A bi-paratopic anti-EGFR nanobody efficiently inhibits solid tumour growth

Abstract

The epidermal growth factor receptor (EGFR) has been shown to be a valid cancer target for antibody-based therapy. At present, several anti-EGFR monoclonal antibodies (mAbs) have been successfully used, among which cetuximab and matuzumab. X-ray crystallography data show that these antibodies bind to different epitopes on the ecto-domain of EGFR, providing a rationale for the combined use of these two antibody specificities. We have previously reported on the successful isolation of antagonistic anti-EGFR nanobodies. In the present study, we aimed to improve on these molecules by combining nanobodies with specificities similar to both cetuximab and matuzumab into a single bi-paratopic molecule. Carefully designed phage nanobody selections resulted in two sets of nanobodies that specifically blocked the binding of either matuzumab or of cetuximab to EGFR and that did not compete for each others binding. A combination of nanobodies from both epitope groups into the bi-paratopic nanobody CONAN-1 was shown to block EGFR activation more efficiently than monovalent or bivalent (monospecific) nanobodies. In addition, this bi-paratopic nanobody potently inhibited EGF-dependent cell proliferation. Importantly, in an in vivo model of athymic mice bearing A431 xenografts, CONAN-1 inhibited tumour outgrowth with an almost similar potency as the whole mAb cetuximab, despite the fact that CONAN-1 is devoid of an Fc portion that could mediate immune effector functions. Compared to therapy using bivalent, mono-specific nanobodies, CONAN-1 was clearly more potent in tumour growth inhibition. These results show that the rational design of bi-paratopic nanobody-based anti-cancer therapeutics may yield potent lead molecules for further development.

Introduction

The epidermal growth factor receptor (EGFR) is a member of a family of four related receptor tyrosine kinases (RTK), named Her- or cErbB1, -2, -3 and -4. The EGFR has an extra-cellular domain which is composed of four sub-domains, two of which are involved in ligand binding and one of which is involved in homo- and hetero dimerisation ^{1, 2}, for review, see ³. EGFR integrates extracellular signals from a variety of ligands to yield diverse intra-cellular responses ^{4, 5}. The major signal transduction pathway activated by EGFR is composed of the Ras-mitogen-activated protein kinase (MAPK) mitogenic signalling cascade. Activation of this pathway is initiated by the recruitment of Grb2 to tyrosine-phosphorylated EGFR ^{6, 7}. This leads to activation of Ras through the Grb2-bound Ras-guanine nucleotide exchange factor (GEF) Son Of Sevenless (SOS). In addition, the PI3-kinase-Akt signal transduction pathway is also activated by EGFR, although this activation is much stronger in case there is co-expression of Her3 ^{8, 9}.

The EGFR is implicated in several human epithelial malignancies, most notably cancers of the breast, lung, colon, head and neck and brain ¹⁰. Activating mutations in the gene have been found, as well as over-expression of the receptor and of its ligands, giving rise to autocrine activation loops (for review, see ¹¹). This RTK has therefore been extensively used as target for cancer therapy. Both small molecule inhibitors targeting the receptor tyrosine kinase and monoclonal antibodies directed to the extra-cellular ligand binding domains have been developed and have shown hitherto several clinical successes, albeit mostly for a select group of patients ¹².

From the crystal structures of the Fab fragments of several therapeutic monoclonal antibodies in complex with the extra-cellular domain of EGFR ^13–15, much knowledge about the working mechanisms of these antibodies has been gathered over the years (for review, see ¹⁶). Most therapeutic antibodies target the ligand binding domain III of the EGFR. One antibody (mAb806) has been reported to recognise an unfolded region of the EGFR that is only exposed when cells either over-express EGFR or express a deletion mutant (de2-7 or vIII) of the receptor ^{17, 18}. The domain III-specific, therapeutically used antibodies, either occupy EGF contact residues directly (shown for cetuximab and panitumumab ^{13, 14}), or bind outside the EGF contact area and sterically inhibit the conformational change necessary for receptor activation (as was shown for matuzumab ¹⁵). The effects of the combined use of the chimeric mAb cetuximab (derived from murine mAb 225) and humanised mAb matuzumab (derived from the murine mAb 425) on EGFR signalling has recently been investigated and this combination was indeed shown to work well in EGFR inhibition ¹⁹.

We have previously reported the successful generation and use of camelid-derived single domain antibody fragments (termed VHH or Nanobody^*) directed to EGFR for the therapy of non-established tumours ²⁰. The single domain nature of these fragments allows for the combination of different nanobodies with different specificities in one molecule: a dual-specific (DS) or bi-paratopic nanobody ^{20, 21} (for review, see ²²). The aim of the current study was to obtain anti-EGFR nanobodies with improved therapeutic efficacy by synthesising bi-paratopic molecules that would combine the specificities of the two domain III specific anti-EGFR whole antibodies cetuximab and matuzumab. We obtained antagonistic nanobodies that competed for the binding of either cetuximab or matuzumab to the receptor. As expected from these epitope-mapping data, these nanobodies did not compete for each others' binding to EGFR. When combined into one single nanobody format, together with an albumin-binding nanobody for in vivo half-life extension ^{20, 23}, this nanobody (named CONAN-1) was shown to be a potent receptor antagonist. Importantly, our results show that in a nude mouse model of established A431 xenografts, this biparatopic nanobody format was very potent in inhibiting tumour outgrowth.

