Structure-activity relationships of butyrophilin 3 ligands

Abstract

Phosphoantigens (pAgs) are small phosphorus-containing molecules that stimulate Vγ9Vδ2 T cells with sub-nanomolar cellular potency. Recent work has revealed these compounds work through binding to the transmembrane immunoglobulin butyrophilin 3A1 (BTN3A1) within its intracellular B30.2 domain. Engagement of BTN3A1 is critical to formation of an immune synapse between cells that contain pAgs and the Vγ9Vδ2 T cells. This review summarizes the structure-activity relationships of pAgs and their implications to the mechanisms of butyrophilin 3 activation leading to Vγ9Vδ2 T cell response.

Graphical Abstract

Stimulating immunity: This review focuses on structure-activity relationships of butyrophilin 3 agonists, or phosphoantigens, which function in infectious disease and cancer. An overview of the naturally occurring compounds and their synthetic analogs that impact receptor binding and cellular activity is presented. The emphasis is on re-interpreting cellular activity in the context of recent discoveries on butyrophilin receptor biology.

1. Introduction

T cells that express the γδ T cell receptor (TCR) heterodimer^[1–2] respond to antigens in ways that could be described as “unconventional” due to their independence from the major histocompatibility complex (MHC).^[3–4] Yet, like conventional T cells the γδ T cells play important roles in initiating an immune response to foreign antigens, particularly in mucosal compartments,^[5] but also systemically. The antigens and their mechanisms of detection by the γδ T cells are in many cases still yet to be identified. However, recent studies into structure-activity relationships underlying Vγ9Vδ2 T cell antigen detection have greatly advanced our understanding of this system.

The Vγ9Vδ2 T cells are the most prevalent form of γδ T cells in human peripheral blood,^[6] representing about 1-10% of total peripheral blood T cells in healthy individuals.^[7–8] These cells respond to non-peptide^[9] phosphorylated^[10–11] small molecules known as phosphoantigens (pAgs).^[12–13] Studies into the immune response to Mycobacterium tuberculosis identified that Vγ9Vδ2 T cells rapidly expand in response to this infection.^[14–15] Fractionation of M. tuberculosis extracts led to identification of the 1-deoxy-D-xylulose 5-phosphate (DOXP) / methylerythritol 4-phosphate (MEP) pathway (Figure 1A) intermediates such as (E)-4-hydroxy-3-methyl-but-2-enyl diphosphate (HMBPP) and isopentenyl diphosphate (IPP) as the active components (Figure 1B).^[16–17] In addition to M. tuberculosis, other pathogens of bacterial and parasitic origin stimulate the Vγ9Vδ2 T cells. These include Plasmodium (during malaria paroxysms Vγ9Vδ2 T cells expand up to 30% of total T cells^[18]), Francisella tularensis (during tularemia Vγ9Vδ2 T cells represent up to 25% of total T cells^[19]), and Ehrlichia (during ehrlichiosis Vγ9Vδ2 T cells comprise up to 60% of peripheral blood lymphocytes^[20]).

Figure 1.

The natural pAgs HMBPP and IPP are derived from the DOXP/MEP pathway in pathogens and the mevalonate pathway in humans. A) The DOXP/MEP pathway produces pAgs IPP and HMBPP from DOXP in pathogens. HMBPP is the immediate precursor to IPP in pathogens and is not found in human metabolism. B) Chemical structures of the natural pAgs HMBPP and IPP. C) The mevalonate pathway produces IPP from mevalonate in humans. IPP is a core building block of larger isoprenoids including FPP, GGPP, polyprenyl-PPs, squalene and cholesterol. Human IPP can be decreased by treatment with statin drugs or increased by treatment with bisphosphonate drugs. IPP is the first intermediate shared between the pathogenic and human pathways. Both IPP and HMBPP can function as direct pAg but HMBPP is 5 to 6 log units more potent.

