A Fab of trastuzumab to treat HER2 overexpressing breast cancer brain metastases

HER2 breast cancer brain metastases are challenging daily practice because of the limited passage of the blood-brain-barrier (BBB) by most treatments [1]. In a pilot pharmacological study, a direct administration of the anti-HER2 antibody trastuzumab into the cerebrospinal fluid (CSF) efficaciously treated HER2-overexpressing breast cancer brain metastases, despite a major brain-to-blood efflux of this therapeutic IgG [2]. We hypothesized that an Fc receptor (FcRn) expressed by endothelial cells of the BBB was responsible for this efflux and that engineering of a Fab antibody would overcome this limitation [3, 4].

Using immunostainings, FcRn receptors were expressed in meningeal and choroid plexus of rats, and mainly at the abluminal side of endothelial cells on human brains (Suppl. Figure 1 A-D). We then did a proof-of-concept study using a commercialized anti-VEGF Fab antibody (ranibizumab) and its corresponding full IgG1 (bevacizumab). We implemented an ultra-sensitive ELISA method to measure ranibizumab concentrations (Suppl. Figure 2 A-B), and an original surgical approach of catheter implantation in the rat cisterna magna to directly inject antibodies into the CSF and perform pharmacokinetic studies (Suppl. Figure 3 A-B). We confirmed the CSF-to-blood rapid clearance of bevacizumab but not of ranibizumab, detected near cerebellar Purkinje cells where VEGF is physiologically expressed [5] (Suppl. Figure 3 C-D).

To target HER2-overexpressing brain metastases, we engineered two Fab of trastuzumab, namely Fab#1 and Fab#2, produced respectively by BIOTEM® and in our research unit (Fig. 1A and Suppl. Figure 4 A). Using fluorescent-labeled antibodies, we confirmed that trastuzumab and the two Fab efficiently bound HER2-overexpressing BT474 cells (Fig. 1B). Flow cytometry showed no significant difference in binding between the three antibodies, and their IC₅₀ was identical, of 8 µg/mL. Inhibition proliferation tests at 8 µg/mL showed similar profiles for the three antibodies (Suppl. Figure 4B-D).

In a patient-derived xenograft model of HER2-overexpressing breast cancer (Suppl. Figure 5), intravenous administration of Fab antibodies or trastuzumab at 2 mg/kg/week significantly inhibited tumor growth compared to untreated mice, with no significant difference between the three antibodies (Fig. 1C). Using Ki67 and CD31 immunostainings, treated mice showed a decreased cancer cell proliferation and microvessel density in tumors (Fig. 1D-E), without any increase in necrotic areas (Fig. 1F), in accordance with a well-known neoangiogenesis inhibition of trastuzumab and not a direct cytotoxic effect on tumor endothelial cells [6]. Pharmacological analysis showed higher serum concentrations at 30 min for trastuzumab compared to Fab#1 antibody (Suppl. Figure 6). When we assessed drug toxicity on normal tissues, no histological damage was identified. We assessed biomarkers of cardiac toxicity, the only known toxicity of trastuzumab [7]. BNP and Adrenomedullin mRNA expressions increased in mice treated with trastuzumab or Fab compared to untreated mice (Suppl. Figure 7 A-B), whereas dipeptidyl-peptidase-3 and cleaved-caspase-3 protein expression was unchanged (Supp. Figure 7 C-D).

After implementing an ultra-sensitive ELISA method to assess low concentrations of the Fab#1 in fluids (Suppl. Figure 8 A-B), two different doses of trastuzumab or Fab#1 were administered in the cisterna magna of rats (36.5 µg first, corresponding to the maximal dose of Fab injected in a volume of 100 µL, and then 1400 µg of a much more concentrated Fab). For the two doses, trastuzumab CSF concentrations rapidly decreased and were almost undetectable at 4 h while it accumulated in the blood. In contrast, and particularly at the higher dose of 1400 µg, CSF concentrations of the Fab#1 remained stable at 4 h with minimal serum detection (Fig. 2A-D). One hour after CSF administration, Fab#1 was almost undetectable in normal lung, liver or kidney (Suppl. Figure 9). When we pooled the 17 animals, the CSF-to-blood ratio steadily increased over time for trastuzumab but remained 10 times lower for the Fab#1 (Fig. 2E). To further investigate the CSF-to-blood efflux, we developed a two-compartment pharmacokinetic model using a population approach (Fig. 5F). We named k₁₀ as the diffusion constant from CSF to brain, k₁₂ as the diffusion constant from CSF to blood, and k₂₀ as the elimination from serum constant. Serum half-life was shorter for Fab#1 compared to trastuzumab (Suppl. Table 1). The k₁₂/k₁₀ ratio suggested greater brain penetration of Fab#1 compared to trastuzumab. In addition, the calculated partition coefficient Kp_{u, ubrain} [8, 9] from brain samples obtained at euthanasia was 12.3% for trastuzumab and 22.7% for Fab#1 (Fig. 2G). Notably, in deeper brain areas (CE1, Fig. 2H), the Kp_{u, ubrain} for Fab#1 was 2.7 times higher than for trastuzumab, indicating enhanced brain penetration (Suppl. Table 1).

In this preclinical study, we successfully engineered two anti-HER2 Fab antibodies and demonstrated their safety and equal efficacy to trastuzumab for the treatment of HER2 overexpressing breast cancer. Our original pharmacokinetic study with intra-ventricular administration of a Fab antibody demonstrated a limited CSF-to-blood efflux and increased brain penetration. Indeed, our anti-HER2 Fab can penetrate deeper brain parenchyma, which shall be explained by its smaller size and the constant production of CSF by the choroid plexus that creates a pressure directing the fluid flow through the ventricular system to the subarachnoid space [10]. Like for xenobiotics, the Fab probably diffuses via the glymphatic system into the cerebral interstitium by convective system, before later elimination into the venous circulation (Graphical abstract) [4]. This innovative approach is promising in view of the development of Fab antibodies for the treatment of brain malignancies but also neurodegenerative diseases [11, 12].