Research ArticleTherapeutic Antibodies

Addressing Safety Liabilities of TfR Bispecific Antibodies That Cross the Blood-Brain Barrier

See allHide authors and affiliations

Science Translational Medicine  01 May 2013:
Vol. 5, Issue 183, pp. 183ra57
DOI: 10.1126/scitranslmed.3005338


Bispecific antibodies using the transferrin receptor (TfR) have shown promise for boosting antibody uptake in brain. Nevertheless, there are limited data on the therapeutic properties including safety liabilities that will enable successful development of TfR-based therapeutics. We evaluate TfR/BACE1 bispecific antibody variants in mouse and show that reducing TfR binding affinity improves not only brain uptake but also peripheral exposure and the safety profile of these antibodies. We identify and seek to address liabilities of targeting TfR with antibodies, namely, acute clinical signs and decreased circulating reticulocytes observed after dosing. By eliminating Fc effector function, we ameliorated the acute clinical signs and partially rescued a reduction in reticulocytes. Furthermore, we show that complement mediates a residual decrease in reticulocytes observed after Fc effector function is eliminated. These data raise important safety concerns and potential mitigation strategies for the development of TfR-based therapies that are designed to cross the blood-brain barrier.


The blood-brain barrier (BBB) continues to hinder the development of therapies for neurological disease. Although efforts to engineer antibodies to cross the BBB have received significant attention (13), this approach has been met with the challenge of demonstrating sustained uptake and associated activity in brain after therapeutic dosing. Whether the pathways being used for brain uptake of molecules have sufficient capacity and/or represent safety limitations to drug development are also unclear. Answering these questions will be essential to identifying areas where molecular engineering may facilitate delivery of antibodies to the brain.

Transferrin receptor (TfR) has been widely studied as a BBB transport molecule. Jefferies et al. (4) were first to report prominent staining of TfR in brain endothelium and proposed that TfR may function to transport transferrin into brain. Numerous groups have since investigated targeting TfR as a potential receptor-mediated transcytosis vehicle for BBB transport (59). However, recent work has concluded that antibodies that bind to TfR are retained in brain endothelium and do not penetrate the central nervous system to appreciable concentrations (1012). By lowering the affinity of antibodies to TfR, we have been able to demonstrate brain uptake and biodistribution of both bivalent TfR antibodies and a bispecific TfR/BACE1 antibody (13). Using BACE1 as a therapeutic target allowed us to acutely measure brain penetration and associated activity of a BBB-crossing bispecific antibody via analysis of amyloid-β (Aβ) concentrations (13, 14). This understanding of the pharmacokinetic (PK) and pharmacodynamic (PD) relationship, combined with a substantial improvement in brain distribution of lower-affinity TfR-targeting therapeutics, established the foundation to expand the mechanistic insights into this approach.

By engineering additional TfR/BACE1 bispecific antibody variants, the roles of affinity for TfR and Fc effector function on drug activity and safety were examined to identify an optimal therapeutic profile. We find that binding affinity to TfR markedly affects the PK/PD profile. Our investigation of the safety profile of TfR-targeting therapeutic antibodies in mice uncovered a liability previously unrecognized, namely, acute clinical signs and a decrease in circulating reticulocyte count after anti-TfR dosing. Through Fc engineering and genetically assessing various pathways, we describe the anti-TfR properties needed to mitigate these liabilities. We also compare TfR expression on circulating reticulocytes in mice, monkey, and humans to assess the potential liability of using TfR approaches in primates. Together, these data serve as important warnings and lend key considerations for future translation of TfR-targeting therapies focused on enhancing antibody uptake in brain.