The Vγ9Vδ2 T cells also exhibit anti-cancer activity.^[21–22] They can target cancer cells following engagement of the Vγ9Vδ2 TCR or activation of other proteins such as C-type lectin receptors. When Vγ9Vδ2 T cells encounter malignant cells that contain elevated levels of pAgs, such as the endogenous mevalonate pathway (Figure 1C) intermediate IPP, they can directly lyse them with a TCR dependent perforin/granzyme response.^[23–25] They are of interest for use in immunotherapy as alternative or complimentary approaches to αβ T cell therapies,^[26] particularly in cases where αβ T cell tumor infiltration may be limited and checkpoint blockade is ineffective,^[27] as their tumor infiltration is not correlated to αβ T cells^[28] and may be more favorable to survival.^[29] Thus, pAgs are relevant to the pathogenesis of both infectious disease and cancer, and understanding their structure-activity relationships (SAR) could enable development of new therapeutics or vaccines for these indications.

2. HMBPP and BTN3A1

2.1. HMBPP is the prototypical pAg

Many species contain the biochemical pathways to produce pAgs,^[30] but only a handful of natural molecules have pAg activity.^[31] Resolving their SAR has been complicated. Not only was the molecular target unknown, but also pAgs are biochemical pathway intermediates that exist in equilibrium with upstream and downstream molecules, other phosphorylated forms, and nucleotide-conjugated forms. They can act directly through binding the pAg receptor BTN3A1 (Figure 2A)^[32–33] or indirectly, either by metabolic conversion to a direct pAg or by increasing cellular levels of endogenous direct pAgs (Figure 2B).^[33–34]

Figure 2.

Direct versus indirect pAg activity. A) Direct pAgs such as HMBPP bind directly to BTN3A1 in its intracellular B30.2 domain. B) Indirect pAgs elevate endogenous IPP levels and promote its engagement of BTN3A1. This can occur via metabolic conversion to produce IPP or by inhibition of downstream enzymes which blocks use of IPP.

Initial reports identified IPP and DMAPP as pAgs.^[16] Their cellular potency falls in the micromolar range, lower than expected relative to M. tuberculosis extract. Additional M. tuberculosis fractionation experiments identified 3-formyl-1-butyl diphosphate^[17] as a putative pAg with nanomolar activity. Around the same time, genetic experiments using E. coli mutants determined that HMBPP (aka (E)-4-hydroxy-dimethylallyl diphosphate or HDMAPP) was produced by bacteria and at least 10,000x more potent than IPP.^[35] Subsequent evaluation showed that the compound initially identified as 3-formyl-1-butyl diphosphate was actually HMBPP.^[36–37]

HMBPP remains the most potent direct pAg found in nature.^[12] With continuous exposure HMBPP stimulates proliferation of human Vγ9Vδ2 T cells at picomolar concentrations. It can be viewed as the prototypical pAg. Though much less potent than HMBPP, IPP remains the most potent endogenous direct pAg. IPP is active at concentrations slightly above physiological levels, so that conditions which elevate IPP cause it to function as a pAg. Other naturally occurring pAgs generally result from metabolism or conjugation of these two core compounds. This includes the natural nucleotidic pAgs,^[38] which are produced in pathogens through conjugation between nucleotides and HMBPP^{[6, 11]} or in human cells through conjugation between nucleotides and IPP.^[39–40] In infectious contexts it is most likely that Vγ9Vδ2 T cells are activated by exogenous HMBPP or HMBPP derivatives produced by the pathogen, due to the potency of HMBPP relative to IPP. However, it remains unclear if endogenous pAgs also contribute to this response.

2.2. BTN3A1 is the pAg receptor

Vγ9Vδ2 T cells can respond quite sensitively to low levels of HMBPP produced by pathogens and also respond to higher levels of IPP produced endogenously. The mechanism of this “unconventional” response is still not completely clear, but it differs from αβ T cells. Structures of the Vγ9Vδ2 TCR are similar to the αβ TCR, but there is no direct interaction between the pAg and Vγ9Vδ2 TCR.^[41] In contrast, antibodies to the butyrophilin protein BTN3A1 can mimic the pAg effect on Vγ9Vδ2 T cells or block the pAg effect on Vγ9Vδ2 T cells.^[42] Depletion of BTN3A1 with siRNA or CRISPR confirm it is required for pAg activity.^[43–46]

BTN3A1 is a single pass transmembrane protein with extracellular N terminal IgV and IgC domains similar to B7 family proteins, though it additionally contains a large intracellular B30.2 domain (Figure 3A/B). Initially, an extracellular HMBPP binding site on BTN3A1 was identified.^[43] However, the extracellular binding site within the IgV domain has not been confirmed, and instead an intracellular binding site for HMBPP within the BTN3A1 B30.2 domain was identified.^{[32–33, 44]} Thus, not only is BTN3A1 essential for pAg activity, but also the intracellular B30.2 domain of BTN3A1 functions as the pAg receptor. How exactly ligand-bound BTN3A1 triggers the Vγ9Vδ2 TCR or other components is not fully understood, though other BTN3 isoforms are likely involved,^[46] and recent studies identified that BTN2A1 is also essential for pAg activity.^[47–48] These butyrophilins likely coordinate and undergo a conformational change upon pAg binding that is detectable by surface interactions with the Vγ9Vδ2 TCR (Figure 3C).

