De novo protein design of calreticulin binders for theranostic applications

Background: Calreticulin (CALR) is an endoplasmic reticulum (ER) protein that can translocate to tumor cell surfaces during partial ER stress. Surface CALR is largely absent on normal cells yet upregulated on many types of tumor cells undergoing stress, making it a tumor-specific target with potential ubiquity and application as a tumor-agnostic biomarker. Targeting surface CALR with theranostic agents could offer a novel strategy to visualize tumors and monitor treatment response. We hypothesized that de novo protein design could generate a novel CALR-specific theranostic agent. Methods: We employed the de novo protein design platform BindCraft to generate small (<10kDa) high-affinity binders against CALR. A streamlined in silico to in vitro screening pipeline enabled isolation of CALR binders within two weeks. Binding affinity and specificity were validated by biolayer interferometry (BLI) and flow cytometry. For flow cytometry, human pancreatic cancer cells were treated with doxorubicin, a known surface CALR inducer. Cells were incubated with two CALR binders for 30 minutes, followed by staining with fluorescent antibody. In a Matrigel plug model, mice bearing CALR and control plugs were injected with 100 μCi of [ 64 Cu]Cu-NOTA-binder and imaged by dynamic PET for 1 hour followed by CT. Results: BLI against CALR identified two high affinity binders, d4 and d6, with calculated dissociation constants of 83.2 nM ( k on = 512790 M -1 s -1 , k off = 0.04265 s -1 ) and 31.2 nM ( k on = 551797 M -1 s -1 , k off = 0.01723 s -1 ), respectively. Flow cytometry showed both CALR binders, d4 and d6, detected increases in surface CALR of 1.6-fold and 2.1-fold, respectively, on doxorubicin-treated cancer cells compared to vehicle. PET imaging revealed significantly higher uptake of [ 64 Cu]Cu-NOTA-d6 in CALR plugs compared to controls as early as 20 minutes post-injection, with sustained retention throughout the imaging period. At 1 hr post-injection, SUV mean of the CALR plug was (0.882 ± 0.270) compared to (0.344 ± 0.048) for the control plug ( p = 0.0272). [ 64 Cu]Cu-NOTA-d6 was cleared renally with insignificant uptake in non-target organs compared to blood pool and an average blood ½-life of 4.2 minutes. Conclusions: This study demonstrates the feasibility of de novo protein design to rapidly generate high-affinity CALR-targeted binders suitable for theranostic applications. Further in vivo studies will evaluate tumor-specific uptake and long-term retention of d6 in preclinical cancer models to support its translational development as a theranostic agent.