Reducing TfR affinity of bispecific anti-TfR/BACE1 improves peripheral exposure leading to sustained brain Aβ reduction

BBB uptake and associated PD responses in brain of TfR/BACE1 bispecific antibodies were explored as a function of anti-TfR affinity. Traditional antibodies bind to a single antigen with both variable regions and are thus bivalent and monospecific. In contrast, the two variable domains of bispecific antibodies each bind a different antigen, often with lower affinity, because monovalent binding can lead to a loss of avidity. For example, in our previous work, the reformatting of bivalent anti-TfRA into a bispecific and monovalent TfRA/BACE1 antibody resulted in reduced TfR affinity and increased brain uptake (13). Here, we explored the effect of varying monovalent affinity for TfR in a bispecific format while keeping the anti-BACE1 arm constant. The TfR affinity of each bispecific TfR/BACE1 variant was reduced relative to the bivalent TfR antibody (Fig. 1A). The affinities of the bispecific anti-TfRA/BACE1, anti-TfRD/BACE1, and anti-TfRE/BACE1 for murine TfR were determined using a competitive enzyme-linked immunosorbent assay (ELISA) [median inhibitory concentration (IC50) values of 18 nM, 588 nM, and ~100 μM, respectively]. These variants enabled us to assess PK/PD properties in mice as a function of a wide range of TfR binding affinities.

Fig. 1 Reducing anti-TfR/BACE1 affinity improves peripheral exposure and brain A β reduction.

(A) The anti-TfR competitive ELISA assay reveals distinct affinity differences of anti-TfR (bivalent TfR binding) and anti-TfR/BACE1 variants (monovalent TfR binding) for TfR. (B to E) Mean ± SEM of plasma (B) and brain antibody (D) concentrations, and mean ± SEM of Aβx–40 concentrations in plasma (C) and brain (E) measured in wild-type (WT) mice after an intravenous injection (50 mg/kg) of control IgG, anti-BACE1, or anti-TfR/BACE1 variants (n = 6 per group).

Peripheral exposure and BACE1 activity, as measured by Aβ concentrations, were evaluated in mice using a single intravenous dose (50 mg/kg) for each bispecific antibody variant. In Fig. 1B, peripheral exposure substantially improved as affinity for TfR was reduced. Coincident with an increase in peripheral exposure, a prolonged reduction in plasma Aβ was also observed with lower-affinity anti-TfRD/BACE1 and anti-TfRE/BACE1 variants compared to anti-TfRA/BACE1 (Fig. 1C). When assessing brain concentration of antibody, similar results were observed for each of the anti-TfR/BACE1 variants (Fig. 1D), with the exception that anti-TfRE/BACE1 showed reduced maximal brain uptake, whereas anti-TfRD/BACE1 exhibited prolonged exposure in brain. Consistent with the increase in duration of brain exposure, the most robust and sustained reduction in brain Aβ was observed with anti-TfRD/BACE1 (Fig. 1E). All bispecific TfR/BACE1 variants displayed higher brain uptake and associated reduction in brain Aβ when compared to either control immunoglobulin G (IgG) or anti-BACE1.

These observations support and extend our previous findings that reduced binding affinity for TfR can improve brain uptake (13). Here, in a bispecific format and with extended time points, exposure in both the periphery (Fig. 1B) and the brain (Fig. 1D) is improved with reduced affinity. Indeed, the high-affinity anti-TfRA/BACE1 was the most rapidly cleared from brain, and the low-affinity anti-TfRE/BACE1 persisted the longest in brain, albeit with reduced maximal uptake. These data point to an optimal TfR affinity because both the total amount of anti-TfRD/BACE1 transported into brain over time (Fig. 1D) and the associated reduction in brain Aβ (Fig. 1E) were superior to higher- and lower-affinity variants of anti-TfR/BACE1. Thus, a fine-tuning of TfR affinity is required to maximize exposure in the periphery and brain. These data also highlight an important point related to BACE1 in vivo potency (IC50 shown in Fig. 1D); by improving BACE1 affinity and associated cellular potency, the in vivo IC50 could be improved, thus allowing for either a reduced dose or an improvement in the duration of Aβ reduction.

Anti-TfR dosing causes acute clinical signs and a decrease in reticulocyte count

Having shown that anti-TfRD/BACE1 had the optimal PK/PD properties, a detailed single-dose safety assessment was conducted comparing anti-TfRD and bispecific anti-TfRD/BACE1. Surprisingly, anti-TfRD dosed at ≥1 mg/kg resulted in acute clinical signs not observed in mice treated with bispecific anti-TfRD/BACE1 at doses as high as 200 mg/kg (Table 1). Within 5 min, all anti–TfRD-treated mice showed profound lethargy characterized by prostration, lack of locomotion and responsiveness, and with a few animals also having occasional limb or whole-body spastic movements. About 20 to 25 min after dosing, these mice were no longer prostrate, but developed a scruffy, hunched appearance. The mice appeared normal within 1 to 2 hours after dosing. Some anti–TfRD-dosed mice were also observed to void reddish brown urine, another observation not seen in bispecific anti–TfRD/BACE1-dosed mice. Given the high TfR expression on immature red blood cells (RBCs) (reticulocytes; Fig. 2A) (15, 16), the acute clinical signs were consistent with hemolysis of circulating red cells, resulting in shock and probable hemoglobinuria.