Figure 3.

Role of butyrophilins in pAg detection by Vγ9Vδ2 T cells. A) There are 3 isoforms of BTN3 in humans. Each contain extracellular IgV and IgC domains. The BTN3A1 and 3A3 isoforms contain an intracellular B30.2 domain. BTN2A1 is similar in domain organization to BTN3A1. The B30.2 domain of BTN3A1 but not BTN3A3 or BTN2A1 binds to pAg. B) All four proteins (BTN3A1 shown) are single pass transmembrane proteins. C) pAg binding activates the BTN2/3 complex, enabling detection by Vγ9Vδ2 T cells.

3. Structure-activity relationships of pAgs and other butyrophilin 3 ligands

3.1. Methods for SAR determination

Early studies evaluated the cellular SAR prior to the identification of BTN3A1 as the receptor. Typically, assessment of cellular SAR involves measurement of the effector functions of Vγ9Vδ2 T cells, such as cell proliferation, cytokine production, or cell-mediated cytotoxicity. Cytokines are usually Th1 origin, such as interferon γ (IFN-γ) or tumor necrosis factor α (TNF-α). Potent pAgs are active in the high picomolar to low nanomolar range in these cellular assays, yet the activity of the same compound can vary significantly among readouts. Evaluation of SAR is complicated by the facts that not all pAgs are butyrophilin ligands (many compounds act indirectly) and not all butyrophilin ligands are pAgs (some compounds bind BTN3A1 without pAg activity). For example, bisphosphonate drugs administered for osteoporosis were found to induce a flu-like acute phase response.^[49] The response is dependent on IPP accumulation in monocytes and could be blocked by IPP depletion with a statin drug.^[50] Thus, bisphosphonates are pAgs but act indirectly. Additionally, HMBPP analogs with increased BTN3A1 binding affinity showed decreased pAg activity.^[51] Thus, BTN3A1 binding in required but binding alone is not sufficient to produce pAg activity.

Since the BTN3A1 B30.2 domain was identified as the molecular target, SAR for pAg binding to BTN3A1 can be evaluated. HMBPP binding has been measured by isothermal titration calorimetry (ITC),^{[32–33, 44, 52]} nuclear magnetic resonance (NMR) spectroscopy,^{[33, 52–53]} x-ray crystallography,^{[32, 51]} small-angle X-ray scattering (SAXS)^[52] and fluorescence polarization (FP).^[54] Collectively, these results show HMBPP binds BTN3A1 with affinity in the low micromolar range.^{[32–33, 44]} This is substantially weaker than had been expected based on the cellular activity of HMBPP which is in the high picomolar to low nanomolar range. The reasons are still being evaluated, but this could include that an additional protein is required to stabilize the binding, or that the binding kinetics serve to control the potency of a signal that is quite sensitively detected by the Vγ9Vδ2 T cells.

Despite the fact that ligand binding is weaker than cellular activity, it is clear that the SAR ranking trends for pAg cellular activity and BTN3A1 binding are generally consistent, however there is an exponential rather than linear relationship between BTN3A1 binding and pAg activity. This is evident because small differences in binding affinity lead to substantial differences in cellular potency, even in similar compounds. The following SAR are primarily derived from the cellular experiments, though BTN3A1 binding is referenced when possible to provide context.

3.2. The HMBPP diphosphate, isoprene unit, and allylic alcohol each contribute to BTN3A1 binding and pAg activity

HMBPP contains several features critical to its function, including the diphosphate, the isoprene unit, and the allylic alcohol (Figure 4). The shallow HMBPP binding pocket is located on the surface of the B30.2 domain (Figure 5). The HMBPP diphosphate interacts with a number of basic amino acid residues to anchor the pAg to the protein.^{[32, 51]} Within the binding pocket, the allylic alcohol interacts with a critical histidine residue^{[32, 51, 53, 55]} and potentially the neighboring tryptophan and tyrosine.^[51] The isoprene unit serves as a linker to position the allylic alcohol relative to the diphosphate, though it also interacts with the protein.