Materials and Methods

E.coli strain and cell lines

The bacterial strain used was TG1 ²⁴. The epidermoid squamous carcinoma cell line A431 ²⁵, carrying an amplification of the EGFR gene ²⁶, was purchased from the ATCC (cat. nr. CRL-1555). Her14 cells are derived from NIH 3T3 fibroblasts and stably express roughly 10⁵ copies of the human EGFR on their cell surface ²⁷. The human head and neck squamous cell carcinoma (HNSCC) cell line UM-SCC-14C (14C) was a kind gift of Dr. T.E. Carey, Ann Arbor, MI, USA. All cells were cultured in Dulbecco's Modification of Eagle Medium (DMEM: Gibco, Invitrogen, Paisley, United Kingdom) containing 7.5% (v/v) foetal calf serum and 2 mM L-glutamine in a humidified atmosphere without antibiotics at 37°C under 5% CO₂. Cells were regularly tested by PCR for the presence of mycoplasma and consistently found to be mycoplasma-free.

Plasmids and constructs

The cDNA encoding scFv 425 ²⁸, cloned as bispecific single-chain Fv antibody fragment in pSecTag ²⁹ was a kind gift of Dr. Van Beusechem (Department of Medical Oncology, VU University Medical Center, Amsterdam, the Netherlands). The cDNA was re-cloned from pSecTag in a bacterial expression vector almost identical to pUR5850 ³⁰, except for lacking the C-terminal biotinylation sequence (LRSIFEAQKMEW). Induction of protein expression in E.coli and purification of scFv from the periplasmic space using IMAC were performed as has been described ²⁴.

The construct encoding the EGFR extra-cellular domain (EGFR-ECD; a.a. 1-614) fused to a human IgG1 Fc gene was a kind gift of Prof. Dr. E.J.J. van Zoelen (Centre for Molecular Life Sciences, Radboud University, Nijmegen, the Netherlands). The construct was used to express EGFR-ECD-Fc fusion protein from an in-house developed expression vector using Hek293E cells. After three days of culture, cellular supernatant was collected and fusion protein was purified by means of prot. G affinity chromatography.

Selection of high affinity- and of cetuximab cross-reactive anti-EGFR nanobodies

EGFR “immune” phage nanobody repertoires used for selections had been synthesised as has been described ²⁰ and were a kind gift of Dr. E.G. Hofman (dept. of Cell Biology, Utrecht University, Utrecht, the Netherlands) ³¹. Selections were performed on recombinant, purified and biotinylated EGFR protein comprising the complete extra-cellular domain (a.a. 1-614, ³²). The protein was biotinylated using biotin amido hexanoic acid 3-sulfo-N-hydroxy succinimide ester (Sigma-Aldrich, Zwijndrecht, the Netherlands) according to the manufacturer's recommendations. Excess non-reacted biotin was removed by dialysis against PBS. For affinity selections, antigen concentrations used were 100pM, 50pM, 20pM, 10pM and 1pM. Phage (roughly 10¹⁰ colony forming units (cfu)) and antigen were mixed in a total volume of 100μl PBS containing 1% (w/v) casein and incubated for 3 hours at room temperature (rt) while shaking. For off-rate selection ³³, a 100-fold molar excess of non-biotinylated antigen (purified EGFR-ECD-Fc fusion) was added and incubated for another 3 hours at rt while shaking. Phage bound to the biotinylated antigen were then captured in an extravidin-coated well (5μg/ml in PBS) of a Maxisorp plate (Nunc, Rochester, U.S.A.) for 15 minutes at rt. Non-bound phage were removed by extensive washing with PBS containing 0.1% (v/v) tween-20 (PBST; 20 times) and bound phage were eluted with trypsin (1mg/ml in PBS) for 10 minutes at rt. Trypsin was finally inhibited by the addition of ABTS (1mM) and selected phage were used to infect exponentially growing TG1 as has been described ³⁴.

For the selection of nanobodies that would compete for the binding of cetuximab to EGFR, the method of competitive elution ³⁵ was used. Briefly, biotinylated EGFR-ECD (4μg/ml in PBS containing 0.5% (w/v) casein) was captured in a neutravidin-coated (5μg/ml, overnight in PBS at 4°C) and blocked (1% casein in PBS for an hour at rt) Maxisorp plate for an hour at rt. All incubations were performed with shaking. Phage were allowed to bind for two hours in PBS containing 0.5% (w/v) casein and plates were subsequently thoroughly washed (as described above). Phage bound to overlapping epitopes on EGFR as the one recognised by cetuximab were then eluted by incubation with 200μg/ml cetuximab in PBS for two hours at rt.

