An epitope is an antibody-recognition site on a target antigen. As such, active immunization of epitope peptides may induce therapeutic efficacy equivalent to the administration of parent antibody medicines. In the present study, we designed peptides based on the epitope recognized by the tumor-suppresive anti-CD98 monoclonal antibody HBJ127, and investigated their efficacy for induction of antitumor immunity. The immune sera showed reactivity against the corresponding peptide–keyhole limpet hemocyanin (KLH) and peptide–bovine serum abumin (BSA) conjugates, although they did not react with CD98-positive HeLa cells or recombinant CD98 heavy chain. To elucidate whether the epitope peptide failed to induce antitumor immunity or not, we constructed the IgG1, κ Fab phage display libraries from spleen cells of immunized mice and tried to retrieve CD98-reactive recombinant Fab (rFab) fragments by panning against either epitope peptide–BSA conjugates or live HeLa cells. RFab fragments retrieved from peptide–BSA panning showed no reactivity to HeLa cells. Their variable-region sequences were different from HBJ127. However, rFab fragments retrieved from HeLa cell panning showed reactivity to CD98 by indirect immunofluorescence and immunoprecipitation. Moreover, they were structurally almost identical to HBJ127. Although the immunogenicity of epitope peptides may be insufficient for induction of expected antitumor activity in vivo, we used antibody phage display to show that IgG antibodies almost identical to HBJ127 were an undetectable population in epitope peptide-induced immune sera. (Cancer Sci 2009; 100: 126–131)
CD98 (GP125) is a heterodimeric protein with a relative molecular mass of 125 kDa, consisting of a glycosylated 85-kDa heavy chain and a non-glycosylated 40-kDa light chain, which are disulfide linked.(1) Analyses of human CD98 cDNA have revealed that CD98 is a type II transmembrane glycoprotein(2) that is disulfide-linked to a non-glycosylated light chain of a member of the permease family.(3) CD98 was identified originally as a cell-surface antigen associated with lymphocyte activation defined by the 4F2 monoclonal antibody (mAb)(1) and is expressed in proliferating normal tissues(4) and in almost all tumor cells.(5) These findings suggest that CD98 is involved functionally in lymphocyte activation, cell proliferation, and malignant transformation. In fact, a mAb against CD98, termed HBJ127, which was raised originally against T24 human bladder cancer(5) was found to inhibit tumor cell growth(6) and lymphocyte proliferation.(7) Furthermore, CD98 heavy chain cDNA-transfected murine fibroblasts showed various malignant phenotypes.(8) Because CD98 expression is detected on almost all cycling cells in both normal and tumor tissues irrespective of the organ of origin, CD98 may be a favorable target for T- or B-cell immune response against tumor cells in cancer patients.
Antibody medicines are used widely for treatment of tumors, autoimmune disorders, and infectious diseases. Antibody medicines such as trastuzumab(9) and rituximab(10) have been used for the treatment of c-erb-B2-positive breast cancers and CD20-positive B-cell lymphoma, respectively. As for rheumatoid arthritis therapy, the anti-tumor necrosis factor (TNF)α mAb infliximab is used for treatment in methotrexate-unresponsive patients.(11) Although antibody medicines show excellent efficacy in the therapy of such diseases, they have some disadvantages, for example, repeated injection may be necessary for maintaining efficacy, or there may be anxieties about anaphylaxis, decreased effectiveness due to the anti-antibody response in the host, or even the very high cost of these medicines.
It is expected that new medicines with equivalent efficacy to antibody medicines but without these drawbacks will be developed. One candidate is the peptide vaccine prepared based on the epitope sequence of antibody medicines or antibodies with biological activities. Induction of protective immunity would be expected by active immunization of the host with epitope vaccines. Epitope-mimicking peptides obtained using a phage display peptide library from cetuximab,(12) rituximab,(13) trastuzumab,(14) and anti-GD2 mAb(15–17) successfully induced protective immunity in the hosts. We have identified the epitope of the tumor-suppressive anti-CD98 monoclonal antibody HBJ127 using a phage display peptide library.(18) In the present study, to evaluate the usefulness of active immunization of HBJ127-derived epitope peptides for induction of antitumor immunity, we characterized the sera from immunized mice, constructed a Fab antibody phage display library from spleen cells of immunized mice, and then characterized recombinant Fab (rFab) fragments retrieved from panning against epitope peptide–BSA conjugates or CD98-positive HeLa cells.
