
61
38
1a
1b
1c
1d
4.0
ND ND
0.60 110
8.6 17
ND ND
0.0054
0.50
740
8
1a:BTN3A1 ratios of 0:1(DH of À51.9 kJ/mol), 10:1 (À49.1 kJ/
980
mol), and 100:1 (À28.5 kJ/mol). Compound 1a also dose-depen-
dently made more favorable the TDS term of HMBPP binding at
IPP:BTN3A1 ratios of 10:1 (À16.2 kJ/mol) and 100:1 (+5.19 kJ/
mol). These data fit well with the cellular data in that 1a is a
more potent phosphoantigen relative to IPP and would be
expected to have a stronger effect. Perhaps surprisingly, the
>10
>19,000
NA
NA, not applicable; ND, not determined.
a
Compound 1a
and the PRY/SPRY domain (Figure 5B) and reached saturation at entropy term became positive at the highest concentration of
an HMBPP:protein ratio of 3:1. Combining ITC and NMR data, 1a. This finding demonstrates that the prenyl alcohol and/or
we have demonstrated that HMBPP interacts with BTN3A1 the olefin position play a more important role in the phosphoan-
at its intracellular domain at 1:1 ratio and is likely capable of tigen-BTN3A1 interaction as compared to the distal phosphate
inducing significant conformational rearrangements in the intra- group. Thus, according to the ITC data, compound 1a may
cellular domain of BTN3A1.
induce conformational changes in the intracellular domain of
To determine whether binding was specific to HMBPP or a BTN3A1 similar to the ones induced by the stronger interaction
general property of phosphoantigens, we also tested whether with HMBPP and absent in the interface with IPP.
another potent phosphoantigen, (5-hydroxy-4-methylpent-3-
To confirm that IPP and compound 1a do have direct inter-
1
15
enyl)-phosphonooxyphosphinic acid (C-HDMAPP) could bind actions with the intracellular domain of BTN3A1, H- N HSQC
to the internal domain (Figure S2). C-HDMAPP is a phospho- titrations of IPP and compound 1a into BFI were performed at
nate analog of HBMPP (Bo e¨ dec et al., 2008). Interestingly, ligand:protein molar ratios of 0:1, 10:1, and 100:1. Both IPP
C-HDMAPP exhibited similar energetics relative to HMBPP (Figure S6A) and compound 1a (Figure S6B) showed concentra-
when binding to the PRY/SPRY domain (K
d
= 1.95 mM, DH = tion-dependent perturbations in the spectra. These two sets of
À47.1 kJ/mol, TDS = À14.5 kJ/mol). The data reveal that both titration data demonstrate that both IPP and compound 1a
HMBPP and C-HDMAPP bind with high affinity to the intracel- bind to BFI at a similar binding site with only a limited number
lular PRY/SPRY (B30.2) domain, but not the extracellular V of distinct differences in amide chemical shift perturbations be-
domain. C-HDMAPP has a slightly weaker interaction, which is tween the two. In addition, the spectrum of compound 1a, but
consistent with its activity in cellular assays (Bo e¨ dec et al., not IPP, shows several chemical shift perturbations identical to
2
008). Therefore, we have demonstrated that two different phos- the ones observed in the titration of HMBPP (marked by arrows
phoantigens bind strongly to the intracellular domain of BTN3A1. on Figure S6C). Thus our NMR and ITC results are mutually
Because IPP also acts as a direct phosphoantigen in cellular consistent and suggest that binding sites for IPP and compound
assays, we tested its ability to interact with BTN3A1. However, 1a on the PRY/SPRY domain are quite similar, although each
we were unable to measure directly the binding of IPP to compound is also involved in unique interactions with BTN3A1.
BTN3A1 by ITC. This is perhaps not surprising; given that the ac-
tivity in cellular assays is approximately 71,000-fold less for IPP Prodrug 1b Stimulates Lysis of Daudi Lymphoma Cells
relative to HMBPP (Table 1), making the projected dissociation by Vg9Vd2 T Cells
constant well outside of the detection limit of ITC. Therefore, To examine if compound 1b can induce T cell killing of malignant
we performed a competition ITC study to determine whether cells, we examined the extent of Vg9Vd2 T cell lysis of Daudi lym-
increasing concentrations of IPP would affect the binding of phoma cells. Daudi cells were labeled with a fluorescent cell
HMBPP to BTN3A1 (Figure S4). IPP dose-dependently membrane dye and mixed with purified Vg9Vd2 T cells at a ratio
decreased the favorable enthalpy of HMBPP binding from a of 5 to 1. Following coculture in the presence or absence of
DH of À51.9 kJ/mol to À47.7 kJ/mol and À38.7 kJ/mol at 0.1 mM test compounds, cells were stained with fluorescein iso-
IPP:BTN3A1 ratios of 10:1 and 100:1, respectively (Table S1). thiocyanate (FITC)-Annexin V antibody and the cell profiles
IPP also dose-dependently made more favorable the TDS term examined by flow cytometry. The Vg9Vd2 T cells in combination
of HMBPP binding, changing it from À18.7 kJ/mol to À12.7 kJ/ with compound 1b resulted in a significant increase in Annexin V
mol and À6.78 kJ/mol at IPP:BTN3A1 ratios of 10:1 and 100:1 positive Daudi cells in comparison to no phosphoantigen
respectively. Taken together, the data show that IPP binds to treatment (Figure 6A). Of the test compounds, only HMBPP
BTN3A1 with a much weaker affinity compared to HMBPP and and compound 1b resulted in significant increase in Annexin V
that both compounds compete for the same binding site within positive Daudi cells (Figure 6B), with compound 1b promoting
the PRY/SPRY domain.
greater T cell cytotoxic activity than HMBPP. In the absence of
Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved 949

Chemistry & Biology
PA Prodrug
Figure 4. Binding of HMBPP and Compound
1a to the Intracellular Domain of BTN3A1
(
A) Isothermal titration calorimetry plots for the
interaction of HMBPP with the extracellular IgV
domain (gray line/dots), the full intracellular domain
(
(
(
BFI) (black line/dots), or the isolated PRY/SPRY
B30.2) domain (red line/dots).
