Photobodies: Light‐Activatable Single‐Domain Antibody Fragments

Abstract Photocaged antibody fragments, termed photobodies, have been developed that are impaired in their antigen‐binding capacity and can be activated by irradiation with UV light (365 nm). This rational design concept builds on the selective photocaging of a single tyrosine in a nanobody (a single‐domain antibody fragment). Tyrosine is a frequently occurring residue in central positions of the paratope region. o‐Nitrobenzyl‐protected tyrosine variants were incorporated into four nanobodies, including examples directed against EGFR and HER2, and photodeprotection restores the native sequence. An anti‐GFP photobody exhibited an at least 10 000‐fold impaired binding affinity before photodeprotection compared with the parent nanobody. A bispecific nanobody–photobody fusion protein was generated to trigger protein heterodimerization by light. Photoactivatable antibodies are expected to become versatile protein reagents and to enable novel approaches in diagnostic and therapeutic applications.

Abstract: Photocaged antibody fragments,t ermed photobodies,h ave been developed that are impaired in their antigenbinding capacity and can be activated by irradiation with UV light (365 nm). This rational design concept builds on the selective photocaging of as ingle tyrosine in an anobody (a single-domain antibody fragment). Tyrosine is af requently occurring residue in central positions of the paratope region. o-Nitrobenzyl-protected tyrosine variants were incorporated into four nanobodies,i ncluding examples directed against EGFR and HER2, and photodeprotection restores the native sequence.Ananti-GFP photobody exhibited an at least 10 000fold impaired binding affinity before photodeprotection compared with the parent nanobody.Abispecific nanobodyphotobody fusion protein was generated to trigger protein heterodimerization by light. Photoactivatable antibodies are expected to become versatile protein reagents and to enable novel approaches in diagnostic and therapeutic applications.
Antibodies recognize their antigens with high affinity and selectivity.T hey are indispensable protein reagents in basic research and biomedicine,f or example to detect and enrich their binding partners and for diagnostic and therapeutic applications.Bispecific antibodies contain two different paratope sites for antigen binding and trigger the heterodimerization of their respective epitopes. [1] Antibody binding is mediated through extended complementary interaction surfaces with the epitope.T he three complementarity-determining regions (CDRs) in the variable domains of the immunoglobulin scaffold represent the hot spots of sequence diversity and structural malleability to provide the most important contributions to binding.
We envisaged the rational design of novel protein reagents in which the binding of the antibody variable domain(s) to its cognate epitope is rendered light dependent. Photoactivatable and photoswitchable molecules are exquisite tools to study biological systems with high temporal and spatial resolution, [2] to develop new biomaterials, [3] and they have great potential for therapeutic purposes in the field of photopharmacology. [4] Previous examples of light-controlled molecules covered small molecules and ligands,enzymes,ion channels,v arious proteins,n ucleic acids,l ipids,a nd others. [5] However,t he design of light-dependent antibodies,w ith ad irect and defined activation of the antibody-antigen interaction, has not been reported yet. [6] Previous efforts to modulate this binding event in al ight-dependent manner include chemical photocaging of the antigen, [7] global, unspecific and uncharacterized coating of the antibody with up to 50 chemical photocaging groups per protein, [8] and photoinduced cleavage of synthetically fused epitopes that act as an oncovalent, bivalent inhibitor. [9] We focused on single-domain antibody fragments,a lso referred to as nanobodies,which are the variable domains of heavy-chain-only antibodies of camelids (V H H). [10] Nanobodies bind their epitope monovalently and typically with affinities in the low nanomolar or sometimes even picomolar range.T hey are highly stable in monomeric form and can be efficiently expressed in Escherichia coli. With these and other favorable properties,n anobodies recently have gained considerable attention. They are being explored for al arge variety of applications,mostly in basic research and diagnostics,b ut they are also considered attractive in therapy. [10a,11] Our concept to rationally design light-activatable nanobodies,t ermed photobodies,i sb ased on the idea of structurally perturbing and thereby impairing the nanobody-antigen interaction by the introduction of as terically demanding photocaging group at ac entral and neuralgic position in the paratope region of the nanobody.L ight-induced cleavage of the cage group would furnish the native nanobody sequence to reestablish binding (Scheme 1). We observed that the CDR loops of nanobodies show ah igh frequency of tyrosine residues.T he same holds true for several additional residues on the nanobody core domain that are often implicated in antigen binding. [12] We further inspected structures of nanobody-antigen complexes deposited in the Protein Data Bank and found that these tyrosines are often part of the nanobody-antigen interaction interface.T herefore,w es elected tyrosine side chains for our photocaging approach.
As the first nanobody test case,w es elected an anti-GFP nanobody (GFP-enhancer;P DB:3 K1K). [13] Thes tructure with its antigen reveals aT yr residue (Y37) located in the center of the extended protein-protein interaction interface ( Figure 1A,B). Thes ide chain of Y37 is in nearly perpendicular orientation to the b-barrel structure of GFP.Remarkably, the Y37 side chain is almost completely engulfed by other residues of the binding interface in the wild-type nanobody-GFP complex. We expected that ap hotolabile protecting group would not only disrupt hydrogen bonding of the hydroxyl group,b ut also prevent the formation of multiple interactions in the vicinity of Y37 through its steric demand and therefore have ad ramatic effect on the binding affinity.