Figure 4.

HMBPP is the prototypical pAg. The chemical structure of HMBPP contains an isoprenoid moiety and a diphosphate moiety. It carries a charge of near −2.5 at physiological pH. The diphosphate contains two bridging oxygen atoms (1 and 3α) and five non-bridging oxygen atoms (1α, 2α, 1β, 2β, and 3β) which provide strength to the BTN3A1 binding interaction. The allylic oxygen (position 4) contributes to both BTN3A1 binding and direct pAg activity.

Figure 5.

HMBPP binding to BTN3A1. A) HMBPP binds within a shallow basic binding pocket in the intracellular B30.2 domain of BTN3A1, a comprised largely of beta strands. B) HMBPP forms interactions with the protein at its diphosphate, linker region, and allylic alcohol. Images of 5ZXK^[51] created with NGL.^[56]

Both phosphates are required for optimal binding and pAg activity as alkaline phosphatase treatment abrogates pAg activity.^[57] Crystal structures revealed six interactions between the non-bridging oxygen atoms of the β phosphate and three arginine residues, while the non-bridging oxygen atoms of the α phosphate also interacted with two residues through bridged water molecules.^{[32, 51]} Neither study found interactions between the protein and the bridging oxygen atoms of HMBPP. In the case of bridging oxygen 1, this is not surprising because replacement with a carbon has minimal negative impact on pAg activity (TNFα release EC₅₀ HMBPP = 0.39 nM versus C-HMBPP EC₅₀ = 0.91 nM) (Figure 6).^[58] Both compounds also bind to BTN3A1 with similar affinities (BTN3A1 B30.2 domain K_D HMBPP = 0.92 μM versus C-HMBPP 0.51 μM).^[32] However, in the case of bridging oxygen 3α, carbon replacement is not well tolerated and results in substantial loss of pAg activity (HMBPCP proliferation EC₅₀ = 5,300 nM).^{[55, 59–60]} Replacement of both bridging oxygen atoms with carbon atoms modestly decreases BTN3A1 binding affinity relative to HMBPP (74 fold) but substantially decreases pAg activity (proliferation EC₅₀ = 26,000 nM).^[54–55] It is possible that oxygen 3α has unrecognized molecular interactions that are important to pAg activity. While many diphosphates have measurable BTN3A1 binding affinity, binding alone does not cause pAg activity.^[54]

Figure 6.

Structures of synthetic phosphonate analogs of HMBPP. Replacement of bridging oxygen 1 with a carbon has little impact on potency. Replacement of bridging oxygen 3α with a carbon retains moderate binding affinity but negatively impacts cellular potency. The EC₅₀ values listed are for continuous exposure experiments. In washout experiments, the phosphonates display superior direct pAg activity.

The allylic alcohol is also important for BTN3A1 binding and for maximal pAg activity. This portion of the pAg interacts with the histidine residue at position 381 of BTN3A1 through hydrogen bonding. Point mutations to this histidine residue in BTN3A1 decrease its pAg sensing function while incorporation of histidine into this position of BTN3A3 results in gain of pAg sensing function.^[32] Removal of the alcohol decreases potency (HMBPP proliferation EC₅₀ = 70 pM versus DMAPP EC₅₀ = 230 nM, ~3300 fold^[61]). On the phosphinophosphonate backbone (CC-HMBPP), removal of the allylic alcohol also decreases BTN3A1 binding and abrogates pAg activity.^[55] Smaller diphosphates that also lack an allylic alcohol show only weak pAg activity (Figure 7),^[62] and an indirect mechanism cannot be excluded. Halohydrins of HMBPP retain potency, despite lacking the olefin (Figure 8).^[59] Within the series, there is a clear SAR trend favoring the larger iodine. The potency of these halohydrins is generally similar to HMBPP. One study showed the pAg activity of HMBPP and BrHPP was equal (proliferation EC₅₀ HMBPP ≈ BrHPP ≈ 10 nM),^[59] while another study found that BrHPP trailed the potency of HMBPP by ~60 fold (proliferation EC₅₀ of 0.39 versus 23.85 nM).^[58]

Figure 7.