B) Isothermal titration calorimetry plots for the
interaction of the full intracellular domain (BFI) of
BTN3A1 with HMBPP in the absence of compound
1
a (red line/dots), or the presence of 10x com-
pound 1a (gray line/dots), or 100x compound 1a
blue line/dots).
(
See also Figures S2 and S3 and Table S1.
results demonstrate that phosphoanti-
gen neutralization dramatically decreases
the time necessary for loading phos-
phoantigen presenting cells and suggest
that charged phosphoantigens enter cells
primarily by active uptake rather than
passive diffusion. There is clearly an op-
portunity for further increases in potency,
T cells, no effect on the viability of Daudi cells was observed which could be generated either through utilizing stronger phos-
when the cells were treated with high concentrations of either phoantigens as templates or through utilizing different prodrug
compound 1a or 1b (Figure 6C) for up to 72 hr at 10 mM. In motifs.
contrast, cell viability was significantly reduced when treated
with vorinostat.
During the course of these studies, two groups reported bind-
ing of phosphoantigens to BTN3A1 via conflicting mechanisms
(
Sandstrom et al., 2014; Vavassori et al., 2013). Initially, De Lib-
DISCUSSION
ero and colleagues (Vavassori et al., 2013) demonstrated extra-
cellular binding, whereas a recent paper by Adams (Sandstrom
In this study we have described the synthesis and evaluation of a et al., 2014) clearly showed intracellular binding. Our studies
phosphoantigen prodrug. Surprisingly, this compound exhibits a are in strong agreement with those from the Adams Laboratory.
large increase in potency relative to the free phosphoantigen and In addition to the calorimetry assay, we also see strong inter-
displays similar, if not greater, potency relative to HMBPP. Addi- actions as assessed by NMR spectroscopy. Our results also
tionally, our analysis of the newly synthesized POM-modified demonstrate the relative contribution of the prenyl alcohol group
prodrug highlights several important conclusions (Figure 6D) and/or olefin position is more important than the second
about the mechanisms that underlie phosphoantigen presenta- phosphate group, because binding of HMBPP to BTN3A1 is
tion: (1) phosphoantigen charge serves as a barrier to both more strongly affected by 1a than IPP. Additionally, we have
cellular entry and exit, (2) phosphoantigens bind to the intracel- compared the effects of different antigen binding on the full intra-
lular, and not extracellular, domain of BTN3A1, and (3) effective cellular domain and the PRY/SPRY domain by NMR. Our results
delivery of phosphoantigens to the cytoplasm of malignant cells indicate that although the binding site for phosphoantigens is
induces their lysis by Vg9Vd2 T cells.
localized in PRY/SPRY domain, the 68-residue membrane prox-
At the start of our studies, the molecular target of phosphoan- imal region, preceding PRY/SPRY, also undergoes major struc-
tigens was not known, so we had deliberately chosen to combine tural rearrangement upon binding (Figure 5C). In the future, it will
a moderate strength phosphoantigen with a well characterized be interesting to determine the molecular details and the role of
prodrug to ensure that we could measure any potency increases the lipid bilayer in this process. While we are still in process of
generated by the prodrug strategy. To that end, it was quite sur- resonance assignments, which can be time consuming for a pro-
prising to see that compound 1b rivals the potency of HMBPP, tein of this size, our data are consistent with the recently pub-
the most potent known phosphoantigen. Additionally, relative lished crystal structure of the PRY/SPRY domain in complex
to the free phosphonic acid 1a, the pivaloyloxymethyl prodrug with HMBPP (Sandstrom et al., 2014). Our findings clearly pro-
1
b is approximately 740 times more potent for expansion of mote a model in which intracellular binding of phosphoantigens
Vg9Vd2 T cells from peripheral blood following a three day stim- to BTN3A1 induces a conformational change, which either
ulation (Table 1). The increased activity is even more prevalent in propagates across the membrane to the extracellular region or
assays using shorter incubation times (Figure 3D), such as the promotes an interaction with another intracellular protein, thus
2 hr PBMC stimulation or the 4 hr Daudi lysis assay (Figure 6B). allowing new transmembrane complex formation necessary for
In both of these assays, strong activity was seen with 0.1 mM recognition by the Vg9Vd2 T cell receptor.
of compound 1b, while the activity of the free phosphonic
Phosphoantigen prodrugs such as these may ultimately
acid 1a at this concentration was not measureable. These become clinically useful. Currently, there has been only one
950 Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved

Chemistry & Biology
PA Prodrug
Figure 6. Bis-POM Prodrug 1b Promotes Lysis of Daudi Lymphoma
Cells by Vg9Vd2 T Cells
(
A) DiD labeled Daudi cells were cocultured for 4 hr in the presence of Vg9Vd2
T cells at a 1:5 ratio and assessed by annexin V-FITC staining. Treatment with
compound 1b increases the number of annexin V positive Daudi cells.