Materials and Methods
Preparation of peptide–carrier protein conjugates. Peptides containing sequences of the predicted HBJ127 epitope (LLHGDFHAFSAG-GGGC, where GGGC is the linker and the predicted epitope sequence is in italics), epitope-related peptides (LLHGDFH-GGGC, FHAFSAG-GGGC), and control unrelated peptide (AHNQVRQVPLQR-GGGC) were synthesized chemically (Torey Research Center, Tokyo, Japan). Peptide–carrier protein conjugates were prepared by reaction of the cysteine residue at the C-terminal end of the peptide and maleimide-activated carrier proteins (Imject maleimide-activated carrier proteins; Pierce, Rockford, IL, USA). The protein content of peptide–carrier protein conjugates was determined by BCA Protein Assay (Pierce), and used for immunization (KLH conjugates) and enzyme-linked immunosorbent assays (ELISA) (BSA conjugates).
Immunization of mice. Female BALB/c mice (6–8 weeks old) (Japan SLC, Hamamatsu, Japan) were immunized intraperitoneally and subcutaneously with epitope peptide–KLH conjugates (50 µg protein/mouse) in saline with Freund's complete adjuvant (Difco Laboratories, Detroit, MI, USA). After 2 and 4 weeks, the same immunization procedure using incomplete adjuvant (Difco Laboratories) was repeated. A third immunization procedure was carried out 2 weeks later, after which elevated serum antibody titer was confirmed and the mice received a booster intravenous injection of 30 µg of epitope peptide–KLH in saline without adjuvant. Three days after the final immunization, the mice were killed by cutting the jugular vein and venous blood was collected. Spleens were also obtained from immunized mice, and were used to make the antibody library.
Library construction. The construction of a Fab phage-display combinatorial library from immunized mouse spleen was carried out as described previously.(19) Briefly, total RNA from spleen cells (1 × 107) was extracted using the RNeasy Mini Kit (Qiagen, Tokyo, Japan). First-strand cDNA was then synthesized with a pd(T)15–18 primer using the a first-strand cDNA synthesis kit for reverse transcription (RT)-polymerase chain reaction (PCR) (AMV; Roche Diagnostics, Tokyo, Japan). PCR amplification of the heavy chain Fd region and whole κ light chain was carried out using the iCycler thermal cycler (Bio-Rad, Tokyo, Japan) with the family-specific variable region and isotype-specific constant-region primers,(20) designed based on the Kabat database.(21) Gel-purified PCR products were double digested with restriction enzymes (SpeI and XhoI for heavy chain and SacI and XbaI for light chain; all enzymes were purchased from Roche Diagnostics) and then ligated sequentially into the phage-display vector pComb3 (provided by Dr Dennis R. Burton, The Scripps Research Institute, La Jolla, CA, USA).(22) The constructed combinatorial library was electroporated into XL1-Blue cells (Stratagene, La Jolla, CA, USA) using a MicroPulser (Bio-Rad, Richmond, CA, USA). The transformed library was grown in super optimal broth with catabolite repression (SOC) medium for 1 h at 37°C and subsequently in superbroth medium containing 10 µg/mL tetracycline and 20 µg/mL carbenicillin for an additional hour after increasing carbenicillin to 50 µg/mL. Phage particle assembly was initiated by addition of VCSM13 helper phage (1012 plaque-forming units). After 2 h of additional culture, kanamycin was added (final concentration 70 µg/mL), and the culture was grown overnight at 30°C.
