Monofunctional trap Fc-fusions each contain a heavy chain Fc region and include the signal sequence (MDWTWRILFLVAAATGTHA) at their N-termini. The DNA coding regions for the constructs were prepared synthetically (Biobasic Inc. or Genescript USA Inc.) and were cloned into the HindIII (5’ end) and BamH1 (3’ end) sites of the pTT5 mammalian expression plasmid vector (39). Fusion proteins were produced by transient transfection of Chinese Hamster Ovary (CHO) cells with the heavy chain T2m or T22d35 fused to the IgG heavy chain (T2m-HC and T22d35-HC, respectively) construct. Briefly, T2m-HC or T22d35-HC plasmid DNAs were transfected into a 2.5 L and 4.6 L culture, respectively, of CHO-3E7 cells in FreeStyle F17 medium (Invitrogen) containing 4 mM glutamine and 0.1% Kolliphor p-188 (Sigma) and maintained at 37°C. Transfection conditions were: DNA (80% plasmid construct, 15% AKT plasmid, 5% GFP plasmid) and PEI (polyethylenimine)pro (Polyplus) (ratio = 1:2.5). At 24 h post-transfection, 10% Tryptone N1 feed (TekniScience Inc.) and 0.5 mM Vaporic acid (VPA, Sigma) were added, and the temperature was shifted to 32°C to promote the production and secretion of the fusion proteins. Cultures were then maintained for 15 days post-transfection after which the cells were harvested. At the final harvest, the cell viability was 89.6%.

Depending on their structures, multifunctional constructs were comprised of a heavy-chain signal sequence MDWTWRILFLVAAATGTHA and a light-chain signal sequence MVLQTQVFISLLLWISGAYG (if the light chain was present in the structure). The DNA coding for constructs were prepared synthetically (Biobasic Inc. or Genescript USA Inc.). Trap constructs comprised of a “headless” Fc, antibody, and D10-Fc were cloned into the EcoR1 (5’ end) and BamH1 (3’ end) sites and those comprising FC5-Fc were cloned into the HindIII (5’ end) and BamH1 (3’ end) sites of the pTT5 mammalian expression plasmid vector (39). The Cet-T2m and Cet-T22d35 constructs were produced by transient co-transfection of CHO cells with the heavy chain (HC)-T2m or (HC)-T22d35 construct combined with the Cetuximab light chain (LC) construct, which then assembled as the Cetuximab-T22d35 (Cet-T22d35) or Cetuximab-T2m (Cet-T2m) fusion proteins. Briefly, CetHC-T22d35 and CetLC plasmid DNAs (ratio = 3:2) were co-transfected into a 10L Wavebag culture of CHO-3E7 cells in FreeStyle F17 medium (Invitrogen) containing 4 mM glutamine and 0.1% Kolliphor p-188 (Sigma) and maintained at 37°C. Transfection conditions were: DNA (50% HC+LC plasmids, 30% ssDNA, 15% AKT plasmid, 5% GFP plasmid): PEI (polyethylenimine)pro (Polyplus) (ratio = 1:2.5). At 24h post-transfection, 10% Tryptone N1feed (TekniScience Inc.) and 0.5 mM Vaporic acid (VPA, Sigma) were added, and the temperature was shifted to 32 °C to promote the production and secretion of the fusion proteins. Cultures were then maintained for 15 days post-transfection after which the cells were harvested. At final harvest the cell viability was 89.6%. Similar transfection and production methods were performed for the other antibody-trap examples listed in Table 1 . For production of the ‘headless’, FC5-, and D10- Fc-fusions the composition of the transfection mixture was modified as follows: DNA (80% plasmid construct, 15% AKT plasmid, 5% GFP plasmid): PEIpro (ratio 1:2.5).

T cells were isolated from mouse spleens and seeded in 96-well plates (5x10⁵ cells/well), and co-cultured in the presence of cultured BALB/c-derived tibia bone marrow dendritic cells (DC) isolated from naïve, non-tumor bearing, untreated mice and 4T1 breast cancer cell lysate (25 μg protein per mL). DCs were obtained by flushing cells from femurs and tibias of naïve, non-tumour bearing, untreated BALB/c mice. Red blood cells were lysed using ACK lysis buffer, after which the remaining cells were cultured in complete RPMI-1640 with L-glutamine, pen/strep/β-mercaptoethanol/10% FBS supplemented with GM-CSF (10 ng/mL) and IL-4 (10 ng/mL). After 72 h, cells were further incubated for an additional 18 h in the presence of ³H-TdR (1 μCi), after which cells were collected onto glass fiber filters and ³H radioactivity was evaluated by liquid scintillography.