Forphotocaged Ty rvariants,weturned our attention to onitrobenzyl-protected Ty r( ONBY,c ompound 1)a nd its methylenedioxy-derivative nitropiperonyl tyrosine (NPY, 2; Figure 1C). Theg enetic incorporation of 1 into proteins by the amber stop codon suppression technology has been previously reported with a Methanococcus jannaschii antityrosyl tRNAs ynthetase (TyrRS, MjTy rRS) mutant selected for ONBY and its cognate MjtRNAf or expression in E. coli. [14] Both 1 and 2 have been encoded in mammalian cells using am utant of pyrrolysine-tRNAs ynthetase. [15] In contrast, to our knowledge,the incorporation of 2 has not yet been shown in E. coli using the MjTyrRS system. [15a,b] To test our concept, we produced both the wild-type anti-GFP nanobody (construct 3)a nd its two caged photobody variants with Y37ONBY (4)a nd Y37NPY (5). Forthe latter two,t he orthogonal pair of the MjTy rRS and MjtRNAf or incorporation of 1 [14] was co-expressed in the E. coli production host. Amino acids 1 or 2 were added to the growth medium at 1mm concentration. NPY (2)w as found to be accepted as a MjTy rRS(ONBY) substrate,albeit with slightly reduced expression levels.F igure 2A shows the purified nanobodies.C onsistent with our previous findings on periplasmic expression, [16] mass spectrometry (MS) analysis of the purified protein fractions suggested the presence of only negligible contents of the reduced o-aminobenzyl forms (OABY and APY; 5% and 3%,respectively;Supporting Information, Figures S1 A,B and S2). No prematurely deprotected photobody with the mass of the wild-type nanobody could be detected. Protein stabilities showed very similar melting temperatures for the wild-type nanobody and the ONBY-photobody (Supporting Information, Figure S1 C), consistent with the idea that photocaging of asurface-exposed tyrosine residue has an egligible impact on protein folding. Photodeprotection (l = 365 nm) occurred virtually quantitatively and within afew seconds ( Figure 2C,D and Supporting Information, Figure S2).
We next addressed the crucial question on the intended loss of binding affinity of the photobodies and the reactivation by light. We performed microscale thermophoresis (MST) experiments with superfolder GFP (sfGFP) as the fluorescent antigen and added the photobody in ad ilution series.T oour delight, we could not reach saturation in binding for the ONBY-photobody (4), indicating an estimated dissociation constant K d of ! 10 mm ( Figure 2E and Support-  ing Information, Figure S3 A). This is at least an approximately 10 000-fold impairment compared to the binding constant reported for the wild-type nanobody (1; K d = 1.1 nm for sfGFP [17] and 0.6 nm for GFP [13] ). Unexpectedly,the NPYcaged photobody (5)w as less impaired (approximately 420fold), with a K d of 0.46 AE 0.02 mm (Supporting Information, Figure S3 B). Ap ossible explanation for this finding is that other binding contributions from the NPY protecting group have accidentally and partially compensated for the impairment caused by the steric effect in this particular case.W e therefore abandoned the NPY variant in the subsequent study and focused on the ONBY-photobody (4), although the sterically more demanding NPY is likely to be useful or even superior in other cases.T oa ssay the photodeprotected photobodies for reconstituted antigen-binding capacity,w e realized the limited sensitivity of our MST assay to measure with sfGFP concentrations below 10 nm,w hich is 10-fold higher than the expected binding affinity.N evertheless,t he assay sufficed to show that the ONBY-photobody (4) regained binding affinity after photodeprotection (Supporting Information, Figure S3 A).
To accurately measure the binding affinities,wedisplayed the nanobodies on the surface of E. coli cells using the AIDA autodisplay system [18] and measured antigen binding (sfGFP) by flow cytometry (Figure 3a nd Supporting Information, Figure S4). E. coli cells presenting the wild-type anti-GFP nanobody bound sfGFP with a K d of 3.0 nm,s uggesting the nanobody was displayed in fully functional form. To display the ONBY-photobody,w ec ombined the AIDAa utodisplay with the amber stop codon suppression technique. [19] Prior to irradiation of the presenting cells,nosfGFP binding could be detected for concentrations up to about 10 mm,consistent with the results from our MST assay.Wethen added sfGFP to cells that had been irradiated for 45 s( l = 365 nm) and could determine a K d of 0.90 AE 0.03 nm ( Figure 3E), nicely fitting with the positive control and consistent with the previously reported binding constant of the wild-type anti-GFPenhancer nanobody. [13,17] Together, these results demonstrated that the photodeprotected photobody regains its full antigen binding affinity.