(
B) The mixture was incubated with various test compounds at 0.1 mM. Data
1
15
Figure 5. H- N HSQC Titration of HMBPP into the Full-Length
Intracellular Domain and the PRY/SPRY Domain of BTN3A1
Compared to the PRY/SPRY domain, the full-length intracellular domain of
BTN3A1 (BFI) includes an additional membrane-proximal region of 68 residues
at the N terminus, which is demonstrated to be involved in the interaction with
HMBPP and to undergo significant structural changes upon binding.
are represented as mean ± SD, n = 3, *p < 0.05 by ANOVA with Tukey’s post-
hoc analysis.
(
C) 72 hr viability of Daudi cells incubated with compounds 1a and 1b relative
to vorinostat as a positive control. The mean of four replicates is presented,
p < 0.05 by ANOVA with Tukey’s post-hoc analysis.
D) proposed prodrug model illustrates pathways for POM-modified
*
(
A
(
A) Chemical shift perturbations for the interaction of HMBPP and BFI at ratios
of 0:1 (black), 1:1 (blue), and 3:1 (red).
B) Chemical shift perturbations for the interaction of HMBPP and the PRY/
SPRY domain at ratios of 0:1 (gray), 1:1 (orange), and 3:1 (green).
C) Evidence for the involvement of the 68-residue membrane-proximal region
compound internalized and presented by antigen presenting cells. Charged
molecules such as phosph(on)ates have restricted membrane diffusion. By
neutralizing charge with a cell cleavable protecting group, one can encourage
diffusion into cells, and upon intracellular release ion trapping results in
elevated intracellular levels.
(
(
in interaction with HMBPP: HSQC spectra of the free BFI (black), BFI + HMBPP
at 1:3 ratio (red), and PRY/SPRY + HMBPP at 1:3 ratio (green). Peaks that
belong to the 68-residue membrane-proximal region are absent in the spec-
trum of PRY/SPRY domain. Arrows indicate the peaks corresponding to the
amides from this region that have showed remarkable shifts upon binding to
HMBPP, thus suggesting significant structural rearrangements.
See also Figures S4–S6.
apy. This dosing regimen was likely a consequence of limited
plasma stability of BrHPP (a diphosphate), because studies in
primates demonstrated rapid degradation in the plasma (Sicard
et al., 2005). Our results clearly show a time dependency to the
uptake of phosphoantigens by peripheral blood cells, with no
direct phosphoantigen that has been tested in human clinical activation seen with HMBPP treatment of 2 hr (Figure 3D). This
trials, 4-bromo-3-hydroxy-3-methylbutyl diphosphate (BrHPP) explains the fact that a very high concentration of BrHPP was
(
Fourni e´ et al., 2013). In these trials, the initial stimulation led to needed to see any activation of Vg9Vd2 T cells in these
an expansion of the Vg9Vd2 population, but this required large patients. We have shown that the phosphoantigen prodrug 1b
intravenous doses and was limited in subsequent cycles of ther- requires a much shorter incubation time. Therefore, this or other
Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved 951

Chemistry & Biology
PA Prodrug
similar neutral compounds may be able to achieve Vg9Vd2 T cell
activation in human blood with significantly lower doses relative
to BrHPP.
ing flash column chromatography (3% MeOH in Et
2
O) to afford the target com-
1
pound in 65% yield (4.31 g) as a clear oil: H NMR d 5.07–5.00 (m, 1H), 3.67
(
d, JHP = 11.1 Hz, 6H), 2.26–2.14 (m, 2H), 1.75–1.63 (m, 2H), 1.62 (s, 3H),
1
3
1
.55 (s, 3H); C NMR d 132.9, 123.1 (d, JCP = 16.7 Hz), 52.2 (2C), 25.7, 25.0
31
d, JCP = 137.9 Hz), 21.1 (d, JCP = 4.8 Hz), 17.7; P NMR d 34.2; high resolution
Alternatively, others have used bisphosphonate drugs as
phosphoantigens in clinical trials (Van Beek et al., 2002); how-
ever, these compounds also have severely restricted phar-
macokinetics (e.g., rapid bone sequestration), low therapeutic
indexes, and the potential for side effects such as osteonecrosis.
Our data (Table 1) clearly show the advantage of using direct
phosphoantigens over an indirect phosphoantigen, as HMBPP
has a selectivity index that is 25 times greater than risedronate.
It is not clear whether other nitrogenous bisphosphonates,
such as zoledronate, have similarly narrow selectivity indexes.
However, we presume this to be the case as both the stimulation
and the cytotoxicity result from cellular inhibition of FPP syn-
thase, and at least one other group, has reported that shorter
durations of stimulation with zoledronate reduces its cytotoxic
effects (Nussbaumer et al., 2013). Compound 1b also has an
improved selectivity index relative to risedronate, and again we
believe it to be possible to make further gains in this area through
compound optimization.
(
mass spectrometry (HRMS) (time-of-flight [TOF] MS electron impact [EI]) m/z
+
calculated for C H O P (M+H) 193.0994, found 193.0973.