Production of selected Nanobodies

After one (affinity-) and two (cetuximab-competitive-) selection rounds, single bacterial clones were picked, grown and nanobody-expression was induced as has been described ³⁴. After four hours of induction, a periplasmic extract was made ²⁴, which was subsequently used to determine EGFR specificity and antagonism by means of ELISA and to measure antibody off-rate by means of SPR using a BIAcore.

Competition ELISA

Maxisorp plates were coated with a rabbit polyclonal anti-human IgG serum (Dako, Glostrup, Sweden; 1:2000 in PBS) over night at 4°C. Next day, wells were washed with PBS, blocked with 2% (w/v) BSA in PBS (PBS/BSA; 30 minutes at rt) and purified EGFR-ECD-Fc was captured at 0.75μg/ml in PBS/BSA for 1 hour at rt. All further incubations were performed in PBS/BSA.

For EGF competition, wells were washed with PBS and a mix of nanobody (either crude periplasmic extract (50% (v/v) in PBS/BSA), or varying concentrations of purified nanobody) in 800pM of biotinylated EGF (Molecular probes/Invitrogen, Carlsbad, U.S.A.) was added. After incubation for 1 hour at rt, wells were washed again and bound EGF was detected with peroxydase-coupled streptavidin (Jackson ImmunoResearch Laboratories, Suffolk, England; 1 in 5000 in PBS/BSA) and staining with OPD/H₂O₂.

For cetuximab, mab 425 or nanobody competition, monoclonal phage were prepared as described ³⁴. Roughly 10¹⁰ phage were mixed with a 100-fold molar excess of either cetuximab, the 424 scFv or nanobody in 2% (w/v) Marvel (skimmed milk powder) in PBS (MPBS) and the mix was added to coated wells containing the EGFR ECD-Fc in triplicate. Bound phage was detected with a peroxydase-coupled antibody to M13 (Amersham/GE Healthcare, Uppsala, Sweden; 1 in 10000 in MPBS) and staining with OPD/H₂O₂. Optical density was read at 490nm.

Re-cloning and expression of selected Nanobody-genes

The cDNA encoding nanobody Alb1 was made synthetically using the sequence information published in patent WO2006/122786 according to described methods ³⁶. For the synthesis of bivalent and trivalent nanobodies, selected nanobody-encoding genes were PCR-amplified using the Expand High Fidelity PCR System (Roche, Mannheim, Germany) with an appropriate primer set, purified, cut with the indicated restriction enzymes and cloned into Sfi1-BstEII cut pUR 5850 ³⁰. Linker sequences (composed of Gly4-Ser (G₄S) repeats) were encoded in the primers, making it possible to vary the length of the linker separating two nanobody genes. Constructs were sequenced ³⁷ to verify that no mutations were introduced by PCR. Protein expression in E.coli TG1 and purification from the periplasmic space by means of IMAC were performed as has been described ²⁴.

Inhibition of EGFR signalling by selected nanobodies and test for EGFR agonism

The assays measuring the inhibition of EGF-induced EGFR activation by nanobodies and detecting possible agonistic effects of nanobodies on the receptor, were performed as described ²⁰. As control for the amount of cell lysate loaded, blots were stained for the total amount of either β-actin, or tubulin.

Inhibition of cell proliferation by nanobodies

Measurement of inhibition of cell proliferation by nanobodies or cetuximab using the sulpho-rhodamine B (SRB) assay ³⁸ was performed as described before ²⁰.

Affinity measurements and pharmacokinetics

Nanobodies were labelled according to the IODO-GEN method ³⁹, as described by Visser et al. ⁴⁰.

For affinity measurements, 14C cells were seeded a day before the assay in 24-well tissue culture plates (Corning, Corning, USA) at 100000 cells per well. ¹²⁵I-labelled nanobodies were diluted in binding medium (DMEM, containing 7.5% (v/v) FCS, 25mM HEPES and 2% (w/v) Marvel) and added to cells in triplicate. After 2 hours of incubation at 4 degrees, non-bound nanobody was removed by washing twice with ice-cold PBS. Bound nanobody was then quantified by lysis of the cells in 1M NaOH and quantification of the radio-activity using a gamma counter (Wallac, Turku, Finland). Results were analysed with the Graph Pad software.

Measurement of in vivo pharmacokinetics with ¹³¹I-7D12-9G8-Alb1 was performed in tumour-bearing mice essentially as described ²³. Next to a therapeutic dose of trivalent nanobody (in group 2 in the therapy study, see below), each mouse was injected intraperitoneally with 0.33 MBq of radio-labelled nanobody (7.5 μg), and blood was drawn at 2 hrs, 6 hrs, 24 hrs, 48 hrs, 72 hrs p.i. to determine detailed pharmacokinetics.

Therapy study

The therapy study was performed essentially as described before ²⁰. However, therapy was started when tumours were established and the average size of the tumours was approximately 100mm³. Group1 received PBS twice a week during 5 weeks, group 2 received nanobody and group 3 received cetuximab. Together with the 2nd and 7th dose of nanobody administration, pharmacokinetics were determined in the nanobody-treated group. Anti-tumour effects were expressed by a growth delay factor (GDF), which was defined as the difference in the median time tumours needed to quadruple in the treated group and in the control group, divided by the median quadrupling time of the control group.