Equilibrium of IPP and DMAPP. DMAPP is interconverted to the direct pAg IPP by the enzyme IPP isomerase (IDI). Other similar synthetic compounds demonstrate cellular pAg activity, but lack the allylic alcohol (or electronegativity) normally required for potent direct pAg activity. These compounds may work indirectly through impacting an enzyme involved in IPP metabolism.

Figure 8.

Synthetic halohydrins as HMBPP analogs. Incorporation of halogens into the pAg scaffold shows a clear trend favoring the larger atoms. These compounds show potent direct pAg activity.

The isoprene unit itself is optimal in the naturally occurring E configuration. In cells, the Z isoprene analog is less potent relative to the E isomer (proliferation EC₅₀ E-HMBPP = 0.39 nM versus Z-HMBPP = 252 nM, ~650 fold) (Figure 9).^[58] Longer chain length compounds are inactive in cellular assays despite measurable binding to BTN3A1^[54] (Figure 10). However, the inactivity of C-HMHP may result from lack of oxygen 3α and the intermolecular interactions in which it may participate. Decreased chain length compounds are also less active or inactive,^{[58, 63]} though in the diphosphate form some shorter length compounds are active (e.g. Figure 7).^[16] Taken together, direct acting pAgs must contain three elements- 1) a diphosphate (or monophosphate), 2) an alcohol (or electronegative functional group), and 3) a short isoprene (or alkyl) linker to connect the first two features.

Figure 9.

Impact of olefin stereochemistry on pAg activity. The naturally occurring E isomer is preferred over the synthetic Z isomer, both on the phosphate and phosphonate scaffold.

Figure 10.

Impact of chain length and phosphonate position on synthetic pAg activity. Removal of the β phosphate produces a moderate pAg, however carbon replacement of the α phosphate abrogates pAg activity.

3.3. IPP, DMAPP, and indirect pAgs

Compared to HMBPP, IPP lacks the allylic alcohol but retains some electron density with a terminal olefin. IPP binding to BTN3A1 is weaker than HMBPP binding (BTN3A1 B30.2 binding K_D HMBPP = 0.92 μM versus IPP 490 μM, ~530 fold),^[32] and the cellular activity of IPP is also weaker than that of HMBPP (proliferation EC₅₀ HMBPP = 0.51 nM versus IPP = 36,000 nM, ~71,000 fold).^[33] Like HMBPP, IPP is more potent in the diphosphate form relative to the phosphate form (proliferation EC₅₀ IP = 700 μM versus IPP = 3 μM, ~230 fold).^[16] In theory, due to the IPP isomerase reaction the cellular pAg activity of IPP and DMAPP should be similar. However, some groups have reported IPP as less potent than DMAPP (proliferation EC₅₀ IPP = 21 μM versus DMAPP = 230 nM, ~91 fold).^[61]

Compounds that increase IPP can indirectly act as pAgs (Figure 11).^[64–65] Bisphosphonates that inhibit farnesyl diphosphate synthase (FDPS) function by this mechanism (Figure 11A). Because FDPS uses IPP as a substrate and depletes sterols, FDPS inhibition increases IPP, which can bind to BTN3A1. Likewise, decreased expression of FDPS can also increase IPP to active levels.^[66] The bisphosphonate SAR mirrors their ability to inhibit FDPS, with the second generation bisphosphonates pamidronate and alendronate having lower FDPS inhibition and lower pAg activity relative to the third generation compounds risedronate and zoledronate (Figure 11B).^[67–68] Some bisphosphonates also inhibit other enzymes of isoprenoid metabolism, like decaprenyl diphosphate synthase (PDSS), which may contribute to their activity in addition to FDPS inhibition (Figure 1C).^[65] For example, a lipophilic pyridinium bisphosphonate inhibits both FDPS and PDSS and is a potent pAg (Figure 11C). PDSS inhibition alone is not likely sufficient for pAg activity. This relationship between FDPS inhibition and pAg activity is not completely correlated, as the zoledronate analog deshydroxyzoledronate is less potent for inhibition of FDPS but more potent for pAg activity relative to zoledronate (Figure 11D).^[69] In contrast, fluorozoledronate was a less active pAg (Figure 11E).^[70] Taken together, ability of bisphosphonates to function as indirect pAgs depends on their potency of FDPS inhibition, profile of off-target enzyme inhibition, and ability to engage the cellular target to elevate endogenous IPP.

