The fibronectin molecule that has advanced the furthest clinically is CT-322 (BMS-844203, pegdinetanib), a vascular endothelial growth factor receptor 2 binder (VEGFR2) developed by Adnexus Therapeutics (Bristol-Myers Squibb). The lead molecule was discovered by mRNA display [15] and affinity matured for human and mouse cross-reactivity [19]. The evolved protein was modified in several ways as follows: (a) seven amino acids at the N-terminal end, VSDVPRD, were removed; (b) the natural seven-amino-acid linker following the tenth type III domain, EIDKPSQ, was introduced; (c) the serine in the C-terminal tail was mutated to cysteine; and (d) the unique cysteine introduced at the penultimate residue was conjugated to a 40-kDa branched poly (ethylene glycol) (PEG). The resultant molecule has 11 nM affinity for VEGFR2 and negligible cross-reactivity for VEGFR1 or VEGFR3 [20]. PEGylation slows the plasma clearance from a terminal half-time of 4 ± 2 h for native protein to 17 ± 2 h with 20-kDa PEG and 50 ± 20 h with 40-kDa PEG.

In the initial publication of in vivo results for CT-322, the molecule reduced the growth of orthotopic MiaPaCa-xenografted tumors by 42 % in nude mice [21]. Physiological manifestations included increased apoptosis, reduced microvessel density, and reduced VEGF-activated blood vessels. However, metastases were still observed at the same frequency as control treatment. Conversely, in a syngeneic Pan02 xenograft model in immunocompetent mice, CT-322 eliminated metastases although the 23 % reduction in primary tumor size was not statistically significant. Combination therapy with gemcitabine elevated efficacy to a 65 % reduction in tumor size, further validating the therapeutic potential of the evolved scaffold.

Evaluation in a glioblastoma xenograft model in nude mice demonstrated a 45 % reduction in tumor growth [20]. Increasing the dose to 30 mg/kg every other day yielded inhibition of tumor growth and microvessel generation equivalent to anti-VEGFR2 antibody DC101. CT-322 was also compared to small-molecule receptor tyrosine kinase inhibitors. Colo-205 xenograft growth inhibition by CT-322 (30 mg/kg biweekly) was comparable to sorafenib and suritinib but with reduced skin rash, weight loss, and death. Vascular growth was significantly inhibited [22]. CT-322 (60 mg/kg thrice weekly) was comparable to mTor inhibitor temsirolimus, and the benefits were partially additive. Also, in a separate model, the incidence of breast cancer metastasis was inhibited 64 % by CT-322.

Evaluation of CT-322 in an intracranial glioblastoma murine model further validated its efficacy [23]. Monotherapy decreased U87 xenograft tumor size and increased survival from 19 to 29 days. Moreover, CT-322 boosts therapeutic efficacy of the DNA-alkylating agent temozolomide by further decreasing tumor size and extending survival from 32 to 47 days. Immunohistochemical staining demonstrates inhibited microvessel formation, increased apoptosis, and decreased proliferation.

Therapeutic efficacy in multiple murine models with concomitant validation of physiological mechanisms motivated further clinical translation of CT-322. Toxicology studies indicated no adverse effects at 5 and 10 mg/kg in rats and monkeys [24], further supporting human trials. A phase I trial identified a maximum tolerated dose of 2 mg/kg intravenously weekly or biweekly, which yielded a lack of severe toxicity and a clearance half-time of 71–98 h [24]. 82 % of patients developed anti-drug antibodies that were shown to bind to the engineered loops of CT-322, which are of synthetic, non-human origin. Nevertheless, the response did not impact CT-322 or VEGF-A concentrations. Of the 34 patients available for therapeutic evaluation, 68 % had stable disease and 12 % demonstrated a reduction in tumor size.

In a phase II study for first-line treatment of advanced non-small-cell lung cancer, patients received paclitaxel/carboplatin plus either CT-322 or bevacizumab (antibody targeting VEGF-A) [25]. Progression-free survival (5.6 months vs. 6.8 months; CT-322 vs. bevacizumab), overall survival (12.5 months vs. 15.2 months), and response rate (25 % vs. 33 %) were all lower with CT-322 than with bevacizumab. Thrombocytopenia and grade ≥3 hypertension increased with CT-322. The efficacy of CT-322 in other clinical indications awaits further study.