CD4⁺ and CD8⁺ T cells were isolated from draining lymph nodes using ThermoFisher mouse pan-T cell DynaBeads, according to the manufacturer’s instructions, stained with PE-labeled anti-CD4 and PE-CY5-labeled anti-CD8 mAbs (30 min, 4°C), washed and suspended in PBS/10%FBS/FITC-labeled Annexin V (15 min, RT). Cell populations were evaluated by flow cytometry.

T cells isolated from mouse lymph nodes (CyTox 96 non-radioactive cytotoxicity assay, Promega) according to the manufacturer’s instructions. Naïve target tumor cells (4T1 mouse breast or B16F10 mouse melanoma cells) were plated and incubated for 4 h with CD8⁺ effector T cells isolated (as describe above) from the lymph nodes of mice treated with saline, T22d35, or T22d35-Fc. The LDH release by the tumor cells in response to the T cells was evaluated over a period of 30 min using various effector (E) and target (T) cell ratios (E:T=10:1, 25:1, and 50:1) by measuring the LDH-mediated reduction of INT dye to blue formazan according to the manufacturer’s instructions (LDH cytotoxic assay kit, Abcam).

We initially linked the T2m module to the C-terminus of a hIgG1, -2, -3, and -4 Fc-fragments via the TβRII natural linker sequence. It should be noted that the hinge regions of the four hIgGs differ in terms of their cysteine content; the hIgG1 and hIgG4 hinges contain 2 cysteines while the hIgG2 and hIgG3 hinges contain, 3 and 11 cysteines, respectively. The bare N-terminal Fc hinge regions of the so-called ‘headless’ trap fusions could, if left unaltered, negatively affect protein expression and cause aggregation during the manufacturing process. We therefore engineered these hinge regions by 1) removing (ΔC) the cysteines altogether or 2) replacing cysteines by serines (S). For the hIgG3 fusions, due to its long hinge region, we made these modifications using a truncated Fc fragment which contains only the last 3 cysteine residues in its hinge region. Using this approach, the hIgG3 fusions thus contain a hinge region that closely matches the hinge region of the other human IgGs ( Supplementary Table S2 ). All fusions could be produced and purified, except for those containing the hIgG4 Fc, with yields that can be ranked as follows: hIgG1 > hIgG2 > hIgG3 ( Supplementary Table S7 ). This allowed us to select hIgG1Fc(SCC)ΔK-T2m (indicated in bold in Supplementary Table S2 ). However, to avoid any potential immunogenicity issues caused by the serine residues, we further engineered the N-terminal sequence by deleting the ‘EPKSS’ sequence segment from its hinge, thus generating our hIgG1Fc(CC)ΔK-T2m lead. Building on the knowledge obtained through the N-terminal designs, we then devised three C-terminal T22d35 Fc-fusions: hIgG1Fc(C)ΔK-T22d35, hIgG1Fc(CC)ΔK-T22d35, and hIg2Fc(CC)ΔK-T22d35 ( Supplementary Table S3 ). Of these three, the hIgG2Fc(CC)ΔK-T22d35 production showed the highest monomeric content (96.75%), the lowest level of aggregation (3.3%) and contained no fragments, compared to the other two fusions ( Supplementary Figure S2 ). We thus selected the hIgG1Fc(CC)ΔK-T2m and hIg2Fc(CC)ΔK-T22d35 as our C-terminally Fc-fused leads for further development, which are subsequently referred to as Fc-T2m (bold in Supplementary Table S5 ) and Fc-T22d35 (bold in Supplementary Table S3 ), respectively.