8
18
3
Compound 1d
To a vigorously stirred solution of compound 4 (2.03 g, 10.6 mmol) in CH
at room temperature were added SeO (0.602 g, 5.4 mmol), 4-hydroxy-
benzoic acid (0.150 g, 1.1 mmol), and t-BuOOH (70% in H O, 5.48 ml,
2.6 mmol). The solution was allowed to stir for 19 hr and a saturated aqueous
solution of NaHCO was added. The solution was extracted (CH Cl ), and the
combined extracts were dried (MgSO
2
Cl
2
2
2
4
3
2
2
4
) and concentrated in vacuo. The crude
ꢁ
material was dissolved in MeOH with stirring, cooled to 0 C in an ice bath,
and NaBH
was allowed to stir for 2.5 hr, saturated aqueous NH
tion was extracted (Et O), and the combined extracts were dried (MgSO
concentrated in vacuo. Final purification by flash column chromatography
4
(0.820 g, 21.6 mmol) was added in several aliquots. The solution
Cl was added, the solu-
) and
4
2
4
(
2
6%–10% MeOH in Et O) afforded the target compound in 36% yield
1
0.781 g) as a clear oil: H NMR d 5.37–5.31 (m, 1H), 3.89 (s, 2H), 3.66
(
(d, J = 11.7 Hz, 6H), 2.32–2.19 (m, 2H), 1.79–1.68 (m, 2H), 1.58 (s, 3H);
HP
1
3
C NMR d 136.3, 122.8 (d, JCP = 15.7 Hz), 67.5, 52.1 (d, JCP = 6.9 Hz, 2C),
3
1
2
4.3 (d, JCP = 138.6 Hz), 20.2 (d, JCP = 4.6 Hz), 13.3; P NMR d 34.3;
+
HRMS (TOF MS EI) m/z calculated for C
31.0745.
Compound 1a
8 17 4
H O PNa (M+Na) 231.0762, found
Finally, it is important to note the functional consequences of
the phosphoantigen prodrug in the inducible lysis of cancer cells
2
(
Figure 6). Compound 1b causes no discernable growth inhibi-
2 2
Compound 1d (0.166 g, 0.80 mmol) was dissolved in CH Cl , and the mixture
tion of Daudi cells at concentrations up to 10 mM and treatment
times up to 72 hr. However, when combined with Vg9Vd2 T cells,
compound 1b can induce the lysis of Daudi cells at 0.1 mM in
only 4 hr. Therefore, we believe that compounds such as 1b,
if effectively delivered to malignant cells, have the potential for
strong chemically induced cancer immunotherapy through re-
cruiting a targeted Vg9Vd2 T cell response.
ꢁ
was cooled to 0 C in an ice bath. Collidine (0.425 ml, 3.2 mmol) and bromotri-
methylsilane (0.419 ml, 3.2 mmol) were added via syringe, and the reaction
was allowed to stir overnight. After the volatiles were removed, toluene was
added and removed in vacuo, and this cycle was repeated three times
providing a white solid. The solid was dissolved in aqueous NaOH (1.49 M,
2
.15 ml, 3.2 mmol), and the reaction mixture was stirred overnight. The mixture
ꢁ
was poured into an equal volume of acetone and cooled at 7 C for a period of
2
2
4 hr to facilitate precipitation. Acetone and H O were removed in vacuo,
In conclusion, these findings reveal several key mechanisms
of antigen recognition in the immune system and point to new
possibilities for development of clinical agents through under-
standing phosphoantigen structure activity relationships at the
molecular and cellular levels.
the resulting solid was washed with acetone, dissolved in H O, and washed
2
with diethyl ether. The aqueous phase was decolorized using activated
charcoal and concentrated in vacuo to provide the target salt in 77% yield
1
137 mg): H NMR d 5.52–5.45 (m, 1H), 3.94 (broad singlet [br s], 2H), 2.23–
(
1
.11 (m, 2H), 1.64 (br s, 3H), 1.45–1.33 (m, 2H); C NMR d 134.0, 128.6
3
2
31
d, JCP = 17.3 Hz), 67.9, 29.5 (d, JCP = 126.9 Hz), 22.92, 13.5; P NMR
(
-
d 22.3; HRMS (TOF MS EI) m/z calculated for C
found 179.0460.
6 12 4
H O P (M-H) 179.0473,
SIGNIFICANCE
Compounds 6 and 7
Phosphonate 4 (0.967 g, 5.0 mmol), sodium iodide (1.56 g, 10.4 mmol), and
pivaloyloxymethyl chloride (2.3 ml, 15.9 mmol) were added to acetonitrile un-
der an atmosphere of argon. The mixture was heated and held at reflux with
vigorous stirring overnight. Water and diethyl ether were added, and the
phases were separated. The aqueous phase was extracted with diethyl ether,
We have described a phosphoantigen prodrug that func-
tions with exquisite potency. The prodrug defines key roles
of phosphoantigen charge, binding, and delivery. This strat-
egy has potential to impact clinical use of phosphoantigens,
which have been limited by poor pharmacokinetics, for
treatment or prevention of cancer and infectious disease.
and the combined organic extracts were dried (Na
vacuo. Final purification by flash column chromatography (30% Et
anes to 50% Et O in hexanes) gave the target compound in 20% yield
2
SO
4
) and concentrated in
2
O in hex-
2
1
(
391 mg) as a clear oil: H NMR d 5.60 (d, JHP = 13.2 Hz, 4H), 5.04–4.97
EXPERIMENTAL PROCEDURES
(
m, 1H), 2.26–2.14 (m, 2H), 1.84–1.71 (m, 2H), 1.61 (s, 3H), 1.53 (s, 3H), 1.16
1
3
(
s, 18H); C NMR d 176.6 (2C), 133.0, 122.2 (d, JCP = 17.9 Hz), 81.1
d, JCP = 6.5 Hz, 2C), 38.6 (2C), 26.6 (d, JCP = 136.8 Hz), 26.7 (6C), 25.4,
Chemical Synthesis
(
Compound 4
3
1
2
0.5 (d, JCP = 4.6 Hz), 17.5; P NMR d 32.7; HRMS (TOF MS EI) m/z calculated
To a stirred solution of n-BuLi (2.4 M in hexanes, 14.6 ml, 35.0 mmol) in tetra-
ꢁ
+
for C18H33O PNa (M+Na) 415.1862, found 415.1868.
hydrofuran at À78 C was added N,N-diisopropylamine (4.82 ml, 34.4 mmol).