An additional growth factor receptor antagonist, a type I insulin-like growth factor receptor (IGF1R) binder with 110 pM affinity, has been studied [26]. This molecule, PEGylated akin to CT-322, reduced vascular surface area by 59 % and reduced tumor growth by 37 % in a xenograft mouse model of Ewing’s sarcoma. Combination treatment with CT-322 did not yield statistically significant gains in tumor inhibition.

This agent was also genetically fused to an epidermal growth factor receptor (EGFR)-targeted fibronectin [27]. PEGylation reduced the affinity of both components 10–20 fold though strong parental affinity yielded 1 and 10 nM final affinity for IGF1R and EGFR. The bispecific agent inhibits H292 lung cancer xenografted tumors in mice compared to EGFR-targeted panitumumab. More notably, the bispecific fibronectin construct inhibits RH41 pediatric rhabdomyosarcoma xenografted tumors, which are not effectively treated by panitumumab. In BxPC3 xenografted tumors, the bispecific yielded greater tumor growth inhibition than EGFR-targeted cetuximab or even a mixture of monospecific fibronectin domains. Thus, bispecific agents may provide benefits over multi-agent treatment approaches. Notably, the fibronectin-based bispecific agent was a simple genetic fusion with a (GS)₁₀ linker and was stable (T _m  = 58 °C) and monomeric. ¹⁴C labeling of the domain along with whole-body autoradiography delineated physiological distribution including variable tumor uptake (1–3 %ID/g in RH41 vs. 6–11 %ID/g in H292) [28].

Research on an alternative multi-specific agent provides data on one possible mechanism for improved efficacy relative to monospecific agents. An antibody–fibronectin construct targeting three independent EGFR epitopes clusters receptors, yielding downregulation of surface EGFR. This construct dramatically reduced the growth of BRAF and KRAS xenografted tumors in mice [29].

On the diagnostic front, the fibronectin domain has demonstrated effective imaging in molecular positron emission tomography (PET). A fibronectin domain engineered for binding EGFR was radiolabeled with ⁶⁴Cu using a 1,4,7,10-tetra-azacyclododecane-1,4,7,10-tetraacetic acid (DOTA) chelator [30]. The resultant molecule is stable in mouse and human serum for at least 24 h. Molecular PET exhibits EGFR-specific tumor uptake within ~20 min and rapid clearance from blood (half-time = 2.1 min) and healthy tissue. 3.4 ± 1.0 %ID/g was achieved in tumor at 1 h post-injection with a 9:1 tumor:background ratio. The primary limitation of the molecule is high renal uptake and retention although the 2.1 mGy/MBq renal dose is still only 2 % of the maximum tolerated dose. Moreover, preliminary work indicates that using an ¹⁸F-labeling approach substantially reduces renal signal. The higher positron yield and more rapid radioactive decay (half-life = 1.8 h) make ¹⁸F a preferable isotope for clinical use. The impact of protein scaffold hydrophilicity and charge on physiological distribution was elucidated using molecular PET to analyze delivery of fibronectin mutants [31]. Renal retention was reduced by selective removal of charged residues. Hydrophilicity reduces hepatic retention but elevates renal retention.

Fibronectin domains have been effectively used as monovalent antagonists, multi-specific fusions, immunotoxins in cell culture [32], and molecular PET diagnostics. A PEGylated fibronectin has proven safe and biologically active in patients. Additional study will continue to identify the efficacy of this compound and others relative to alternative therapeutic molecules in various indications including cardiovascular disease [33]. The small size and rapid clearance of non-PEGylated fibronectins are well suited for molecular imaging, which has demonstrated preclinical efficacy. Efficacy in both therapy and imaging coupled with robust, efficient evolution of stable, high affinity binders to a multitude of targets poise the fibronectin scaffold for continued successful development.

Several natural knottins have been explored for clinical use. For example, ziconotide, isolated from the cone snail, is an N-type calcium channel antagonist. It is approved by the United States Food and Drug Administration (FDA) as an analgesic for severe chronic pain. It has also demonstrated 44 % reduction in ischemic cerebral damage in a rabbit model [38]. In mouse models of cachexia, a 50-amino-acid portion of agouti-related peptide (AgRP) targeting melanocortin receptor has effectively inhibited appetite suppression and weight loss induced by radiation or tumor growth [39]. The 36-amino-acid chlorotoxin targets annexin A2, resulting in preferential binding to multiple malignant cell types relative to healthy tissue. Conjugation with ¹³¹I enables single-photon emission computed tomography (SPECT) imaging of glioma in human patients [40] and has completed phase I and II trials for radiotherapy [41, 42]. Conjugation of the near-infrared fluorophore Cy5.5 has enabled fluorescence molecular imaging of xenografted glioma, spontaneous medulloblastoma, primary prostate tumors, and lung metastases in mice [43]. Decoration of superparamagnetic iron oxide nanoparticles with chlorotoxin permits molecular magnetic resonance imaging [44].