Library panning. For cell panning, the CD98-positive human cervix cancer cell line HeLa(23) was used as the target of library panning. HeLa cells (1 × 106) were transferred to a microfuge tube and packed by centrifugation. After removal of the supernatant, 500 µL of phage solution was added and mixed by inverting the tube several times. The phage–cell mixture was incubated for 2 h on ice with occasional mixing. The phage solution was subsequently removed and cells were washed with ice-cold phosphate-buffered saline (PBS). Bound phage were eluted with glycine-HCl (pH 2.2), and neutralized with 2 mol/L Tris base. Eluted phage were reamplified for the next round of panning as described previously.(22) The panning procedure was repeated four times. Phagemid DNA of the last round of panning was digested with NheI and SpeI to excise the cpIII gene, and reconstructed phagemid was used to obtain soluble Fab fragments. For epitope peptide panning, the wells of 3690 EIA/RIA plates (Costar Corning Life Sciences, Tokyo, Japan) were coated with peptide–BSA conjugates (20 µg/mL in PBS) and incubated overnight at 4°C. The wells were then blocked with 3% BSA in PBS for 1 h at 37°C. After discarding the blocking solution, the wells were treated with phage solution (50 µL/well) and incubated for 2 h at 37°C. The wells were washed with PBS containing 0.05% Tween-20 (T-PBS). Peptide–BSA-bound phage were eluted with glycine-HCl (pH 2.2) and neutralized with 2 mol/L Tris base. The following steps were the same as those described in the cell panning section.
Screening of rFab clones. RFab cultures were grown from single colonies overnight at 30°C, and bacterial lysates was prepared by freeze-thawing.(22) Screening of rFab clones obtained by cell panning was carried out by indirect immunofluorescence (IIF). Packed HeLa cells (2 × 105) were incubated on ice with 20 µL rFab lysates (1:1 diluted with 1% BSA-PBS) for 1 h. Cells were washed three times with ice-cold PBS by centrifugation for 5 min at 200g). Packed cells were treated with 30 µL fluorescein isothiocyanate-labeled rabbit antihuman IgG F(ab′)2 (1:300 diluted in 5% normal rabbit serum in PBS). After 1 h incubation on ice, cells were washed three times as above. The reactivity of rFab clones was examined by confocal laser microscopy (LSM510; Carl Zeiss Japan, Tokyo, Japan). Screening of rFab clones obtained by epitope peptide panning was carried out by direct ELISA. The wells of 3690 EIA/RIA plates were coated with peptide–BSA conjugates and incubated overnight at 4°C. The wells were then blocked with 1% BSA in PBS for 1 h at 37°C. After discarding the blocking solution, the wells were treated successively with rFab lysates, alkaline phosphatase (ALP)-labeled rabbit antimouse IgG F(ab′)2 (Jackson ImmunoResearch, West Grove, PA, USA), and p-nitrophenyl phosphate (4-nitrophenyl phosphate disodium salt hexahydrate 5 mg tablet; Sigma-Aldrich, Tokyo, Japan) in 1 mol/L diethanolamine buffer (pH 9.8). The absorbance of the resultant p-nitrophenol in each well was determined at 405 nm using a Model 680 Microplate Reader (Bio-Rad, Tokyo, Japan).
DNA sequencing. The DNA sequence encoding the variable regions of the rFab fragment was sequenced using a PRISM 310 genetic analyzer (Applied Biosystems, Foster City, CA, USA) with the BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems). The primer used to determine the heavy chain sequence was SeqT3 (5′-ATT AAC CCT CAC TAA AG-3′), hybridizing to the negative strand. The primer used for the light chain was KEF (5′-GAA TTC TAA ACT AGC TAG TTC G-3′), hybridizing to the negative strand. Comparison with the reported immunoglobulin germline sequences from GenBank European Molecular Biology Laboratory (EMBL) DNA Data Bank of Japan (DDBJ) was carried out using NCBI's Ig BLAST analysis and DNAPLOT analysis.
Immunoprecipitation. Cellular antigens that were reactive with the rFab fragment were evaluated by immunoprecipitation using HeLa cells. Biotin-labeled HeLa cells (5 × 106) were treated with cell lysis buffer (PBS containing 1% Nonidet P40 (NP)-40 [Nakarai, Tokyo, Japan], 1:100 diluted protease inhibitor cocktail [Nakarai, Tokyo, Japan], 1:5000 diluted Benzonase [Merk], and 0.1% NaN3) by incubation on ice for 1 h with occasional mixing. Cell lysates was centrifuged for 15 min at 14 800g at 4°C, and cleared supernatant was used for immunoprecipitation. Preclear of biotinylated lysates was carried out by successive treatment with non-immune mouse IgG Fab fragment (Jackson ImmunoResearch), rabbit antimouse IgG F(ab′)2 (Jackson ImmunoResearch), and protein G-sepharose 4FF (50% suspension in PBS) (Amersham, GE Healthcare, Tokyo, Japan). Precleared supernatant was then treated successively with 5 µg rFab fragment, 5 µg rabbit antimouse IgG F(ab′)2, and protein G-sepharose 4FF. Immunoprecipitated materials were eluted from packed beads with 1× sodium dodecylsulfate–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer by heating for 5 min at 95°C. Eluted proteins were separated by SDS-PAGE (12.5% separating gel) followed by transfer to polyvinyldine difluoride (PVDF) membrane (Immobilon-N, Millipore, Tokyo, Japan) using a semi-dry blotter (Transblot SD Cell; Bio-Rad). Protein-blotted membrane was blocked with Block Ace (DS-Pharma Biomedical, Tokyo, Japan), and treated with ALP-ABC (Vector Laboratories, Burlingame, CA, USA). After five washes with T-PBS, proteins on the membrane were visualized by treatment with 5-bromo-4-chloro-3-indolyl-phosphate/nitro blue terazolium (BCIP/NBT)-based substrate solution (AP conjugate substrate kit; Bio-Rad).