Table 1 Acute clinical signs observed after monovalent anti-TfR treatment in mice at multiple-dose levels.
View this table:
Fig. 2 Anti-TfR dosing decreases reticulocyte count.

(A) Red cell maturation in the bone marrow proceeds from the pro-erythroblast (Pro-EB), basophilic erythroblast (Baso-EB), polychromatic erythroblast (Poly-EB), orthochromatic erythroblast (Ortho-EB), and reticulocyte. Reticulocytes are released from the bone marrow to the circulation where they mature to RBCs. During later stages of maturation, erythroid precursors synthesize the iron-containing protein hemoglobin, which requires a concomitant increase in TfR expression. TfRs are down-regulated when cells have completed hemoglobin synthesis and cease to proliferate. (B) Percentage of the immature reticulocyte fraction from whole blood of WT mice 1 hour after intravenous injection of anti-TfRD, anti-TfRD/BACE1, or control IgG at the indicated dose (n = 6 per group). (C) Total reticulocyte count 1 or 7 days after intravenous injection of anti-TfRA/BACE1, anti-TfRD/BACE1, or control IgG at the indicated dose (n = 6 per group). (D) Brain Aβx–40 concentrations in WT mice dosed with anti-TfRA/BACE1 (left) or anti-TfRD/BACE1 (right) compared to control IgG. Error bars represent means ± SEM. P values were determined using two-tailed unpaired t test, compared to control IgG. *P < 0.05, **P < 0.001, ***P < 0.0001. n.s., not significant.

To establish whether target binding on reticulocytes underlies the observed acute clinical signs following monospecific anti-TfR, but not bispecific anti-TfR/BACE1 administration, we measured circulating blood cell counts after a single intravenous injection of anti-TfRD, anti-TfRD/BACE1, or control IgG at a range of doses (1, 5, or 50 mg/kg; Fig. 2B). The immature reticulocyte fraction was measured with an automated hematology analyzer on the basis of the fluorescence intensity of polymethine dye bound to cellular RNA (17). At 1 hour after dosing, anti–TfRD-treated animals showed both acute clinical signs as previously described and a marked decrease in immature reticulocyte count (to about 10% of control IgG–treated mice), with a similar magnitude and incidence at all doses tested (≥1 mg/kg; Fig. 2B and Table 1). Mice treated with anti-TfRD/BACE1 (1 or 5 mg/kg) maintained an immature reticulocyte count similar to those observed in control IgG–treated mice. Mice treated with anti-TfRD/BACE1 (50 mg/kg) showed an about 50% reduction in immature reticulocyte count (Fig. 2B and Table 1), with no accompanying clinical signs.

To assess the time course of reduced reticulocyte count and to relate these findings to TfR affinity, we conducted a similar experiment comparing anti-TfRA/BACE1, anti-TfRD/BACE1, and control IgG. Total reticulocytes were counted at 1 and 7 days after single intravenous injection of antibody (5, 25, or 50 mg/kg) (Fig. 2C). Samples were evaluated at 24 hours instead of 1 hour after dosing in this and all subsequent experiments to measure the nadir in total reticulocyte count (Fig. 2B). At 24 hours after dosing, mice dosed with anti-TfRA/BACE1 showed a marked reduction in total reticulocyte count (to less than 10% control) at all doses evaluated (Fig. 2C). Dose level–related reductions in reticulocytes were also observed in mice administered anti-TfRD/BACE1, although reductions in antibody affinity (for example, anti-TfRA to anti-TfRD) as well as dose level appeared to attenuate reticulocyte destruction.