7
The resulting solution was allowed to stir for 10 min and dimethyl methyl-
phosphonate (3.7 ml, 34.1 mmol) was added slowly via syringe. Once addition
was complete, the mixture was allowed to stir for 15 min during which time a
white precipitate formed. Prenyl bromide 3 (5.17 g, 34.7 mmol) was added
slowly via syringe, and after addition was complete the solution was allowed
Compound 6 was accompanied by compound 7 and isolated in 30% yield
1
(0.46 g) as a clear oil: H NMR (300 MHz, CDCl
3
) d 5.67 (d, JHP = 13.4 Hz,
2H), 5.14–5.05 (m, 1H), 3.75 (d, JHP = 11.1 Hz, 3H), 2.36–2.20 (m, 2H), 1.91–
1
3
1.74 (m, 2H), 1.68 (s, 3H), 1.62 (s, 3H), 1.24 (s, 9H); C NMR (75 MHz,
CDCl ) d 176.6 (s), 134.7 (d, JCP = 1.8 Hz), 122.4 (d, JCP = 17.4 Hz), 81.4
(d, JCP = 5.9 Hz), 51.5 (d, JCP = 7.0 Hz), 38.4 (s), 26.6 (s, 3C), 25.3 (s), 25.0
3
ꢁ
to stir for an additional 30 min at –78 C, allowed to warm to room temperature
3
1
unassisted, and stirred overnight. A saturated aqueous solution of NH
added, the solution was extracted with Et O, and the combined extracts were
dried (Na SO ) and concentrated in vacuo. Final purification was achieved us-
4
Cl was
(d, JCP = 35.1 Hz), 20.2 (d, JCP = 4.8 Hz), 17.3 (s); P NMR (121 MHz,
+
+
CDCl
3 13 25 5
) 33.7; HRMS (ESI ) m/z calculated for C H O PNa (M+Na)
2
315.1337, found: 315.1337.
2
4
952 Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved

Chemistry & Biology
PA Prodrug
Compound 1b
60 mM. The concentration of protein was examined after each ITC experi-
ment and a good solubility (>85%) was shown. Data analysis was done in
NanoAnalyze Software suite using an ‘‘independent’’ model. In all cases, a
stoichiometry of 1 ± 0.1 was revealed.
2 2
Compound 6 (0.361 g, 0.92 mmol) was dissolved in CH Cl followed by addi-
tion of SeO
and 4-hydroxybenzoic acid (0.013 g, 0.094 mmol). The mixture was stirred
vigorously overnight and then an aqueous solution of saturated NaHCO
was added. The mixture was extracted with CH Cl , the combined organic ex-
tracts were dried (Na SO ), and concentrated in vacuo. Final purification by
flash column chromatography (10% hexanes in Et O) afforded the target prod-
uct in 45% yield (168 mg) as a clear oil: H NMR d 5.60 (d, JHP = 12.9 Hz, 4H),
.34–5.28 (m, 1H), 3.91 (br s, 2H), 2.34–2.19 (m, 2H), 1.89–1.77 (m, 2H), 1.58
2
(0.05 g, 0.45 mmol), t-BuOOH (70% aqueous, 0.47 ml, 3.65 mmol),
3
2
2
Nuclear Magnetic Resonance Spectroscopy
1
5
2
4
N-labeled recombinant proteins were overexpressed in M9 medium supple-
1
5
2
mented with NH
samples of 100–200 mM were prepared in a buffer of 50 mM Tris, pH 7.5,
100mMNaCl,5mMBME,7%D O,1mM4,4-dimethyl-4-silapentane-1-sulfonic
4
Cl as the sole nitrogen source and purified as above. NMR
1
5
2
1
3
1
15
(s, 3H), 1.17 (s, 18H); C NMR d 176.9 (2C), 136.6, 122.7 (d, JCP = 15.7 Hz),
acidasthe internalstandard. H- N HSQCtitrationexperimentswere performed
on Varian Inova 600 MHz spectrometer equipped with inverse-triple resonance
81.2 (d, JCP = 6.2 Hz, 2C), 67.9, 38.6 (2C), 26.4 (d, JCP = 138.2 Hz), 26.7
31
ꢁ
(6C), 20.2 (d, JCP = 5.2 Hz), 13.5; P NMR 32.6; HRMS (TOF MS EI) m/z calcu-
cyroprobe at 25 C. All spectra were processed with NMRPipe (Delaglio et al.,
+
lated for C18
Compound 1c
Compound 7 (0.45 g, 1.54 mmol), SeO
H
34
O
8
P (M+H) 409.1992, found 409.1995.
1995) and analyzed in CcpNmr Analysis (Vranken et al., 2005) software suite.
2
(0.09 g, 0.77 mmol), 4-hydroxyben-
Daudi Killing Assay
zonic acid (0.02 g, 0.15 mmol), and t-BuOOH (0.79 ml, 6.15 mmol) were
dissolved in dichloromethane (10 ml), and the reaction was allowed to stir over-
night. The solution was quenched by addition of water and extracted with
diethyl ether. The organic portions were combined and washed with brine
The Daudi killing assay was performed based on (Gomes et al., 2010) with slight
modification. Vg9Vd2 T cells were expanded from PMBCs by stimulation with
0.01 mM HMBPP for 3 days and expansion for 8À11 additional days as above.