In addition to the C-terminal Fc-fusions we also devised one T2m and five T22d35 N-terminally hIgG1 and -2 Fc-fusions while applying a variety of linker designs ( Supplementary Table S4 ). All fusions were produced and purified, and showed a monomeric content of >98%, with a low % of aggregates, and no fragment content ( Supplementary Figure S2 ). Of these variants we selected the T2m-hIgG2Fc(CCCC)ΔK and the T22d35-hIgG2Fc(CC)ΔK as our N-terminally Fc-fused leads for further assessment; these variants are, further referred to as T2m-Fc and T22d35-Fc, respectively (bold in Supplementary Table S4 ). Supplementary Figure S1 shows the purification profiles of the selected C-terminal Fc-T2m (A) and Fc-T22d35 (B) and N-terminal T2m-Fc (C) and T22d35-Fc (D) lead fusions. Size exclusion Chromatography (SEC) elution profiles after Protein-A affinity purification revealed that these fusions are relatively pure and devoid of aggregates, and were shown to be >95% monomeric by UPLC-SEC, with the exception of the T22d35-Fc, which is ~87% monomeric. Further SDS-PAGE assessment, confirmed the expected molecular weight of these fusions: ~60 kDa and ~90 kDa for Fc-T2m and T2m-Fc, and ~90 kDa and ~150 kDa for Fc-T22d35 and T22d35-Fc, under respectively reducing and non-reducing conditions, respectively.

The purpose of this manuscript is also to demonstrate the modularity and potential therapeutic functionality of TGF-β trap in the context of a bifunctional moiety by fusing the trap to N-terminus of several antibodies or other functional entities such as a blood-brain barrier (BBB) crossing (FC5V_HH) (37) or a bone targeting moiety (D10) (32, 38). For the generation of bifunctional antibody-trap fusions ( Supplementary Table S5 ), we used the well-characterized monoclonal antibodies Cetuximab, Herceptin, Avastin and Synagis, which all showed very similar production and purification profiles. Supplementary Figure S1 shows the purification profile of Cet-T2m (E) and Cet-T22d35 (F) as representative examples, showing a monomeric purity of these fusions by UPLC-SEC of >99% after Protein-A and SEC purification, and their expected molecular weight by SDS-PAGE [Cet-T2m: ~242 kDa (non-reducing), and ~78 kDa (HC) and ~27 kDa (LC) under reducing conditions; Cet-T22d35: ~250 kDa (non-reducing), and ~110 kDa (HC) and ~30 kDa (LC) under reducing conditions].

To generate other bifunctional non-antibody trap fusions, we C-terminally linked the T2m and T22d35 to an already existing N-terminally mIgG2a Fc-fusion of the single domain FC5V_HH (37), and we also fused a functional ‘homing’ peptide, the D10 poly-aspartate sequence (32, 38), to the N-terminus of the earlier described Fc-T2m ( Supplementary Tables S5 and S6 ). It should be noted that the hinge region of the mIgG2a Fc in the FC5V_HH fusions was not re-engineered. Four D10 fusion were designed using various sequences linking the D10 to the N-terminus of the Fc-T2m ( Supplementary Table S6 ). These fusions were expressed at a small scale with very similar yields and monomeric purities >98% (data not shown), indicating that neither the D10, nor the linker sequences used were detrimental for the production and purification of this fusion. We selected the D10-hIgG1Fc(CC)ΔK-T2m fusion for further functional assessment.

The selected FC5V_HH-Fc-T2m, FC5V_HH-Fc-T22d35 ( Supplementary Table S5 ) and D10-FchIgG1(CC)ΔK-T2m ( Supplementary Table S6 ) fusions are, from this point, referred to as FC5-Fc-T2m, FC5-Fc-T22d35 and D10-Fc-T2m, respectively. Supplementary Figure S1 shows the purity of the FC5-Fc-T2m (G), FC5-Fc-T22d35 (H) and D10-Fc-T2m (I) to be >94%, 83%, and >99% monomeric by UPLC-SEC after Protein-A and SEC purification, and their expected molecular weight by SDS-PAGE [FC5-Fc-T2m: ~100 kDa (non-reducing), and ~80 kDa under reducing conditions; FC5-Fc-T22d35: ~155 kDa (non-reducing), and ~80 kDa under reducing conditions; D10-Fc-T2m: ~150 kDa (non-reducing), and ~80 kDa under reducing conditions]. Duplicate bands are likely the result of differences in glycosylation.

An overview of the production and purification for all antibody-, FC5- and D10-trap fusions that were generated for this study is shown in Supplementary Table S9 . Our data clearly demonstrates that, using the described design strategies, our T2m and T22d35 traps, either as a N- or C-terminal fusion in the context of an Fc fragment alone, combined with an additional targeting moiety, or as an antibody can be produced at relatively high titers and can be purified by Protein-A followed by SEC. This demonstrates good manufacturability of both the mono- and bifunctional T2m and T22d35 N- and C-terminally Fc-fused proteins. It should be noted that the C-terminal lysine residue of the Fc fragment was deleted from all trap fusions to avoid cleavage of C-terminal linked sequences.