Cell growth inhibition assay. HeLa cells were seeded in triplicate into flat-bottomed 96-well culture plates at 1 × 104 cells per well. The rFab fragment crosslinked with the rabbit antimouse IgG F(ab′)2 was incubated with cells for 72 h at 37°C. The rFab fragment was used at a final concentration of 50 µg/mL, and an equimolar concentration of HBJ127 was used as a positive control. The cell growth inhibitory effect of the rFab fragment was evaluated using the alamar blue assay (Cosmo Bio, Tokyo, Japan) according to the manufacturer's instructions.
Preparation of epitope peptide–KLH conjugates and immunization to mice. We previously identified the epitope of HBJ127 using a phage display random peptide library.(18) We prepared four peptides as shown in Table 1: #1 is the dodecapeptide including the predicted epitope sequence; #2 is the N-terminal heptapeptide of #1; #3 is the C-terminal heptapeptide of #1; and #4 is the dodecapeptide with the sequence unrelated to of #1. These peptides were conjugated with maleimide-activated KLH via the cysteine residue of the linker. The mice were immunized with epitope peptide–KLH at days 0, 14, and 28. We collected blood samples from immunized mice at day 42, and tested the reactivity against the corresponding epitope peptide–KLH conjugate by direct ELISA. All mice showed a high titer (>1:10 000) against epitope peptide–KLH. Three days after intravenous injection of the epitope peptide–KLH conjugate, blood samples and spleens were collected from the immunized mice. Sera separated from blood samples were further characterized, and dissected spleen cells were used for construction of the antibody Fab library.
Table 1. Sequence of peptides designed based on the epitope recognized by HBJ127
Predicted epitope sequences are in italics.
Characterization of antisera from epitope peptide–KLH-immunized mice. We determined the reactivity of antisera against the corresponding epitope peptide–KLH and epitope peptide–BSA by direct ELISA. As shown in Table 2, the titer of antisera was between 1:1000 and 1:10 000. These results indicate that an immune response occurred not only against the carrier protein but also against hapten in immunized mice. We determined the specificity of antisera by testing the reactivity against four peptide–BSA conjugates having different epitope sequence. As shown in Figure 1, antisera against #1-KLH (#1S) showed crossreactivity to #4-BSA. Antisera against #3-KLH (#3S) showed crossreactivity to #2-BSA and #4-BSA. On the other hand, antisera against #2-KLH (#2S) and #4-KLH (#4S) were found to be sequence specific because they reacted only with epitope peptide–BSA, which included a sequence that was the same as the immunogen. We determined the reactivity of antisera against recombinant human CD98 heavy chain by direct ELISA. #1S and #2S showed reactivity to recombinant human CD98 heavy chain in a concentration-dependent manner, although #3S showed no reactivity at any determined concentrations (data not shown). Moreover, none of the antisera reacted either with CD98-positive HeLa cells or with human CD98 heavy chain-transfected mouse NIH3T3 cells(8) (data not shown). These results suggest that antibodies reactive with native CD98 were not induced or were induced but presented as an undetectable population in immune sera. To confirm this hypothesis, we constructed a Fab phage display library from spleen cells of immunized mice and characterized rFab clones retrieved from panning against epitope peptide–BSA or live HeLa cells.