Results

The present study set out to improve the inhibitory capacity of anti-EGFR nanobodies by generating bi-paratopic molecules that are specific for different epitopes on domain III of the EGFR. Therefore, phage nanobody selections were performed using the purified ecto-domain of EGFR as target antigen and `immune' phage nanobody repertoires ^{20, 31}. To obtain the different nanobody specificities, we made use of the mAbs cetuximab and matuzumab that were shown to bind to different and non-overlapping epitopes on domain III of the EGFR ^{13, 15, 16}. We first set out to obtain antagonistic nanobodies with the highest affinity possible, by performing phage selections on very low amounts of biotinylated antigen in solution, combined with “off-rate selections” ³³. Selected nanobody clones were subsequently tested for their ability to inhibit EGF binding to the EGFR in a competition ELISA, followed by screening for nanobodies with a low off-rate using surface plasmon resonance (SPR). Out of 180 clones screened, 6 clones were found to inhibit the binding of EGF to the EGFR (data not shown). These were subsequently screened for their kinetic dissociation rate constant (k_off). Dissociation rate constants were found to vary between 2 and 150 10⁻⁴ s⁻¹ (Table 1A). Because of their low dissociation rate constants, clones 9G8 and 38G7 were then selected for further characterisation (the amino acid sequences of all the selected nanobodies can be found in the supplemental data).

Table 1.

Affinity characteristics and growth delay factors of different nanobodies

A. Kinetic dissociation rate constants of selected antagonistic anti-EGFR nanobodies as measured by SPR using a BIAcore
Nanobody:	koff(10−4 s−1)
27H7	31.2
27E5	20.4
27C7	12.4
27E8	17.3
9G8	4.0
38G7	2.0
7C12	150
7D12	25

B Affinity values of anti-EGFR nanobodies as measured by binding of 125I labelled nanobody to live cells
Nanobody:	KD (nM) on 14C cells	KD (nM) on A431 cells
9G8	14.4	13.8
7D12	10.4	25.7
7D12-7D12	2.1	4.6
9G8-9G8	2.8	10.7
9G8-7D12	7.1	7.5
7D12-9G8	3.1	5.4

C Growth delay factors (GDF) for quadrupling of tumour volume after different treatments
Treatment	GDF
cetuximab	2.19
CONAN-1	1.52
7D12-7D12-Alb1	1

After protein production and -purification, the IC50 for EGF binding to EGFR for both nanobodies was measured in ELISA and found to be 6–7 nM for the 9G8 nanobody and 10 nM for the 38G7 nanobody (Fig. 1A). Since the 9G8 nanobody showed a lower IC50 for EGF binding, was expressed to a much higher level in E.coli and because the 38G7 amino acid sequence contained some very unusual amino acids at several key positions (data not shown), the 9G8 nanobody was selected for further engineering. To gain insight into the epitope specificity of the selected anti-EGFR antagonists, selected nanobodies were tested for their competition for binding to the EGFR with the whole antibody cetuximab or the scFv of the 425 antibody (matuzumab). Surprisingly, all the selected nanobodies competed with the 425 scFv for binding, but not with cetuximab (shown for 9G8 in Fig. 1B).

Figure 1. In vitro characterisation of selected monovalent anti-EGFR nanobodies.

A The binding of biotinylated EGF (800pM) to an EGFR-ECD-Fc fusion was tested in the presence of increasing amounts of the 9G8, 7D12, 7C12 or 38G7 nanobodies, or without nanobody. Receptor-bound EGF was then detected via peroxydase-coupled streptavidin and staining with OPD/H₂O₂.

B–D The binding to EGFR of the 9G8, 7D12 and 9G8 nanobody respectively (B–D), expressed on phage, was detected in a 100-fold molar excess of the indicated antibody (cetuximab) or antibody fragments.

We then specifically sought to select nanobodies that would recognise an epitope that overlapped with that of cetuximab by using the method of competitive elution ³⁵ with the antibody cetuximab. Selected nanobodies were subsequently tested for EGF antagonism, as well as for competition with cetuximab for binding to EGFR in ELISA. Two nanobodies were selected (called 7C12 and 7D12) that blocked the binding of EGF to the EGFR (Fig. 1A) and indeed competed for the binding of cetuximab and not for that of the scFv of matuzumab (shown for the 7D12 nanobody in Fig. 1C). Based on its lower IC50 for EGF binding (8 versus 30nM: Fig. 1A) and lower off-rate (2.5 × 10⁻³ versus 1.5 × 10⁻² s⁻¹: Table 1A), the 7D12 nanobody was then selected for further engineering. However, the fact that the 7D12 and 9G8 nanobodies competed for binding to EGFR with two mAbs that do not block each others binding ^{13, 15}, does not necessarily mean that they did not compete for each others binding. To confirm that the 7D12 and 9G8 nanobodies indeed had different epitope specificities, they were tested for competitive binding to immobilised EGFR. Indeed, and as expected, 7D12 and 9G8 did not compete for each others' binding to EGFR (Fig. 1D).