For all TGF-β isoforms ( Figure 1D , TGF-β1; Figure 1E , TGF-β2; Figure 1F , TGF-β3), the potency of both N- and C-terminally Fc-fused T22d35 and the N-terminally fused Fc-T2m is superior to that of the C-terminally fused Fc-T2m, which in turn behaves similarly to the non-Fc-fused T22d35 single-chain trap. Monofunctional Fc-fused traps can therefore be ranked as follows: Fc-T22d35 ≈ T22d35-Fc > Fc-T2m > T2m-Fc ≈ T22d35. The calculated IC₅₀ values against TGF-β1 ( Table 1 ) for Fc-T22d35 and T22d35-Fc (0.006297 nM and 0.0033116 nM) demonstrate their potencies to be at least ~520- and ~920-fold better than the IC₅₀ values calculated for T22d35 (3.253 nM). When comparing their TGFβ1 IC₅₀ to those calculated for Fc-T2m (0.02293 nM) and T2m-Fc (2.500 nM), the Fc-T22d35 is ~3.6-fold better than the Fc-T2m while the T22d35-Fc is ~760-fold more potent than the T2m-Fc. This indicates that the orientation of the trap, with respect to the hIgG Fc fragment, is very important for its TGF-β1 neutralization. In addition, our data also shows that Fc-T22d35 and T22d35-Fc neutralize TGF-β2 more effectively compared to either the Fc-T2m or T2m-Fc constructs; however, this is still to a much lesser extent than TGF-β1 and -β3. It should be noted that, although the neutralization potency of the Fc-T22d35 and T22d35-Fc trap are very similar for TGF-β1 and -β3, the T2m-Fc variant displayed a ~27-fold higher neutralization potency for TGF-β3 compared to TGF-β1 (0.0943 nM versus 2.500 nM, respectively), suggesting that the T2m conformation as an N-terminal Fc-fusion is better suited to neutralize TGFβ-3 compared to TGF-β2. Moreover, the neutralization potency of the C-terminal T2m and T22d35 fusions, and to a lesser extend the N-terminal fusions, significantly increased compared to the non-fused T2m and single-chain non-fused T22d35 protein, which are non-neutralizing with respect to TGF-β2 (28). Fusing more than two T2m domains to an Fc fragment (i.e. more than the two T2m domains that are present in the T22d35 construct) did not further increase the TGF-β1 neutralization potency of these fusions when compared to Fc-fused T22d35 (data not shown).

To demonstrate the modularity of the T2m and T22d35 traps we generated a series of bifunctional antibody fusions in which the C-terminus of the heavy chain was linked to the trap ( Figure 2A ; Cet, Cetuximab; Her, Herceptin; Ava, Avastin; Syn, Synagis), and evaluated both the antibody and trap function in the context of the fusion.

To confirm the relative TGF-β neutralization of antibody fusions, we used the earlier described A549 IL-11 release assay. In these experiments ( Figure 2B ), the TGF-β1 neutralization potencies of Cet-T2m and Cet-T22d35, Her-T22d35 and Syn-T22d35 were compared to T22d35, Fc-T2m and Fc-T22d35. As anticipated, the neutralization potencies of all antibody -T22d35 fusions were practically identical to that of the C-terminal Fc-T22d35 fusion ( Table 2 ), exhibiting IC₅₀ values in the range of 0.0005 nM, whereas the IC₅₀ value for the T2m fusions (e.g., Cet-T2m) and the non-Fc-fused T22d35 were calculated as ~0.05 nM and ~0.5 nM, respectively. As such antibody -T22d35 and Fc-T22d35 fusions displayed a ~100-fold increase in TGF-β1 neutralizing potency, but also demonstrated that constructs with two C-terminally fused TGF-βRII-EDs (i.e., T22d35) are ~10-fold more potent than similar constructs containing a single TGF-βRII-ED (i.e., T2m).

We then evaluated the binding of the Cet-T22d35, Her-T22d35 and Syn-T22d35 fusions to their intended target antigen by SPR, by comparing these to the non-fused antibody s. Our data shows that the K_D values of al antibody fusions are very similar to the ones calculated for the respective non-fused parental antibodies ( Table 3 ), clearly demonstrating that fusion of one (Cet-T2m) or two (Cet-T22d35, Her-T22d35, or Syn-T22d35) TGF-βRII-ED(s) to the C-terminus of an antibody Fc region does not significantly alter antigen-binding affinities and K_D values of the antibody. These data suggest that the ectodomain(s) can be readily fused to any antibody without compromising the ability of the antibody to bind its target antigen.