Figure 11.

Indirect pAg activity resulting from FDPS inhibition. A) Bisphosphonate drugs inhibit farnesyl diphosphate synthase (FDPS). Because FDPS utilizes IPP as a substrate and also increases isoprenoid flux by depleting sterols, inhibition of this enzyme results in elevated cellular IPP, which can engage BTN3A1. B) The inhibition of FDPS by clinical bisphosphonate drugs mirrors their pAg activity. C) Lipophilic bisphosphonates show excellent cellular pAg activity. D) Removal of the central bisphosphonate alcohol increases pAg activity. E) Replacement of the alcohol with fluorine modestly decreases pAg activity.

PAg activity does not occur with inhibitors of downstream isoprenoid pathway enzymes like geranylgeranyl diphosphate synthase (GGDPS),^[65] because they do not deplete sterols and thus elevate IPP.^[71] However, it is possible that inhibition of other IPP metabolizing enzymes^[72] could result in indirect pAg activity. Furthermore, the structures of direct BTN ligands are often similar to the structures of isoprenoid pathway inhibitors so there is a probability that pAg activity of some compounds could result from both direct and indirect components. Compounds lacking any of the three core pAg features described above are assumed to function indirectly.

4. Further roles of the diphosphate

4.1. The diphosphate is a barrier to membrane diffusion

HMBPP is highly acidic and carries a significant charge at physiological pH that reduces movement across cellular lipid bilayers.^[45] Cellular and extracellular levels of HMBPP are regulated by transporters and endocytosis/pinocytosis. Secretion of HMBPP^[73] from intracellular pathogens such as M. tuberculosis would position it within the cytosol to bind to the B30.2 domain of BTN3A1.^[4] Secretion of IPP from bisphosphonate treated cells using the ABCA1 transporter has also been reported, and the authors of that study suggested that some cells might also secrete HMBPP after uptake of pathogens.^[74] Extracellular HMBPP/IPP can be internalized through energy dependent uptake mechanisms such as endocytosis/pinocytosis, in which case acidification of endocytic vesicles promotes diffusion or transport into the cytosol and access to the B30.2 domain.^[45] Even in the absence of endocytosis, such as in fixed cells or cells kept at low temperatures, diffusion of HMBPP into the cytosol can occur given sufficient exposure time or high enough extracellular concentrations. Thus, early reports showing fixed cells can be “pulsed” with diphosphate pAgs are probably accurate, but rather than confirming cell surface pAg presentation they are more likely measuring transmembrane diffusion of potent direct pAgs in sufficient quantity to engage BTN3A1.^{[4, 73]}

To promote intracellular delivery of pAgs, the negatively charged non-bridging oxygen atoms can be protected with charge neutral and cell-cleavable functional groups to form pAg prodrugs (Figure 12).^[33] These modifications dramatically increase diffusion across the cell membrane. Once internalized, release of the charged pAg occurs, which itself is unable to diffuse out of the cell. Prodrugs such as POM₂-C-HMBP are more potent relative to its corresponding free acid (proliferation EC₅₀ POM₂-C-HMBP = 5.4 nM versus C-HMBP = 4 μM, ~740 fold).^[33] Additional prodrug forms such as the aryl esters and aryl amides increased cellular potency even further (proliferation EC₅₀ POM/1-NAP-C-HMBP= 0.79 nM^[75–76] versus 1-NAP/GlyOMe C-HMBP = 0.44 nM^[77–78]).

Figure 12.

Synthetic pAg prodrugs. The chemical structures of C-HMBP and its various prodrug forms. These prodrugs increase the direct pAg activity.

Indirect pAgs that contain a bisphosphonate also face a barrier to membrane diffusion. Since the enzymes of isoprenoid metabolism are compartmentalized, these compounds may need to cross more than one membrane to engage their target. Similar to the direct pAg prodrugs, POM (pivaloyloxymethyl) prodrug forms of the indirect-acting bisphosphonates such as tetrakis-pivaloyloxymethyl 2-(thiazole-2-ylamino)ethylidene-1,1-bisphosphonate (PTA) ^[79–81] also have strong cellular pAg activity compared to the free acid (TA) form (TNFα EC₅₀ TA = 3,000 nM versus PTA = 3 nM, ~1000 fold) (Figure 13).^[79]

Figure 13.