The linking of two nanobody `heads' into bivalent molecules has already been shown to increase the potency of such molecules ^{20, 41}. To find the optimal bivalent anti-EGFR nanobody combination, the 7D12 and 9G8 nanobody-encoding genes were re-formatted into bivalent molecules ²¹ in all possible combinations: either bivalent, mono-specific, or dual-specific / bi-paratopic, using a standard flexible linker of 10 amino acids (in G₄S repeats). The affinities of the monovalent nanobodies, as well as that of the bivalent and bi-paratopic nanobodies were then determined by binding of ¹²⁵I labelled nanobody to live cells. As expected, bivalent molecules had higher affinities than the corresponding mono-valent counterparts (Table 1B). The mono-specific 7D12-7D12 had the highest affinity, followed by the 7D12-9G8 bi-paratopic molecule. Subsequently, all constructs were tested for their capacity to inhibit EGF-induced EGFR phosphorylation and EGF-dependent cell proliferation. All nanobodies dose-dependently inhibited the EGF-induced phosphorylation of tyrosine (Y) 1068 of the EGFR (Fig. 2). As phosphorylation of tyrosine 1068 of the EGFR has been reported to be the initiation of signalling towards Ras ⁶, this phosphorylation site of EGFR was measured. The best inhibition was achieved with the bivalent 7D12 nanobody and bi-paratopic molecules 9G8-7D12 and 7D12-9G8. However, the first two showed a slight increase in receptor phosphorylation at the highest nanobody dose used (Fig. 2), a phenomenon that was not observed for the 7D12-9G8 bi-paratopic molecule. These results clearly show that the bi-paratopic anti-EGFR nanobody 7D12-9G8 performed best in inhibiting EGFR signalling.

Figure 2. Inhibition of EGF-induced EGFR phosphorylation by mono- and bivalent nanobodies.

EGF (8nM) was mixed with increasing amounts of nanobody (monovalent 7D12 and 9G8; bivalent 7D12 and 9G8 and bi-paratopic nanobodies 7D12-9G8 and 9G8-7D12) and the resulting mixtures were added to Her14 cells. After stimulation for 15 minutes, cell lysates were made, proteins were size-separated and blotted to PVDF membranes. Membranes were then stained for phosphorylated EGFR (Y1068) and for the total amount of tubulin as loading control.

When tested for their capacity to inhibit tumour cell proliferation, all nanobodies inhibited the growth of A431 cells (Fig. 3). For comparison, the whole mab cetuximab was used as reference in all experiments. The bi-paratopic nanobody 7D12-9G8 proved as effective as cetuximab in reducing the growth of A431 cells (Fig. 3D). Both bivalent, mono-specific nanobodies 7D12-7D12 and 9G8-9G8 proved less effective in inhibiting the proliferation of A431 cells (Fig. 3A and B) than cetuximab or the 7D12-9G8 nanobody. For both bivalent, mono-specific molecules increased cell proliferation was observed at higher nanobody concentrations, consistent with the observed increased level of EGFR phosphorylation (Fig. 2). Also, mixtures of mono-valent 7D12 and 9G8 or of bivalent 7D12 and bivalent 9G8 were not as potent as the bi-paratopic 7D12-9G8 molecule in inhibiting A431 cell proliferation (Fig. 3). The bi-paratopic 9G8-7D12 molecule also slightly stimulated the growth of A431 cells at low nanobody concentrations, again in agreement with increased levels of phosphorylated EGFR observed on blot (Fig. 2). These data all provide strong support for the choice of the 7D12-9G8 as the most effective bivalent, bi-paratopic nanobody combination. In addition, they show that the order of the two `heads' was critically important for the activity of this bi-paratopic anti-EGFR nanobody. Based on these results, the 7D12-9G8 bi-paratopic molecule was selected for further optimisation and in vivo testing.

Figure 3. Inhibition of A431 tumour cell proliferation by mono- and bivalent nanobodies.

A–D The proliferation of A431 cells in the presence of increasing amounts of nanobody (A: mono- and bivalent 7D12; B: mono- and bivalent 9G8; C: mixes of mono- and bivalent 7D12 and 9G8 and D: bi-paratopic molecules 7D12-9G8 and 9G8-7D12) was measured using a sulphorhodamine-based stain of total cellular protein after TCA precipitation. Proliferation is plotted as percentage of maximal growth (cells left without treatment). The whole antibody cetuximab was used as `standard' in every test. Data points where the value of the control is statistically different from that of the nanobody-treated group are indicated by an asterix. C: only different between 7D12+9G8 and control; D: only different between 9G8-7D12 and control.