Synthetic knottins have been engineered by grafting binding motifs into knottin loops and evolving the grafted peptides. Thrombopoietin peptide mimics were grafted into knottin frameworks based on AgRP or Ecbalium elaterium trypsin inhibitor II (EETI-II) and crosslinked [45]. The resultant dimers stimulated proliferation of megakaryoblasts. A 200 ng dose of an EETI-II-based dimer was able to double the platelet count in a mouse model.

The Cochran lab in conjunction with the Cheng and Gambhir labs has performed extensive murine studies on knottins for molecular imaging of integrins. The RGD tripeptide motif was inserted into loop 1 of the EETI-II scaffold, and directed evolution with yeast surface display was used to identify clones with ~20 nM affinity for integrin α_vβ₃ [46]. These clones were then used for development of molecular imaging agents. Conjugation with Cy5.5 or ⁶⁴Cu enabled near-infrared fluorescence imaging or PET of xenografted U87 tumors in nude mice [47]. At 4 h post-injection, 4 %ID/g tumor and a 20:1 tumor:muscle ratio were achieved for the radiolabeled agent. Simultaneous labeling with the fluorophore and radioisotope was also achieved while retaining activity [48]. To better match the rapid biodistribution and clearance of the 4-kDa knottin, clone 2.5D was radiolabeled with ¹⁸F using N-succinimidyl-4-fluorobenzoate chemistry [49]. Effective tumor imaging was retained although tumor uptake (2.6 ± 0.7 %ID/g at 0.5 h, 1.5 ± 0.4 %ID/g at 1 h), and specificity (5:1 tumor:muscle at 0.5 h, 3:1 tumor:muscle at 1 h) reduced slightly. ⁶⁴Cu-labeled clone 2.5F exhibited superior specificity relative to ¹⁸F-fluorodeoxyglucose for imaging nascent lung nodules in a transgenic mouse model (6.0 ± 0.6 tumor:background for ⁶⁴Cu-knottin vs. 4.4 ± 0.7 for ¹⁸F-fluorodeoxyglucose) [50]. The 2.5D clone was also used for molecular ultrasound [51]. Conjugation to perfluorocarbon-filled microbubbles yielded specific detection of integrin α_vβ₃-expressing tumors in murine models. Targeting was superior to microbubbles labeled with integrin α_vβ₃ antibodies or cyclic RGD peptides and was molecularly specific as verified by blocking and cellular control studies.

The diversity of the knottin family was also demonstrated as the AgRP scaffold was used to engineer nanomolar affinity binders to integrin α_vβ₃ [52] and label them with ⁶⁴Cu for PET [53]. Imaging of xenografted U87 tumors in nude mice was effective although tumor uptake was slightly lower (2.7 ± 1.1 %ID/g at 1 h vs. 4.6 ± 0.6 %ID/g) and renal signal much higher (60 ± 18 %ID/g vs. 4 ± 1 %ID/g) than with EETI-II. ¹¹¹In labeling reduced renal signal to 34 ± 8 %ID/g but hindered tumor retention [54]. ¹⁸F labeling using 4-nitrophenyl-2-fluoropropionate reduced renal signal to 20 %ID/g [55]. Evolution of an additional scaffold, the Momordica cochinchinensis trypsin inhibitor-II (MCoTI-II), further substantiated the breadth of the knottin scaffold, including application to a new integrin target, α_vβ₆ [56]. Clones with low nanomolar affinity were evolved by loop grafting and directed evolution with yeast surface display. Ligands were labeled with ⁶⁴Cu for molecular PET of nude mice with subcutaneous or orthotopic xenografted pancreatic cancers. The original, arginine-rich scaffold yielded 5 ± 1 %ID/g tumor with an 8 ± 2 tumor:background ratio at 1 h post-injection albeit with 75 ± 5 %ID/g kidney. Transfer of the evolved loop sequence to a serine-rich scaffold decreased renal uptake to 18 ± 3 %ID/g but also decreased tumor uptake to 1.8 ± 0.4 %ID/g. Renal retention was substantially reduced with ¹⁸F radiolabeling using N-succinimidyl-4-fluorobenzoate chemistry (9 ± 3 %ID/g kidney) albeit with slight tumor reduction (2.3 ± 0.6 %ID/g) [57].