To gain insight into whether the binding of TGF-β2 was affected when the TβRII-ED was fused to an antibody, we analyzed TGF-β2 binding by Cet-T2m, Cet-T22d35, and T22d35 in a competitive SPR binding experiment ( Figure 2C ). Our data clearly shows that the binding of TGF-β2 by TβRII-ED dramatically increases when it is fused to the C-terminus of an antibody such as Cetuximab, with a ~200-fold and ~100-fold increase in the TGF-β2 EC₅₀ values for Cet-T22d35 (EC₅₀ = 0.50 nM) and Cet-T2m (EC₅₀ = 1.17 nM), respectively, when compared to non-fused T22d35 (EC₅₀ > 100 nM). This indicates that the antibody fusion of either T2m or T22d35 improves the affinity for TGF-β2, and agrees with the increased degree of TGF-β2 neutralization we observed for Fc-T2m and Fc-T22d35 fusions in the A549 IL-11 neutralization assay ( Table 1 ).

To further demonstrate that the antibody function per se is not affected by the presence of C-terminally fused TGF-βRII ectodomain(s), we further evaluated other attributes of the Cetuximab fusions (using Cet-T2m and Cet-T22d35 as examples).

To assess the ability of Cetuximab to maintain its therapeutic function when fused to either T2m or T22d35, we evaluated the Epidermal Growth Factor (EGF)-induced phosphorylation of the Epidermal Growth Factor Receptor (EGFR) expressed by human non-small lung cancer A549 cells by its EGF ligand. As shown in Figure 2D , Cetuximab, Cet-T2m and Cet-T22d35 all inhibited EGFR phosphorylation to a similar extent at various doses, relative to the EGF control, whereas T22d35 alone had no effect on the EGF-induced receptor phosphorylation. These results confirm that fusion of the C-terminal T2m or T22d35 modules to Cetuximab does not interfere with the antibody’s ability to block EGF-induced EGFR phosphorylation.

It is known that A549 cells exposed to a combination of EGF+TGF-β1 undergo EMT (45), a phenomenon that is characterized by these cells transitioning from an epithelial-like ‘cobble-stone’ morphology to that of a more elongated mesenchymal morphology ( Figure 2E ). This transition is accompanied by changes in the cell-surface expression levels of the adherens junction proteins such as E- and N-cadherin, with E-cadherin levels being down- and N-cadherin levels being up-regulated. We thus assessed the cell surface expression of both E- and N-cadherin by flow cytometry in A549 cells after a 3-day exposure to EGF+TGF-β1 in the presence or absence of Cet-T22d35, Cetuximab+T22d35, and Cetuximab and T22d35 alone. Figure 2F clearly shows that Cet-T22d35 significantly inhibits the up- and down-regulation of respectively N-cadherin and E-cadherin, which is better than Cetuximab or T22d35 alone, or even the Cetuximab+T22d35 combination. These results thus demonstrate the superior EMT neutralization potency of the Cet-T22d35 fusion.

Disruption of the autocrine EGFR signaling cascade by Cetuximab has been shown to result in varying degrees of cytotoxicity in EGFR-expressing cells (46). To evaluate whether Cet-T22d35 retained this function, we compared the cytotoxicity induced by Cet-T22d35 to that of Cetuximab or T22d35 alone in MDA-MB-468 human breast cancer and HaCaT immortalized human keratinocyte cells. It is known that both cell lines exhibited significant Cetuximab cytotoxicity (47) due to their intrinsic dependence on the EGF signaling cascade for growth. The dose-response curves of Cetuximab, Cet-T22d35 and T22d35 of the HaCaT ( Figure 2G , left panel) and MDA-MD-468 ( Figure 2G , right panel) cell lines show a similar cytotoxic response to both Cetuximab and Cet-T22d35 with calculated IC₅₀ values ( Table 4 ) in the 0.2-1.4 nM range, while T22d35 elicited no cytotoxic effects. These results further confirm that fusing a TβRII-ED to Cetuximab, and likely other antibody s, does not interfere with the function of antibody itself (https://en.wikipedia.org/wiki/Cetuximab).