An important characteristic of the 7D12-9G8 bi-paratopic molecule is the linker length and linker composition connecting the two nanobody `heads'. This linker is supposed to provide sufficient space/length and freedom to allow the two nanobodies to bind simultaneously (`chelate' ⁴²) to the same ecto-domain of EGFR. We therefore analysed the effect of flexible linkers consisting of G₄S repeats (as found in the filamentous bacterio-phage pIII protein) varying in length from 5 to 30 a.a.. The resulting bivalent constructs were then functionally tested for their ability to inhibit A431 cell proliferation. Surprisingly, the length of the linker used between the two nanobody “heads” turned out to be almost inversely correlated with the efficacy of the molecule in inhibiting A431 cell proliferation, with 5 and 10a.a. linkers being optimal (Fig. 4A). Therefore, a linker of 10a.a. was chosen as the optimal format for the anti-EGFR 7D12-9G8 bi-paratopic nanobody.

Figure 4. In vitro optimalisation and characterisation of trivalent bi-paratopic nanobody CONAN-1.

A The proliferation of A431 cells in the presence of increasing amounts of bi-paratopic nanobody 7D12-9G8 (with linkers varying in length between the two nanobody units) was measured using a sulphorhodamine-based stain of total cellular protein after TCA precipitation. Proliferation is plotted as percentage of maximal growth (cells left without treatment). Error bars have been omitted to increase the clarity of the figure. The different linker lengths between the two nanobodies are indicated.

B Purified EGFR-ECD-Fc fusion protein was immobilised via a coated anti-human Fc anti-serum and a concentration range of either bivalent (7D12-9G8) or trivalent (CONAN-1) nanobody was bound. The binding of biotinylated mouse serum albumin (MSA) was subsequently detected with peroxydase-conjugated streptavidin and staining with OPD/H₂O₂.

C The proliferation of A431 cells in the presence of increasing amounts of nanobody was measured using the SRB assay. The effect of the presence of human serum albumin (HSA: 1%) in the medium was assessed.

D EGF (8nM) or nanobody (1mM) was added to serum-starved Her14 cells (-). After stimulation for 15 minutes, cell lysates were prepared, proteins were size-separated and blotted to PVDF membranes. Membranes were then stained for phosphorylated EGFR (Y1068) and for the total amount of tubulin as loading control.

E EGF (8nM) was mixed with increasing amounts of nanobody and the resulting mixture was added to Her14 cells. After stimulation for 15 minutes, cell lysates were prepared, proteins were size-separated and blotted to PVDF membranes. Membranes were then stained for phosphorylated EGFR (Y1068) and for the total amount of tubulin as loading control.

Since the 7D12-9G8 nanobody has a molecular weight of roughly 30 kDa, it would be cleared very rapidly in vivo via the kidneys once injected into the blood stream of mice ⁴³. To prolong the in vivo half-life of small proteins, binding to albumin has been reported to be an excellent option ^{20, 23, 44}. Therefore, the gene encoding the anti-MSA/HSA nanobody Alb1 (this nanobody recognises both MSA as well as HSA with a KD of 6.5nM for MSA and 0.57nM for HSA) was made synthetically ³⁶ and fused C-terminally to the bi-paratopic 7D12-9G8 nanobody using two different linker lengths (15 or 30 amino acids in repeats of Gly4-Ser) between the anti-EGFR nanobodies and Alb1. First, trivalent nanobodies were tested for their functionality in ELISA: the binding of biotinylated MSA to EGFR-bound 7D12-9G8 and 7D12-9G8-Alb1 was assessed and shown to depend on the presence of the Alb1 nanobody (Fig. 4B). These results show that at least one of the anti-EGFR `heads' together with the Alb1 nanobody present in the same trivalent molecule could simultaneously bind antigen. Since in an in vivo (therapy) situation, albumin will be abundantly present, both trivalent nanobodies (containing either a 15- or 30 a.a. linker before the Alb1 nanobody) were then tested for their capacity to inhibit A431 cell proliferation in the presence of 1% (w/v) HSA. The construct containing a 15 amino acid linker before the Alb1 nanobody significantly lost potency in the presence of HSA (data not shown). However, for the construct containing a 30 residue linker between the anti-EGFR units and Alb1 unit, efficacy in inhibition of cell proliferation was not affected by the presence of HSA (Fig 4C). Therefore, this version was selected for further in vivo testing and was named COoperative NANobody-1 (CONAN-1).

To check whether the CONAN-1 nanobody by itself did not induce activation of the EGFR, the nanobody was given as ligand at high concentration to EGFR over-expressing cells (Her14 cells). Figure 4D shows that CONAN-1 did not cause receptor activation in the absence of EGF, while cells readily responded to EGF stimulation. Finally, the CONAN-1 nanobody was shown to strongly inhibit EGF induced signalling (Fig. 4E): not only was receptor phosphorylation inhibited (Fig. 2), but the phosphorylation of MAPK was also reduced to background levels. These results show that the CONAN-1 nanobody functioned as a true receptor antagonist and it was therefore tested in an in vivo therapy study.