At that time, Vg9Vd2 T cells were purified by negative selection using a kit.
Resulting populations were typically 98%À99% Vg9Vd2 T cells as assessed
by flow cytometry. Daudi cells were stained with DiD (2 min in 4 mM DiD in
BSA/PBS), quenched by addition of an equal volume of FBS, and washed twice
and aqueous sodium thiosulfate, the solution was dried (MgSO
4
), filtered,
and concentrated in vacuo. The residue was purified via flash chromatography
(
ether), and the desired product 1c was isolated as a yellow oil in 32% yield
1
0.15 g, 43% based on recovered starting material): H NMR (300 MHz,
3
4
(
in T cell media. Daudi cells (6 3 10 ) were mixed with 3 3 10 T cells and test
compounds to a final volume of 100 ml. Mixtures were incubated for 4 hr at
37 degrees then placed on ice for 5 min. Annexin V FITC (3 ml) was added
for 15 min on ice, then cells were diluted by addition of 200 ml binding buffer
(Becton Dickinson) and immediately analyzed by flow cytometry.
CDCl
3
) 5.71–5.61 (m, 2H), 5.44–5.34 (m, 1H), 3.98 (s, 2H), 3.75 (d, JHP =
1
1.3 Hz, 3H), 2.85 (br s, 1H), 2.38–2.28 (m, 2H), 1.92–1.81 (m, 2H), 1.66
13
(s, 3H), 1.24 (s, 9H); C NMR (75 MHz, CDCl
3
) d 177.0 (s), 136.4 (d, JCP
=
=
1
.4 Hz), 122.9 (d, JCP = 16.2 Hz), 81.5 (d, JCP = 6.3 Hz), 67.8 (s), 51.8
(d, JCP = 7.2 Hz), 38.6, 26.8 (s, 3C), 26.6 (d, JCP = 20.3 Hz), 20.3 (d, JCP
31
+
4
.8 Hz), 13.5 (s); P NMR (121 MHz, CDCl
3
) 33.5; HRMS (ESI ) m/z calculated
Proliferation Assay
+
PNa (M+Na) 331.1286, found: 331.1291.
for C13
H
25
O
6
Daudi proliferation was performed as described with the following parameters
(
Wiemer et al., 2013). Cells/well (10,000) were added to 96-well plates in 100 ml
Vg9Vd2 T Cell Isolation and Culture
in the presence of test compounds. Cells were allowed to proliferate for 72 hr.
During the last 2 hr, cells were labeled with 5 ml of CellQuantiBlue reagent and
were scanned with a Victor plate reader.
Human peripheral blood mononuclear cell (PMBCs) were isolated from blood
from Research Blood Components using density centrifugation. Cells were
frozen in freezing media (10% DMSO, 90% fetal bovine serum [FBS]) in liquid
6
nitrogen until needed. 2À4 3 10 cells were resuspended in fresh T cell media
Statistical Analysis
(RPMI-1640, 10% heat-inactivated FBS, 1x 4-(2-hydroxyethyl)-1-piperazinee-
One-way ANOVA was used to calculate significance. Comparisons were done
relative to the control or between pairs of conditions. Columns in bar graphs
represent the mean ± SD. An a level of 0.05 was used.
thanesulfonic acid, pyruvate, nonessential amino acids, b-mercaptoethanol
BME]) and added to 6-well plates. Cells were stimulated for 3 days unless
[
stated otherwise. Cells were cultured for another 11 days after compound
removal. Human IL-2 (5 ng/ml) was supplemented every 3 days. All experi-
ments were performed at least three times independently using at least two
different blood donors.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Supplemental Experimental Procedures,
six figures, and one table and can be found with article online at http://dx.
doi.org/10.1016/j.chembiol.2014.06.006.
Flow Cytometry
Cells were washed and suspended in 100 ml of fluorescence-activated cell
sorting buffer (2% BSA in PBS). Total gamma delta T cell receptor (TCR)
was labeled using either FITC-anti-human Pan gamma delta TCR antibody
or anti-human Pan gamma delta TCR antibody followed by peridinin chloro-
phyll-eFluor 710-anti-mouse immunoglobulin G1. Cells were costained with
PE-conjugated anti-human CD3. Each step was done at 4 degrees for
AUTHOR CONTRIBUTIONS
C-H.C.H. and A.J.W. performed molecular biology and cell-based experi-
ments. R.J.B. and R.R.S. performed the chemical synthesis under supervision
of D.F.W. Protein purification, ITC, and HSQC NMR were performed by X.L.
under supervision of O.V. Cell viability experiments were performed by J.L.
The manuscript was written by C-H.C.H. and A.J.W. with the support of X.L,
O.V., and D.F.W.
30 min, washed twice, and then fixed in 3% paraformaldehyde before flow
analysis. Data were analyzed using FlowJo.
Cloning and Purification of Butyrophilin 3A1 Constructs
BTN3A1 constructs were purified based on the method of De Libero and col-
leagues (Vavassori et al., 2013) with some modifications (see Supplemental
Experimental Procedures).
ACKNOWLEDGMENTS
Financial support from the University of Connecticut Department of Pharma-
ceutical Sciences (A.J.W.) and from the Roy J. Carver Charitable Trust as a
Research Program of Excellence (D.F.W.) is gratefully acknowledged. A.J.W.
and D.F.W. are cofounders of Terpenoid Therapeutics, and D.F.W. is on its
Board. The current work did not involve the company.