To determine whether the Cet-T2m remains intact in circulation, we carried out a PK study in normal, healthy BALB/c mice. Animals were injected with a single dose of Cet-T2m, and the collected serum (15 min to 168 h post-injection) was analyzed by LC-MS/MS MRM using peptides specific for the Fc fragment of Cetuximab (‘ALP’ peptide) and the fused T2m (‘LPY’ peptide). Table 5 shows that the calculated serum concentrations (Beta_hI) for both peptides are very similar, indicating that the Cet-T2m fusion remains intact over time in circulation, with a relatively long circulating half-life of at least 100 h, which is very similar to the mean serum half-life of ~114 h reported for Cetuximab (https://en.wikipedia.org/wiki/Cetuximab).

The TGF-β1 neutralization potency of both the FC5-Fc-T2m and FC5-Fc-T22d35 was compared to that of the Fc-T2m, Fc-T22d35 and the non-Fc-fused T22d35 single-chain trap in the described A549 IL-11 release assay. The data presented ( Figure 3B ), shows that efficient TGF-β1 neutralization was conserved in the both fusions ( Table 6 ), with K_D values of 5.688 pM for FC5-Fc-T2m and 2.351 pM for FC5-Fc-T22d35. These values are in agreement with those for the other trap fusions, and confirms that TGF-β1 neutralization can be achieved to a much higher degree with a T22d35 fusions compared to a Fc-T2m fusion or non-fused T22d35.

We then assessed the ability of the two TGF-β trap fusion proteins, FC5-Fc-T2m and FC5-Fc-T22d35, to undergo receptor-mediated transcytosis across the BBB in vitro. Specifically, we assessed the efficiency and apparent permeability (P_app) of FC5-Fc-T22d35 and FC5-Fc-T2m in comparison to control antibodies including FC5-Fc, T22d35, T2m, and the non-transcytosing A20.1 antibody, raised against Clostridium difficile toxin A with no know mammalian receptor (48). The BBB transcytosis studies were performed using a trans-well setup (49) ( Figure 3C ), wherein the SV-ARBEC cells were seeded on collagen I-coated semi-permeable inserts (42) with the test antibodies being added to the top apical (input) and collected from the top and bottom basolateral (output) chambers at defined time points (15, 30, 45 and 60 min). The transcytosis efficiency and P_app were quantified by highly sensitive multiplexed nanoLC-SRM (31, 43). The transcytosis efficiency was calculated as (output/input) × 100%, representing the percentage of the applied compound that successfully traversed the SV-ARBEC monolayer into the basolateral compartment over the experimental time frame. The transcytosis efficiency for FC5-Fc-T22d35 (top middle) was very similar to the FC5-Fc control (top left) ( Figure 3D ). In contrast, the transcytosis of the FC5-FcT2m (top right) was lower, but much higher than those of T2m (bottom middle) and T22d35 (bottom left) alone which showed negligible BBB permeability, similar to the non-crossing A20.1 control fusion ( Figure 3D ). These results are supported by the P_app values at 60 min, which demonstrate that the permeability of FC5-Fc-T22d35 and the FC5-Fc control were approximately 50% higher than those observed for FC5-Fc-T2m ( Figure 3E ). Nevertheless, the levels of FC5-Fc-T2m detected were still ~4-fold higher than that of the negative controls (T2m, T22d35, and A20.1). Overall, this data thus shows that the FC5 V_HH retains its functional capacity to mediate receptor-dependent transcytosis when fused to TGF-β traps, enabling efficient transport of FC5-Fc-T22d35 across the BBB.

Fusion of the poly-aspartate D10 motif to the N-terminus of the Fc-T2m construct did not impact TGF-β1 neutralization ( Figure 4B , blue graphs), showing an IC₅₀ value that is very similar to the Fc-T2m fusion ( Table 7 ; 3.794 pM versus 2.895 pM, respectively). However, CF770 labelling of the Fc-T2m and D10-Fc-T2m fusion reduced TGF-β1 neutralization by ~4-fold ( Figure 4B , red graphs), which was not entirely surprising. The TβRII-ED/TGF-β binding interface contains several lysine residues, and although our CF770 conjugation strategy was aimed at foremost labeling lysines in the Fc fragment, those in the receptor/ligand interface were likely also labeled, thereby compromising TGF-β neutralization to some extent. It should be noted that unlabeled D10-Fc-T2m and Fc-T2m were used for both the PK and in vivo efficacy studies (described in following sections), hence the TGF-β neutralizing capacity of the constructs used in those studies is thus not compromised.