Table 2. Serum titer against corresponding peptide–KLH and peptide–BSA conjugates
Construction and panning of Fab phage display library from epitope peptide–KLH-immunized mice. We constructed a Fab phage display library from #1- and #3-KLH immunized mice spleen cells (libraries were denoted as #1Lib and #3Lib, respectively), as both peptides contained the predicted epitope sequence recognized by HBJ127.(18) Heavy chain Fd and light chain fragments of IgG1, κ were amplified by RT-PCR from total RNA of spleen cells, and cloned into the phage display vector pComb3.(22) The obtained sizes of #1Lib and #3Lib were 8.5 × 107 and 1.9 × 108 colony forming unit, respectively. The phage display libraries were panned against the corresponding peptide–BSA conjugate. Four rounds of panning produced up to 63- and 17-fold amplification of #1Lib and #3Lib, respectively, in eluted phage. Forty clones of rFab from each library tested for reactivity against corresponding peptide–BSA conjugates by direct ELISA. All tested clones reacted with corresponding peptide–BSA. We then carried out panning of the same library against live HeLa cells. Four rounds of panning produced up to 12- and 4-fold amplification of #1Lib and #3Lib, respectively, in eluted phage. Ten clones of rFab from each library tested for reactivity against live HeLa cells by IIF. Nine out of 10 clones from each library showed reactivity to HeLa cells. RFab clones stained the cell surface of HeLa cells as well as HBJ127.
Characterization of rFab clones obtained from panning against epitope peptide or HeLa cells. RFab clones obtained by epitope peptide panning were designated ‘EP’ and those obtained by HeLa cell panning were designated ‘CP’. We tested the identity of positive clones by BstNI fingerprinting(24) and selected CP1-5 and CP1-10 from #1Lib and CP3-10, EP3-1, EP3-6, and EP3-31 from #3Lib for further characterization. All CP clones reacted with live HeLa cells by IIF (Fig. 2). EP3-6 and EP3-31 reacted with #1- and #3-BSA but not with #2- and #4-BSA, respectively. However, EP3-1 reacted with #2-BSA as strongly as with #1- and #3-BSA but not with #4-BSA (Fig. 3). None of the EP clones showed reactivity to live HeLa cells by IIF, whereas none of the CP clones reacted with any peptide–BSA conjugates by direct ELISA (data not shown). We then tested the reactivity of all EP and CP clones against CD98 in HeLa cell lysates by capture sandwich ELISA (Fig. 4). CP clones but not EP clones reacted with HeLa lysates. Only CP clones reacted with soluble recombinant CD98 heavy chain–green fluorescent protein (GFP) fusion protein by direct ELISA (data not shown). A definitive 125-kDa band of human CD98 was detected by immunoprecipitation of biotinylated HeLa cell lysates with CP clones but not with EP clones (Fig. 5). Moreover, CP clones showed a cell growth inhibitory effect against HeLa cells in vitro (Fig. 6).
Molecular structural analysis of EP and CP clones. To elucidate the differences in reactivity between EP and CP clones, we sequenced and compared the variable heavy (VH) and variable light (VL) domains of the EP and CP clones. These sequence data have been submitted to the DDBJ/EMBL/GenBank databases under accession numbers AB447359–70. The derived germline sequence and deduced amino acid sequence of EP and CP clones are shown in Figure 7. The VH of all CP clones were derived from VHQ 52, a5, 13, and the VL were derived from IgVk8–30. The VH (AB056115) and VL (AB056116) of HBJ127 are also derived from the same germline. The VH and VL of CP clones were found to be almost identical to those of HBJ127 (Table 3). However, the VH of EP clones were derived from two different germlines, and the VL were derived from three distinct germlines (Fig. 7). Homology of the VH and VL of the EP clones with those of HBJ127 was quite low compared to the CP clones (Table 3). From these results, we first confirmed that IgG antibodies almost identical to HBJ127 were induced but were presented as an undetectable population in epitope peptide-immunized sera.