Isothermal Titration Calorimetry
Calorimetric measurements of BTN3A1 domains were performed on a low vol-
ume Nano ITC. All experiments were carried out in a buffer of 50 mM Tris, pH
ꢁ
7.5, 100 mM NaCl, 5 mM BME at 25 C with a stirring speed of 200 rpm to mini-
mize the precipitation. Each titration consisted of twenty 2.5 ml injections with
00 s time intervals. The concentrations of protein samples were determined
Received: April 24, 2014
Revised: June 12, 2014
Accepted: June 18, 2014
Published: July 24, 2014
3
by UV, and the compounds were prepared in the same buffer from the dilution
of 10 mM or 100 mM stock. The protein concentration ranged from 30 mM to
Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved 953

Chemistry & Biology
PA Prodrug
REFERENCES
Morita, C.T., Jin, C., Sarikonda, G., and Wang, H. (2007). Nonpeptide anti-
gens, presentation mechanisms, and immunological memory of human
Vgamma2Vdelta2 T cells: discriminating friend from foe through the recogni-
tion of prenyl pyrophosphate antigens. Immunol. Rev. 215, 59–76.
Bo e¨ dec, A., Sicard, H., Dessolin, J., Herbette, G., Ingoure, S., Raymond, C.,
Belmant, C., and Kraus, J.L. (2008). Synthesis and biological activity of phos-
phonate analogues and geometric isomers of the highly potent phosphoanti-
gen (E)-1-hydroxy-2-methylbut-2-enyl 4-diphosphate. J. Med. Chem. 51,
Nussbaumer, O., Gruenbacher, G., Gander, H., Komuczki, J., Rahm, A., and
Thurnher, M. (2013). Essential requirements of zoledronate-induced cytokine
and gd T cell proliferative responses. J. Immunol. 191, 1346–1355.
1747–1754.
Bonneville, M., O’Brien, R.L., and Born, W.K. (2010). Gammadelta T cell
effector functions: a blend of innate programming and acquired plasticity.
Nat. Rev. Immunol. 10, 467–478.
Sandstrom, A., Peign e´ , C.M., L e´ ger, A., Crooks, J.E., Konczak, F., Gesnel,
M.C., Breathnach, R., Bonneville, M., Scotet, E., and Adams, E.J. (2014).
The intracellular B30.2 domain of butyrophilin 3A1 binds phosphoantigens to
mediate activation of human Vg9Vd2 T cells. Immunity 40, 490–500.
Cundy, K.C. (1999). Clinical pharmacokinetics of the antiviral nucleotide ana-
logues cidofovir and adefovir. Clin. Pharmacokinet. 36, 127–143.
Serafinowska, H.T., Ashton, R.J., Bailey, S., Harnden, M.R., Jackson, S.M.,
and Sutton, D. (1995). Synthesis and in vivo evaluation of prodrugs of 9-
[2-(phosphonomethoxy)ethoxy]adenine. J. Med. Chem. 38, 1372–1379.
De Clercq, E. (2004). Antiviral drugs in current clinical use. J. Clin. Virol. 30,
115–133.
Delaglio, F., Grzesiek, S., Vuister, G.W., Zhu, G., Pfeifer, J., and Bax, A. (1995).
NMRPipe: a multidimensional spectral processing system based on UNIX
pipes. J. Biomol. NMR 6, 277–293.
Sicard, H., Ingoure, S., Luciani, B., Serraz, C., Fourni e´ , J.J., Bonneville, M.,
Tiollier, J., and Romagn e´ , F. (2005). In vivo immunomanipulation of V gamma
9V delta 2 T cells with a synthetic phosphoantigen in a preclinical nonhuman
primate model. J. Immunol. 175, 5471–5480.
Fourni e´ , J.J., Sicard, H., Poupot, M., Bezombes, C., Blanc, A., Romagn e´ , F.,
Ysebaert, L., and Laurent, G. (2013). What lessons can be learned from gd
T cell-based cancer immunotherapy trials? Cell. Mol. Immunol. 10, 35–41.
Sireci, G., Espinosa, E., Di Sano, C., Dieli, F., Fournie
´
, J.J., and Salerno, A.
(2001). Differential activation of human gammadelta cells by nonpeptide phos-
phoantigens. Eur. J. Immunol. 31, 1628–1635.
Gao, Y., Yang, W., Pan, M., Scully, E., Girardi, M., Augenlicht, L.H., Craft, J.,
and Yin, Z. (2003). Gamma delta T cells provide an early source of interferon
gamma in tumor immunity. J. Exp. Med. 198, 433–442.
Thompson, K., and Rogers, M.J. (2004). Statins prevent bisphosphonate-
induced gamma,delta-T-cell proliferation and activation in vitro. J. Bone
Miner. Res. 19, 278–288.
Gomes, A.Q., Correia, D.V., Grosso, A.R., Lan c¸ a, T., Ferreira, C., Lacerda, J.F.,
Barata, J.T., Silva, M.G., and Silva-Santos, B. (2010). Identification of a panel
of ten cell surface protein antigens associated with immunotargeting of leuke-
mias and lymphomas by peripheral blood gammadelta T cells. Haematologica
Van Beek, E.R., L o¨ wik, C.W., and Papapoulos, S.E. (2002). Bisphosphonates
suppress bone resorption by a direct effect on early osteoclast precursors
without affecting the osteoclastogenic capacity of osteogenic cells: the role
of protein geranylgeranylation in the action of nitrogen-containing bisphosph-
onates on osteoclast precursors. Bone 30, 64–70.
95, 1397–1404.