The bone-binding capacity of the D10-fusions was confirmed by their direct binding to hydroxyapatite, an inorganic mineral present in both human bone and teeth. Transmission electron microscopic images (TEM; Figure 4C ) show an abundance of 14 nm gold-particle-protein A conjugate along the surface (black dots) of the hydroxyapatite crystals that were incubated the D10-Fc-T2m and CF770_D10-Fc-T2m. Only few gold particles were associated with the hydroxyapatite crystals that were incubated with Fc-T2m and CF770_Fc-T2m, which lack the D10 sequence, and the hydroxyapatite that was incubated with gold-particle-protein A alone (conjugate negative control). These data indicates that the D10 moiety retains its bone-binding characteristics in the context of the trap fusions.

To evaluate the ability of the D10 fusions to home to and be retained in the bone we injected healthy mice with CF770_Fc-T2m or CF770_D10-Fc-T2m. Figure 4D shows representative images of mice (n=3) injected with a single dose of either CF770_Fc-T2m (top) or CF770_D10-Fc-T2m (bottom). These results demonstrate that addition of the D10 peptide to the N-terminus of the Fc-T2m significantly enhances bone localization and retention, and images taken between 5 min to 120 h post-injection show a clear accumulation of CF770_D10-Fc-T2m, but not CF770_Fc-T2m, in the skull bones, hind legs and vertebrae. Ex vivo imaging of the brain, heart, lungs, liver, kidneys, spleen and hind legs 120h post-injection, and the average Fluorescent Intensity (FI) for the indicated organs further confirms the specific accumulation of the D10-Fc-T2m but not the Fc-T2m in the bones of the hindlimbs ( Figures 4E, F ). The fluorescent signals observed in the kidneys and liver are similar for both fusions, indicating that accumulation in these organs is not driven, nor affected by the presence of the D10 sequence.

To determine the serum half-life of the Fc-T2m and D10-Fc-T2m trap fusions, we carried out an in vivo PK study in normal, healthy BALB/c mice. Mice were injected with a single dose of Fc-T2m or D10-Fc-T2m and the collected serum samples (15 min to 168 h post-injection) were analyzed by LC-MS/MS MRM using the method described. Our results indicate that the D10-Fc-T2m has a much shorter serum half-live than Fc-T2m (~50 h and ~130 h, respectively; Table 8 ), the latter of which is similar to the value calculated for Cet-T2m ( Table 5 ; ~120 h). The shorter serum half-life of D10-Fc-T2m is likely attributed to the presence of the poly-aspartate motif causing its accumulation and retention over time in the mouse bones ( Figures 4D-F ).

To further investigate whether the D10-Fc-T2m fusion could affect the formation of breast cancer bone metastases, we used MDA-MB-231 human breast cancer cells that were implanted in the mammary fat pad (MFP) of NOD SCID gamma (NSG) mice as an in vivo model for spontaneous bone metastasis. Starting one day post tumor cell implantation, Fc-T2m or D10-Fc-T2m trap fusions were administered intravenously (iv, tail vein) each week, using PBS as a negative control ( Figure 4G ). The growth of primary mammary tumors was similar for the PBS and D10-Fc-T2m infused mice; whereas Fc-T2m infused mice exhibited a slight delay in tumor growth ( Figure 4H ). This is likely due to the fact the effective circulating Fc-T2m levels are higher than the D10-Fc-T2m (due to the fact that D10-Fc-T2m accumulates in the bone), which correlates with its PK profile ( Table 8 ). Tumors were resected once mammary tumors reached approximately 1 cm³ in size, and bone metastatic lesions were allowed to form until mice exhibited paralysis. Using μCT imaging, we quantified the degree of bone destruction that could be observed in the proximal tibia of the indicated mice. The osteolytic lesions that are formed in mice receiving either the Fc-T2m or D10-Fc-T2m trap fusions injected mice were significantly smaller and less destructive when compared to the metastases formed in the PBS injected mice ( Figure 4I ). However, the D10-Fc-T2m fusion was significantly more effective in impairing the formation of osteolytic bone lesions than the Fc-T2m trap, which is evident by the higher bone volumes and bone mass density in D10-Fc-T2m treated animals ( Figure 4J, K ). Overall, these results thus suggest that the D10-Fc-T2m trap fusion, which localizes and accumulates to the bone surface, can effectively impair the formation of osteolytic bone metastatic lesions.