To measure blood pharmacokinetics of the trivalent CONAN-1 molecule, the nanobody was radio-labelled with ¹³¹I. Together with a therapeutic dose of nanobody, a trace amount of radio-labelled nanobody was injected intravenously (iv) in the tail vein of tumour-bearing mice. Blood sampling in time revealed an in vivo half-life of the protein of approximately 48 hours (Fig. 5A), when fitted to a mono-exponential decay. In a murine model of established A431 human xenografts in athymic mice, the administration of CONAN-1 intra-peritoneally (ip) as therapy was compared with therapy using cetuximab. Based on the measured pharmacokinetics of CONAN-1 (Fig. 5A), therapy was given twice weekly. CONAN-1 had an efficacy in inhibiting the growth of established A431 tumours that was largely comparable to that of cetuximab during treatment (Fig. 5B). The mean tumour volume in the cetuximab-treated group diminished slightly after 4 days, up to 2 weeks of treatment (a phenomenon not observed in the nanobody-treated group), but then increased at approximately the same speed as that of the nanobody-treated group. After treatment had been stopped, tumours of mice in the nanobody-treated group did not re-grow significantly faster than tumours of mice in the cetuximab-treated group. During therapy, CONAN-1 induced a significantly stronger effect than cetuximab during the first 6 days, but the cetuximab-treated group showed a significantly stronger anti-tumour effect after day 8 (p<0.05; independent sample T-test). A comparison of the growth-delay factors for quadrupling of the tumour volume revealed that CONAN-1 induced a strong response in tumour growth inhibition (GDF = 1.52) but that this response was slightly stronger in the group treated with cetuximab (GDF = 2.19: Table 1C). In a separate, but equal experimental setup, therapy using the trivalent anti-EGFR nanobody 7D12-7D12-Alb1 was also compared to that using the whole mAb cetuximab. A comparison of the GDF's calculated for both nanobody treatments revealed that CONAN-1 induced a much stronger anti-tumour response than the 7D12-7D12-Alb1 nanobody (GDF = 1, Table 1C). Importantly, this proves again that the combination of different paratopes and thereby different modes of EGFR inhibition into a single molecule was superior to the use of mono-specific targeting for the inhibition of EGFR.

Figure 5. In vivo pharmacokinetics and therapy using CONAN-1 nanobody in athymic (nu/nu) mice bearing subcutaneous A431 xenografts.

A Together with a therapeutic dose of trivalent nanobody (1mg), ¹³¹I-labeled CONAN-1 (7.5 μg, corresponding to 0.33 MBq) was injected intra-peritoneally (i.p.) in tumour-bearing athymic (nu/nu) mice to measure detailed blood pharmacokinetics. Blood pool radio-activity was measured in time and plotted as percentage of injected dose per gram of tissue as a function of time.

B Athymic (nude) mice bearing subcutaneously implanted A431 xenografts were treated with phosphate buffered saline (PBS; diamonds), cetuximab; barbed rounds or trivalent nanobody 7D12-9G8-Alb1 (CONAN-1; crosses). Treatment consisted of bi-weekly injections of 1mg of protein intra-peritoneally (arrows).

Discussion

The present study set out to combine nanobodies into a bi-paratopic format in order to improve their efficacy in tumour growth inhibition. The results show the efficacy of the optimised CONAN-1 nanobody in therapy to be better than that of mono-specific 7D12-7D12-Alb1 and almost equal to that of the well-characterised anti-EGFR mAb cetuximab, although the former is devoid of immune effector functions.

To obtain anti-EGFR nanobodies that recognize non-overlapping epitopes on the ligand binding domain 2 of EGFR (domain III), we made use of two mAbs of which the binding sites were previously shown to be different by crystallography ^{13, 15}. First, we selected nanobodies for high affinity binding ³³. Surprisingly, these selections only resulted in clones that cross-reacted with the scFv of matuzumab (shown for the 9G8 nanobody in Fig. 1). The nanobodies Ia1, L2-3.40 and IIIa3, previously found using EGF elution ²⁰, were also found to compete for binding to EGFR with the 9G8 nanobody and the 425 scFv (data not shown). The high prevalence of nanobodies recognising the part of domain III also bound by matuzumab is probably a reflection of the immunogenicity of that particular part of the receptor. However, by carefully designed phage nanobody selections, cetuximab cross-reactive nanobodies were readily obtained. These results underline the power of the combination of active immunisation of Llama, nanobody repertoire cloning and phage nanobody selections to obtain nanobodies that recognise defined epitopes on a given antigen (for review, see ²²).

Secondly, phage nanobody selections were performed in combination with specific elution using cetuximab. Only two nanobodies were found that blocked binding of EGF to EGFR and that of cetuximab to EGFR: clones 7D12 and 7C12. These nanobodies have previously been described by Gainkam et al. ⁴⁵. The reported affinity of the 7D12 nanobody (2.3nM: ⁴⁵) differs from the affinity value we obtained (10 nM: Table 1B). This discrepancy might well be due to the different methods used to measure the affinity. We deliberately chose for cell binding experiments, as the affinity values obtained by this method are probably more representative for the in vivo situation. The EGFR makes numerous contacts with other receptors, integrins and lipids ³¹ which may shield particular epitopes on the ecto-domain that are accessible in an in vitro SPR setup. However, the ¹²⁵I-labelling of the nanobodies may have influenced their affinity, as tyrosines in the antigen binding sites may have been altered, thereby lowering the affinity.