Harly, C., Guillaume, Y., Nedellec, S., Peign e´ , C.M., M o¨ nkk o¨ nen, H.,
M o¨ nkk o¨ nen, J., Li, J., Kuball, J., Adams, E.J., Netzer, S., et al. (2012). Key
implication of CD277/butyrophilin-3 (BTN3A) in cellular stress sensing by a
major human gd T-cell subset. Blood 120, 2269–2279.
Vavassori, S., Kumar, A., Wan, G.S., Ramanjaneyulu, G.S., Cavallari, M., El
Daker, S., Beddoe, T., Theodossis, A., Williams, N.K., Gostick, E., et al.
(
2013). Butyrophilin 3A1 binds phosphorylated antigens and stimulates human
gd T cells. Nat. Immunol. 14, 908–916.
Hecker, S.J., and Erion, M.D. (2008). Prodrugs of phosphates and phospho-
Vranken, W.F., Boucher, W., Stevens, T.J., Fogh, R.H., Pajon, A., Llinas, M.,
Ulrich, E.L., Markley, J.L., Ionides, J., and Laue, E.D. (2005). The CCPN data
model for NMR spectroscopy: development of a software pipeline. Proteins
nates. J. Med. Chem. 51, 2328–2345.
Kim, M., Kleckley, T.S., Wiemer, A.J., Holstein, S.A., Hohl, R.J., and Wiemer,
D.F. (2004). Synthesis and activity of fluorescent isoprenoid pyrophosphate
analogues. J. Org. Chem. 69, 8186–8193.
5
9, 687–696.
Wang, T., Gao, Y., Scully, E., Davis, C.T., Anderson, J.F., Welte, T., Ledizet, M.,
Koski, R., Madri, J.A., Barrett, A., et al. (2006). Gamma delta T cells facilitate
adaptive immunity against West Nile virus infection in mice. J. Immunol. 177,
Kornberg, R.D., McNamee, M.G., and McConnell, H.M. (1972). Measurement
of transmembrane potentials in phospholipid vesicles. Proc. Natl. Acad. Sci.
USA 69, 1508–1513.
1
825–1832.
Li, J., Cui, L., and He, W. (2005). Distinct pattern of human Vdelta1 gammadelta
Wang, H., Henry, O., Distefano, M.D., Wang, Y.C., R a¨ ikk o¨ nen, J., M o¨ nkk o¨ nen,
J., Tanaka, Y., and Morita, C.T. (2013). Butyrophilin 3A1 plays an essential role
in prenyl pyrophosphate stimulation of human Vg2Vd2 T cells. J. Immunol. 191,
1029–1042.
T cells recognizing MICA. Cell. Mol. Immunol. 2, 253–258.
Li, H., Xiang, Z., Feng, T., Li, J., Liu, Y., Fan, Y., Lu, Q., Yin, Z., Yu, M., Shen, C.,
and Tu, W. (2013). Human Vg9Vd2-T cells efficiently kill influenza virus-infected
lung alveolar epithelial cells. Cell. Mol. Immunol. 10, 159–164.
Wiemer, A.J., Yu, J.S., Shull, L.W., Barney, R.J., Wasko, B.M., Lamb, K.M.,
Hohl, R.J., and Wiemer, D.F. (2008). Pivaloyloxymethyl-modified isoprenoid
bisphosphonates display enhanced inhibition of cellular geranylgeranylation.
Bioorg. Med. Chem. 16, 3652–3660.
Maalouf, M.A., Wiemer, A.J., Kuder, C.H., Hohl, R.J., and Wiemer, D.F. (2007).
Synthesis of fluorescently tagged isoprenoid bisphosphonates that inhibit pro-
tein geranylgeranylation. Bioorg. Med. Chem. 15, 1959–1966.
Marcellin, P., Chang, T.T., Lim, S.G., Tong, M.J., Sievert, W., Shiffman, M.L.,
Jeffers, L., Goodman, Z., Wulfsohn, M.S., et al.; Adefovir Dipivoxil 437 Study
Group (2003). Adefovir dipivoxil for the treatment of hepatitis B e antigen-pos-
itive chronic hepatitis B. N. Engl. J. Med. 348, 808–816.
Wiemer, A.J., Wernimont, S.A., Cung, T.D., Bennin, D.A., Beggs, H.E., and
Huttenlocher, A. (2013). The focal adhesion kinase inhibitor PF-562,271
impairs primary CD4+ T cell activation. Biochem. Pharmacol. 86, 770–781.
Wiemer, D.F., and Wiemer, A.J. (2014). Opportunities and challenges in devel-
opment of phosphoantigens as Vg9Vd2 T cell agonists. Biochem. Pharmacol.
89, 301–312.
McKenna, C., Higa, M., Cheung, N., and McKenna, M. (1977). Facile dealkyla-
tion of phosphonic acid dialkyl esters by Bromotrimethylsilane. Tetrahedron
Lett. 18, 155–158.
Wrobel, P., Shojaei, H., Schittek, B., Gieseler, F., Wollenberg, B., Kalthoff, H.,
Kabelitz, D., and Wesch, D. (2007). Lysis of a broad range of epithelial tumor
cells by human gamma delta T cells: involvement of NKG2D ligands and
T-cell receptor- versus NKG2D-dependent recognition. Scand. J. Immunol.
66, 320–328.
Miyagawa, F., Tanaka, Y., Yamashita, S., and Minato, N. (2001). Essential
requirement of antigen presentation by monocyte lineage cells for the activa-
tion of primary human gamma delta T cells by aminobisphosphonate antigen.
J. Immunol. 166, 5508–5514.
954 Chemistry & Biology 21, 945–954, August 14, 2014 ª2014 Elsevier Ltd All rights reserved