A bisphosphonate prodrug. The chemical structures of zoledronate, an analog (TA), and its tetra-POM prodrug form (PTA). The prodrug form increases cellular FDPS inhibition and indirect pAg activity.

Beyond prodrug approaches, increased hydrophobicity can be used to promote diffusion across the membrane. This effect has been observed with both direct pAgs such as the bulky HMBPP analogs (Figure 14)^[51] or indirect pAgs such as the lipophilic bisphosphonates (Figure 11C).^[65] Ultimately, whether direct- or indirect-acting, enhancements to permeability increase cellular pAg activity.

Figure 14.

Bulky HMBPP methyl replacements. The chemical structures of HMBPP and some synthetic analogs with bulkier substituents at the methyl position. These bulky analogs block ternary complex formation for reduced direct pAg activity, despite higher binding affinity to BTN3A1.

4.2. The diphosphate is an on-off switch

The metabolic stability of the HMBPP diphosphate influences its pAg activity. Diphosphates are quite susceptible to esterases, both inside and outside of cells. Hydrolysis of either of the HMBPP bridging anhydrides reduces pAg activity and limits the duration of signaling. HMBPP washout experiments show a complete loss of response within twelve hours even when cells are highly loaded with HMBPP.^[45] The cellular half live of HMBPP is sufficient to turn off pAg signaling if infection is resolved. The instability of the HMBPP diphosphate is also important in plasma. During in vivo experiments, diphosphate pAgs are typically dosed in continuous infusion because compound is depleted from plasma within minutes.^[82]

While esterases function to turn off pAg activity, kinases may play a role in turning it on. At least some level of phosphorylation likely occurs because even the diol 1,4-dihydroxy-3-methyl-2-butene shows cellular pAg activity.^[61] The β phosphate is apparently not required,^[58] but loss of the β phosphate results in ~2800 fold decrease in potency (TNFα release EC₅₀ C-HMBP = 2530 nM versus C-HMBPP = 0.91 nM). It is unclear whether some cellular phosphorylation of C-HMBP occurs, though even minor phosphorylation could be relevant given the difference in potency between the mono- and diphosphate forms.

Replacement of bridging oxygen 1 with the more metabolically stable methylene bridge to form a phosphonate (Figure 6) dramatically increases the metabolic stability of the structure without significant loss of potency during continuous exposure experiments.^[58] This confirms that the oxygen at position 1 is a key point of HMBPP metabolism in cells and plasma. Because of the metabolic instability of the diphosphate at this position, pAg SAR studies that utilize the most common approach of continuous compound exposure come with a major caveat- potent compounds containing a diphosphate are not good candidates for in vivo application regardless of potency. In cellular washout experiments, which more closely resemble typical in vivo dosing regimens, the superiority of the phosphonate backbone becomes evident. For example, despite being slightly less potent in continuous exposure experiments, in a four hour treatment washout experiment C-HMBPP retained activity in the nanomolar range while HMBPP was largely inactive (C-HMBPP EC₅₀ = 280 nM versus HMBPP EC₅₀ > 100 μM, >350 fold).^[83] Thus, carbon replacement of bridging oxygen 1 can block cells from switching off the pAg and extend the duration of pAg activity, providing a form of kinetic selectivity to these molecules.^[84]

Interestingly, the nucleotidic pAgs contain features that may impact both uptake and intracellular stability (Figure 15). Though the uptake is not well understood, addition of the nucleotide decreases the charge to mass ratio relative to the free pAg, which could increase transmembrane diffusion. At the same time, the nucleotide could increase transporter-mediated uptake. In cellular experiments using continuous exposure, the pAg potency of HMBPP, ApppH, and dTpppH is quite similar.^[38] In washout experiments, dTpppH clearly retains potency for a longer duration than the other compounds. However, BTN3A1 binding affinity of ApppH and dTpppH is reduced relative to C-HMBPP (C-HMBPP K_D = 1 μM versus ApppH KD = 100 μM and dTpppH = 200 μM). Taken together, certain nucleotidic conjugates of HMBPP are likely metabolized to release HMBPP as the active component, but provide advantages with respect to duration of activity which may involve protection against phosphoesterases that act on HMBPP.

Figure 15.