Radiolabeled knottins were also evaluated for radiotherapy. ¹⁷⁷Lu-labeled integrin α_vβ₃ binders 2.5D and 2.5F were administered to mice bearing U87 tumor xenografts [58]. 61 mGy/mCi absorbed dose was achieved in the tumor with 18 mGy/mCi in the kidneys and <2 mGy/mCi elsewhere.

Natural knottins have demonstrated efficacy in a range of applications but are limited to targeting their native binding partners. Loop grafting and evolution in multiple knottin frameworks have proven highly efficacious for integrin targeting in a variety of murine models, primarily across multiple modalities of molecular imaging but also for radiotherapy. The breadth of impact for the knottin family will depend on the ability to robustly and efficiently develop high-affinity binders to a multitude of targets. To aid in affinity, an integrin binding motif was simultaneously grafted into two loops of the EETI-II scaffold, which enhanced binding and maintained effective molecular PET [59]. Looking to expand the evolutionary capacity of the scaffolds beyond integrins, the EETI-II scaffold has been evolved for binding to seven different targets albeit with limited affinity evaluation [60]. If high affinity can be readily achieved from naïve libraries without grafted motifs, the stable, small knottin family of scaffolds has tremendous potential for molecularly targeted imaging and therapy.

DARPins have demonstrated efficacy in a variety of applications. A DARPin was evolved by ribosome display to achieve 90 pM affinity for human epidermal growth factor receptor 2 (HER2) [61]. The ligand was labeled with ^99mTc and yielded effective localization of a murine tumor xenograft via SPECT/CT [62]. Biodistribution quantitation indicated 9 ± 2 %ID/g tumor with 13 ± 3 tumor:blood at 1 h; 8 ± 3 %ID/g tumor, 27 ± 15 tumor:blood at 4 h; and 8 ± 2 %ID/g tumor, 62 ± 28 tumor:blood at 24 h. Kidney and liver backgrounds were 239 ± 33 %ID/g and 7.3 ± 1.0 %ID/g, respectively, at 4 h. Extensive characterization was performed to quantify the impacts of affinity and size. While protein dimerization had a nearly negligible impact on plasma clearance (half-time of ~3 min), PEGylation drastically slowed clearance (half-time = 19 h for 20 kDa PEG; 50 h for 40 and 60 kDa PEG). This pharmacokinetic modulation decreased tumor:blood ratio from 13 ± 3, 27 ± 15, and 72 ± 27 tumor:blood at 1, 4, and 24 h post-injection for native protein to 0.2 ± 0.1, 0.6 ± 0.3, and 4.7 ± 3.1 for the PEGylated protein. Tumor uptake correlated with affinity for (a) the 14.5 kDa native proteins from 1 to 48 h and (b) the PEGylated constructs at later time points.

The 90 pM HER2 binder was also efficacious for in vitro diagnostics. The DARPin exhibited superior sensitivity and specificity relative to a validated antibody in immunohistochemistry of breast cancer tissue microarrays [63]. Picomolar affinity was critical to success as a DARPin mutant with 1 nM affinity was significantly less effective.

HER2 DARPins were also effective in viral targeting. Fusion of a hemagglutinin mutant with an anti-HER2 DARPin enabled molecular targeting of lentiviral vectors [64]. Vector particles were systemically administered into mice with subcutaneous tumor xenografts of HER2-positive (SKOV3) or HER2-negative (U87MG) tumors. Using a luciferase reporter, HER2-specific viral targeting was evident within 3 days and was retained for at least 14 days. Oncolytic recombinant measles virus was also effectively targeted with an anti-HER2 DARPin [65]. Tumor inhibition following intratumoral injection of a DARPin-targeted virus was superior to an antibody-targeted virus; inhibition was comparable to untargeted virus, but without associated toxic side effects. Also, a dual-targeted (HER2 and epithelial cell adhesion molecule (EpCAM)) bispecific virus demonstrated improved resistance to tumor modulation in cell culture.