We next aimed to demonstrate the potential of our novel anti-GFP photobody for light-controlled protein dimerization in ac ellular assay.W ee nvisioned ab ispecific nanobody in which only one binding site is photocaged. We devised such nanobody-photobody fusion (construct 6)c onsisting of the EgA1 nanobody,w hich is directed against domain 3o ft he epidermal growth factor receptor (EGFR), and the anti-GFP photobody,and obtained it by expression in E. coli (Figure 4). We transiently transfected HeLa cells with the transmembrane and extracellular domains of EGFR fused to the redfluorescent protein mCherry.W ea dded the nanobodyphotobody fusion 6 (10 nm)t ot he HeLa cells,a llowed for binding of the anti-EGFR nanobody portion of 6,a nd then washed the cells.T he cells were irradiated (20 s, l = 365 nm) to activate the photobody portion of 6,w hereas in control samples no irradiation was performed. We then added sfGFP (10 nm)a nd again washed the cells.V isualization of the cells by confocal fluorescent microscopy showed that binding of sfGFP could only be detected on transfected cells and with the photoactivated nanobody-photobody ( Figure 4D;s ee Figure S5 in the Supporting Information for the control experiment with the non-caged bivalent nanobody construct 7). Together, these results demonstrate that ap hotobody can be used in ac ellular context and to design lightdependent protein dimerizers based on abispecific antibody.
Finally,w es ought to generate more examples of our photobody design concept, including photobodies with potential therapeutic relevance.T he aforementioned EgA1 nanobody binds to EGFR, which is upregulated or mutated in certain tumors.T wo tyrosines in the nanobody,T yr32 (in the CDR1 loop) and Ty r119 (at the end of the CDR3 loop), appeared highly promising for photocaging based on structural considerations (PDB:4 KRO; Supporting Information, Figure S6). [20] We prepared abispecific anti-EGFR-anti-GFP photobody-nanobody fusion (8), similar to construct 6,h owever,t his time with ONBY in the anti-EGFR nanobody at position Ty r119 (Supporting Information, Figure S6). Indeed, the photocaged photobody-nanobody 8 did not bind to HeLa cells transfected with an EGFR-mCherry construct, but after photodeprotection, efficient binding to the transmembrane receptor could be monitored using confocal fluorescent microscopy (Supporting Information, Figure S7). We next

Angewandte Chemie
Communications 1508 www.angewandte.org selected the 2Rs15d nanobody,w hich binds domain 1o f human epidermal growth factor receptor 2(HER2). HER2 is overexpressed in several types of breast cancer. [21] We tested Ty r37 for our approach based on the crystal structure with the antigen (PDB:5MY6; [21] Supporting Information, Figure S8). AY 37ONBY-photobody-sfGFP fusion (9)w as produced and labeled with Cy5. Prior to photodeprotection, the photobody (9)w as unable to bind at detectable levels to BT-474 cells overproducing the HER2 receptor;h owever, it specif-ically bound to these cells following light-activation (Supporting Information, Figure S9). Finally,w ec hose another anti-GFP nanobody (GFP-minimizer;P DB:3 G9A). [13] We identified Ty r113 in the CDR3 loop,w hich contacts the antigen only in aside-on orientation, in contrast to the mostly pointed orientations found in the other examples presented (Supporting Information, Figure S10). TheY 113ONBY-photobody (10)e xhibited a K d = 2.28 AE 0.03 mm for sfGFP,w hich is approximately 1000-fold higher than the affinity reported for the wild-type nanobody ( Supporting Information, Figure S11 D). [13] Photodeprotection reconstituted the binding activity of the photobody for the purified sfGFP antigen (Supporting Information, Figure S11) as well as binding of aC y5-labeled photobody 10 to GFP expressed as af usion protein on the surface of HeLa cells (Supporting Information, Figure S12). Together,t hese data demonstrate the broad applicability of our photobody design concept and the versatile utility of the new protein reagents.
In conclusion, we report photoactivatable antibody molecules,t ermed photobodies,t hat were rationally designed with as ingle photocaging group in the binding region. The generation of four such photobodies and the frequent occurrence of suitable tyrosines in key positions of nanobody paratopes suggests this approach to be quite general. We expect this design concept to be extendable to other antibody and antibody-like formats,s uch as full-length IgG,s cFv, diabodies,ormonobodies,which similarly contain tyrosine as af requent residue in their paratope regions, [22] using either tyrosine or other side chains as caged entities.P hotoactivatable antibody-like molecules will find many applications derived from the binding,t argeting, dimerization, and oligomerization properties of their uncaged parent proteins. Caging groups that are sensitive to longer wavelengths will be desirable in the future to enable reduced phototoxicity and deeper penetration of biological material. We have already shown the incorporation of NPY (2)w ith ar ed-shifted absorption maximum compared to ONBY (1). Ultimately, even further red-shifted or two-photon decaging groups [4b,23] would open new avenues for applications in cell biology and live organisms,oreven in patients.F or example,weenvision the combination of photobodies with concepts from antibody-drug conjugates or CAR-T cells [10c, 11a] to achieve spatial and temporal control for these therapeutic strategies.