By 7 days after dosing, reticulocyte count had returned to control levels in all anti–TfR/BACE1-treated groups, with the exception of animals given anti-TfRD/BACE1 (50 mg/kg), which had a sustained reduction in reticulocyte count (about 50% relative to controls) (Fig. 2C). These data suggest that the relative affinity of the antibody for TfR at 7 days after dose (anti-TfRA/BACE1 > anti-TfRD/BACE1) was not as important as the persistence of the antibody in the bloodstream for this later time point (fig. S1B; plasma concentration at day 7 for anti-TfRA/BACE1 = 6 nM, anti-TfRD/BACE1 = 255 nM).

As we observed a reduced reticulocyte count dose response (Fig. 2C), we wanted to relate the dose levels to the associated ability of TfR/BACE1 antibodies to reduce Aβ in brain. Sustained reduction in brain Aβ at both 25 and 50 mg/kg for anti-TfRD/BACE1 was observed (Fig. 2D), whereas anti-TfRA/BACE1 showed an acute reduction in brain Aβ at all three doses. These data are consistent with the PK in both periphery and brain (fig. S1) and show that anti-TfRD/BACE1 (25 mg/kg) is sufficient to reduce brain Aβ for greater than 4 days after a single dose.

Elimination of Fc effector function prevents acute clinical signs and mitigates reticulocyte reduction

In addition to TfR affinity and valency, there are several differences between the monospecific anti-TfR and the bispecific anti-TfR/BACE1 variants. Anti-TfRA (Fc+) and anti-TfRD (Fc+) could elicit the acute clinical signs and decreased reticulocyte count through antibody-dependent cell-mediated cytotoxicity (ADCC) and/or complement-dependent cytotoxicity (CDC). In contrast, the bispecific TfR/BACE1 antibodies (Fc) lack glycosylation because of mutations introduced in the Fc region that are required for Fcγ receptor (FcγR) binding (13), and therefore lacked ADCC functionality.

The relative contribution of effector function to the acute clinical signs observed with anti-TfRD,Fc+ was examined by introducing the same Fc mutations present in bispecific TfRD/BACE1 (that is, anti-TfRD,Fc−). Anti-TfRD,Fc− administered to wild-type mice at 1 or 25 mg/kg did not produce acute clinical signs at either dose, although a dose-dependent reduction of reticulocytes was still observed (Fig. 3A). A complementary study in which anti-TfRD,Fc+ or anti-TfRD/BACE1Fc− was administered to mice lacking functional FcγR (18) at 25 mg/kg showed similar results, with elimination of acute clinical signs despite reductions in reticulocyte count (Fig. 3B).

Fig. 3 Elimination of effector function prevents acute clinical signs and partially mitigates reticulocyte reduction.

(A to C) Clinical signs and total reticulocyte count measured in whole blood 1 day after treatment in (A) WT mice dosed with anti-TfRD without (Fc) effector function, (B) FcγR knockout (KO) mice treated with anti-TfRD with (Fc+) or without (Fc) effector function, and (C) WT mice dosed with anti-TfRD/BACE1 bispecific antibody engineered with (Fc+) or without (Fc) effector function. Dose concentrations (mg/kg) are indicated on the graph (n = 6 per group). Error bars represent means ± SEM. P values were determined using two-tailed unpaired t test, compared to control IgG. **P < 0.001, ***P < 0.0001.

These results suggested dissociation between the mechanisms mediating acute clinical signs and the mechanisms leading to decreased reticulocyte count after anti-TfR administration. To further understand the contribution of effector function to reticulocyte reduction, we administered an effector-positive (Fc+) anti-TfRD/BACE1, constructed with a wild-type Fc (anti-TfRD/BACE1Fc+), to mice. A low dose (5 mg/kg) of anti-TfRD/BACE1Fc− was compared with an equivalent dose of anti-TfRD/BACE1Fc+ in wild-type mice (Fig. 3C). Strikingly, the introduction of effector function resulted in acute clinical signs with similar severity to those seen in animals dosed with anti-TfRA,Fc+ or anti-TfRD,Fc+, with a marked decrease in reticulocytes observed with the effector-positive (Fc+) bispecific antibody at a lower dose relative to the effectorless (Fc) version. Furthermore, the rapid appearance of clinical signs and reticulocyte effects, along with the specific observation of reddish brown urine suggesting hemoglobinuria, is consistent with an ADCC-mediated mechanism of intravascular hemolysis.