Table 3. Comparison of gene usage and structural homologies of the variable heavy (VH) and variable light (VL) domains of the epitope peptide panning (EP) and cell panning (CP) clones
VH gene usage
VL gene usage
Homology to germline (%)
Homology to germline (%)
In the present study, we describe the evaluation of epitope peptide as a potential inducer of IgG antibodies against the CD98 oncoprotein. Almost all previous reports have described epitope-mimicking peptide (mimotope) immunization.(12–16,25) We used an epitope peptide, not a mimotope peptide, because an antibody repertoire that was not only functionally but also structurally equivalent to the parent antibody would be induced by epitope peptide immunization. An antihapten (epitope peptide) response occurred in immunized mice because antisera reacted with peptide–BSA conjugates as well as peptide–KLH conjugates (Table 2). No reactivity of antisera against CD98-positive HeLa cells or recombinant CD98 heavy chain (data not shown) indicates that epitope peptides failed to induce antibodies reactive with native CD98, or these were induced but undetectable in immune sera. In our previous study about HBJ127 epitope analysis using a phage display peptide library,(18) phage clones bearing epitope-unmatched sequences (mimotopes) (HPMHFPS, 11/30; YPRWIQP, 6/30; SVFWMIP, 6/30) appeared frequently. However, only one phage clone bearing an epitope-matched sequence (HHYAFSV, 1/30) appeared. HBJ127 reacts with mimotope sequences in preference to epitope sequence. The epitope peptide failed to induce the anti-CD98 antibody repertoire as a major population in immune sera. These observations indicate that epitope peptide is poorly immunogenic compared with mimotope peptides or full-length protein.
We constructed an IgG1, κ Fab-phage display library from spleen cells of immunized mice to confirm whether epitope peptides induced antibodies against native CD98 or not. RFab retrieved from panning against epitope peptide–BSA showed no reactivity to CD98 and were structurally different to HBJ127. However, those retrieved from panning against live HeLa cells showed reactivity to CD98 and were structurally almost identical to HBJ127. Moreover, CD98-reactive rFab showed no reactivity to any epitope peptides. These results indicate that epitope peptide immunization induces both antipeptide and anti-CD98 antibodies although the latter is too small to detect in immune sera. Because unconstrained peptide was used as a hapten of immunogen, it might be difficult to induce an antibody repertoire reactive with the native CD98 molecule. Although unconstrained peptide has any possible conformation, the native protein molecule has a strict and stable conformation of its own. If the conformation of the epitope peptide was coincidentally the same as that of the epitope region of the native peptide then it may induce CD98-reactive antibodies in vivo. For induction of efficient anti-CD98 activity by epitope peptide immunization, we should determine the three-dimensional structure of the epitope region of the natural CD98 antigen and design a structurally mimicking peptide. This approach may be ideal but is actually very difficult and time consuming. The mimotope peptide approach shown above is based on the molecular mimicry theory,(26) and seems to be easier and more practical.
Because a peptide is too small to be sufficient for induction of immunity by itself, the hapten-carrier system is used generally for the induction of immunity against low molecular weight compounds (hapten). In the present study, we used KLH as a carrier protein, and successfully induced immunity not only against KLH but also against the peptides. For the clinical application of epitope peptide immunization, we should avoid unexpected immune reactions induced by carrier proteins. Although proteins of human origin would be suitable as carriers, they may be insufficient for induction of strong immunity caused by so-called immune tolerance.(27) Using tetanus toxoid as a carrier protein, Riemer et al. successfully induced anti-Her2/neu immunity by immunization of a mimotope of trastuzumab.(14) We can use poly-l-lysine as a backbone for peptide immunization.(28) We can also use a shuttle vector with a glutathione S-transferase (GST) expression system(29) or liposomes as a nanocarrier.(30) Development of suitable carriers for inducing strong antipeptide immunity with minimum side effects is a critical point for the clinical application of epitope peptide-based tumor immunotherapy.
In conclusion, epitope peptides were insufficient for inducing anti-CD98 activity in immune sera, although they induced an anti-CD98 antibody repertoire as a minor population. Induced antibodies were not only functionally but also structurally equivalent to the parent antibody. This may indicate that our previously predicted epitope on CD98 was correct.(18) For efficient induction of serum anti-CD98 immunity, mimotope peptides instead of epitope peptides are promising as haptens of immunogens. Evaluation of HBJ127 mimotope peptides for possible induction of serum anti-CD98 and antitumor activities is now in progress.
This work was supported in part by Grants-in-Aid for Scientific Research (C) from The Japan Society for Promotion of Science (17590467, 20590063). The authors thank Dr Dennis R. Burton of The Scripps Research Institute for providing the pComb3 vector.