The order in which the two nanobodies were linked together in a bivalent or bi-paratopic molecule was shown to be an important parameter for the efficacy of the anti-EGFR bi-paratopic molecule (Figs. 2 and 3). In addition, the linker composition also strongly influenced the characteristics of the nanobody, as a hinge-derived sequence between 9G8 and 7D12 caused this bi-paratopic molecule to act as a receptor agonist (results not shown). The same phenomenon can slightly be observed for the 9G8-7D12 bi-paratopic nanobody containing a flexible linker and the 7D12-7D12 mono-specific molecule (Fig. 4). The engineering and thorough testing of this type of therapeutic molecules is therefore critically important to avoid artificial receptor activation.

Surprisingly, the 7D12-7D12 bivalent nanobody had the highest affinity, yet was not the most potent in inhibiting tumour cell proliferation (Table 1B and Fig. 3). This may be explained by artificial activation of the EGFR by cross-linking of receptors, caused by inter-molecular binding of the nanobody. The 7D12-9G8 nanobody did not cause artificial receptor activation (Figs. 2, 3C and 4D) and therefore functioned as true receptor antagonist. However, simultaneous binding of the two `heads' in the bi-paratopic nanobody 7D12-9G8 to one and the same EGFR molecule (`chelating' binding 42) could not be demonstrated. Size exclusion chromatography and cross-linking experiments did not give conclusive results (data not shown). Preliminary X-ray crystallographic data confirm the predicted location of the epitopes of 7D12 and 9G8 (details of co-crystal structures of EGFR fragments with 7D12 and 9G8 will be published separately, Ferguson et al., manuscript in preparation). These structures also suggest that simultaneous binding of both heads of 7D12-9G8 may be possible with a 10 amino acids linker. However, “chelating” binding will be highly dependent on the linker composition. Based on the estimated locations of the C-terminus of 7D12 and N-terminus of 9G8 in these crystal structures, a linker of only 10 amino acids would have to pass very closely to the surface of EGFR. It is conceivable that such a linker does not permit the bi-paratopic nanobody to `chelate' <sup>42</sup> the EGFR, but that the 7D12-9G8 nanobody binds the receptor with only one `head' at any given time. However, the presence of the second specificity within the same molecule probably results in fast re-binding once the first nanobody unit dissociates from its epitope, thereby resulting in increased apparent affinity (Table 1B: compare 7D12 and 9G8 with 7D12-9G8) and increased potency in EGFR inhibition (Fig. 3). Also, binding of the bi-paratopic 7D12-9G8 to EGFR may induce some conformational change and thus an `induced fit', thereby permitting both `heads' to bind simultaneously. For the mono-specific, bivalent 7D12-7D12 molecule, simultaneous binding of both `heads' would require a cluster of receptors, where domain III is available in two adjacent EGFR molecules. Pre-dimers of EGFR in the absence of EGF have been demonstrated <sup>46, 47</sup>. However, the EGFR in such complex will most probably not be bound bivalently by 7D12-7D12, since the L2 domains are located on either side of the `back-to-back' dimer ², pointing away from each other.

The in vivo half-life of the trivalent CONAN-1 nanobody was measured in athymic tumour-bearing mice and found to be approximately 48 hrs. This half-life is very comparable to that of a similar nanobody construct reported by Tijink et al. ²³ and to that of directly radio-labelled MSA (data not shown). Since the affinity of the anti-albumin unit (Alb1) for HSA is even higher than that for MSA (6.5 (MSA) versus 0.57nM (HSA)), it is expected that the CONAN-1 nanobody would circulate in humans with the same kinetics as that of HSA (its half-life being 10–14 days). The pharmacokinetics would then be very comparable to that of a whole IgG.

The potency of the trivalent CONAN-1 nanobody in inhibiting the in vivo outgrowth of established A431 tumours was almost comparable to that of the mAb cetuximab. Treatment with cetuximab resulted in tumour regression after 4 days, which lasted until approximately day 14 (when tumour started to re-grow). This is indicative of activation of immune effector cells (cetuximab, being a chimeric antibody, can efficiently recruit murine immune effector cells). Despite the fact that the trivalent CONAN-1 nanobody cannot interact with the immune system, its efficacy was largely comparable to that of cetuximab. This could be partially due to better tumour penetration ²³ in combination with similar (or even better) pharmacokinetics (Fig. 6 and ²³). The modular nature of nanobodies permits the addition of `effector' heads (e.g. a nanobody to recruit effector cells). It would be interesting to test such constructs in comparison with cetuximab.

The above-mentioned modularity of nanobodies in combination with the power and possibilities of phage display also permit the design and synthesis of combinations of nanobodies recognising well-defined epitopes on other receptors (e.g. the combination of nanobodies recognising the pertuzumab- and trastuzumab epitopes on Her2/Neu). Additionally, it permits the synthesis of multi-specific molecules capable of inhibiting two (or more) signal transduction pathways simultaneously, which may well lead to a higher efficacy of nanobody-mediated tumour growth inhibition. In conclusion, our results show that the rational design and synthesis of multivalent nanobody molecules is a promising option to develop new cancer treatment modalities.