Nucleotidic pAgs. The chemical structures of HMBPP, C-HMBPP, and some nucleotidic analogs. The nucleotide conjugates of HMBPP show decreased binding to BTN3A1 but increased duration of cellular activity in washout experiments.

5. Unexpected SAR trends and implications to mechanism

5.1. Probes of cooperativity

Because BTN3A1 may exist in a dimeric or multimeric complex^[46] like some other B7 proteins,^[85] it is possible that multiple binding events coordinate to trigger full activation. Cooperativity in binding could be revealed through a non-linear relationship between pAg binding to BTN3A1 and cellular pAg activity. For example, the cellular pAg activity of HMBPP and IPP differ by ~71,000 fold,^[33] yet the compounds bind to BTN3A1 with ~530 fold difference.^[32] Since the diphosphate moiety is identical in these compounds, the differential binding and activity solely arise from the presence or absence of the allylic alcohol and the position of the olefin. However, metabolism of IPP may cloud this interpretation. Also supporting this hypothesis is the pattern of activity observed between HMBPP and CC-HMBPP.^[55] These compounds exhibit very similar molecular structures, identical at the allylic alcohol and olefin, and likely similar cell internalization. Their continuous exposure cellular activity differs by ~70,000 fold, while their binding affinities were found to differ by only about ~100 fold. However, unrecognized interactions of oxygen 3α with another protein may cloud this interpretation. Additionally, it remains possible that uptake differences resulting from unknown transporter recognition influence the relative cellular activity of these compounds. Further direct pAgs and SAR studies are needed to more fully understand this relationship.

5.2. Probes of quaternary structure

Because HMBPP binds to a shallow binding pocket in the B30.2 domain of BTN3A1, the possibility exists that portions of the bound ligand can interact with another protein to form a ternary complex. One possible ternary complex constituent is another butyrophilin molecule. Indeed, an asymmetric dimer has been observed in prior crystal structures of the BTN3A1 B30.2 domain.^{[32, 86]} Supporting this notion is the pattern of activity observed with HMBPP analogs that contain bulky extensions of the methyl group that are predicted to extend into solvent exposed area. Here, replacement of the methyl group with larger lipophilic substituents increases binding to BTN3A1 (HMBPP K_D = 3300 nM and 4-methylbenzyl HMBPP K_D = 260 nM, ~13 fold). The extra lipophilic character is also expected to increase cell permeability. However, these extensions decrease cellular pAg activity and disrupt the ability of BTN3A1 to form an asymmetrical dimer^[51] (Figure 14).

6. Summary and Outlook

In conclusion, studies over the past decade have now established that HMBPP is the prototypical pAg and the BTN3A1 B30.2 domain is the HMBPP receptor. Direct acting pAgs interact with the B30.2 domain in the intracellular portion of the protein, and function to propagate signals to the extracellular portion of the BTN3A1 complex, which also includes roles for BTN2A1, 3A2, and 3A3. These butyrophilins interact both on the cell surface and intracellularly, the latter interaction may involve multiple subunits binding the ligand in a ternary complex and/or multiple ligand binding events. Once the BTN3A1 complex is activated by intracellular ligand binding, the extracellular changes can be detected at low levels by the Vγ9Vδ2 T cells using their TCR.

The identification of BTN3A1 as the HMBPP receptor has allowed new perspectives on the receptor binding and cellular activity of new compounds, and enables future efforts at structure-based drug design. It also allows reinterpretation of prior cellular SAR in the context of the binding partner. It appears that their SAR patterns are clearly correlated, but the correlation is not linear and the processes by which moderate binding affinity leads to potent cellular activity are still not fully understood. For direct pAg activity, a diphosphate, an alcohol, and hydrophobic linker are required. Modifications to these functional groups may enhance potency and stability but some key positions cannot be modified without negatively impacting activity. Additional challenges in development of synthetic butyrophilin ligands for therapeutic application include improvements to stability, internalization, and cell-type specificity.

Biography

Andrew J. Wiemer was born and raised in the United States of America. He received his B.S. from the University of Notre Dame (1999), his M.S. from Colorado State University (2002), and his Ph.D. from the University of Iowa under Raymond Hohl (2008). He was a postdoctoral fellow at the University of Wisconsin under Anna Huttenlocher. In 2012, he was appointed Assistant Professor at the University of Connecticut and promoted to Associate Professor in 2018. His research interests are in cancer immunology, drug discovery, and immune receptors.