To confirm these in vivo observations and address cellular mechanism, we assessed in vitro ADCC activity using the high-affinity anti-TfRA,Fc+ and the effectorless anti-TfRA/BACE1 (fig. S2A). Only anti-TfRA,Fc+ induced ADCC in a murine erythroblast cell line expressing high amounts of TfR, consistent with the conclusion that TfR binding combined with maintained anti-TfR effector function (Fc+) was sufficient to induce cellular toxicity.

These data strongly suggest that the acute clinical signs observed in mice are linked to the effector status of the antibody, such that administration of effectorless antibodies in wild-type animals (or dosing in Fcγ knockout mice) can completely mitigate these clinical observations. Furthermore, these results indicate that although effector function is not necessary to drive decreases in reticulocytes, it is sufficient to profoundly reduce reticulocytes at a relatively low dose. These observations also strongly implicate an additional mechanism contributing to the residual decrease in reticulocytes observed in effectorless (Fc) dosing paradigms.

Bispecific antibodies with the BACE1 arm replaced by a control IgG (anti-TfRA/control IgG and anti-TfRD/control IgG) also reduced reticulocyte counts to the same degree as anti-TfRD/BACE1 (Fig. 4A), demonstrating that the BACE1 arm does not contribute to the residual decrease in reticulocyte counts.

Fig. 4 Residual decrease in reticulocyte count is mediated by complement.

(A to C) Total reticulocyte count measured in WT mice (A and B) or C3 KO mice (C) 24 hours after intravenous injection of the indicated antibody and dose (n = 6 per group). (D) Total reticulocyte count measured in C3 KO or WT mice 24 hours after intravenous injection of the anti-TfRD/BACE1 or control IgG. Error bars represent means ± SEM. P values were determined using two-tailed unpaired t test, compared to control IgG. *P < 0.05, **P < 0.001, ***P < 0.0001.

Reducing affinity to TfR mitigates reticulocyte reduction

As previously described in Fig. 2C, affinity and dose contribute to the degree of reticulocyte loss observed after anti-TfR/BACE1 administration. In a follow-up experiment, anti-TfRE/BACE1 had essentially no impact on reticulocytes at either 25 or 50 mg/kg, whereas similar doses of anti-TfRA/BACE1 or anti-TfRD/BACE1 reduced the reticulocyte count (Fig. 4B). On the basis of antibody exposure and associated Aβ reduction described in Fig. 1, anti-TfRE/BACE1 had better sustained plasma exposure and persisted in the brain, but showed less maximal brain uptake relative to higher-affinity anti-TfRD/BACE1. Indeed, anti-TfRE/BACE1 was able to reach the calculated in vivo IC50 for BACE1 inhibition (Fig. 1D) and thus partially reduced Aβ concentrations in brain. In summary, lower dose levels and/or lower TfR affinities will mitigate the residual decreases in reticulocytes; however, if affinity to TfR is too low (for example, anti-TfRE/BACE1), brain penetration and associated activity are limited.

Residual decrease in reticulocyte count is mediated by complement

Antibody effector function is a primary driver of the acute clinical signs and is sufficient to induce robust reticulocyte reductions seen after anti-TfR dosing. As best illustrated in Fig. 3C, when full effector function (Fc+) and effectorless (Fc) TfRD/BACE1 bispecific antibodies are dosed in wild-type mice at subtherapeutic doses (5 mg/kg), effector function was both necessary and sufficient for the acute clinical signs and was sufficient to induce a profound decrease in reticulocyte count. Nevertheless, multiple lines of evidence support the conclusion that mechanisms other than antibody effector function are necessary for the residual decrease in reticulocytes seen with effectorless TfR antibodies, in particular, the observation that reticulocytes are still reduced in FcγR knockout mice dosed with effectorless anti-TfRD/BACE1 (Fig. 3B).

Perplexed by this non–effector function–mediated decrease in reticulocytes, we next explored if the complement cascade is contributing to either the acute clinical signs or the observed reticulocyte reduction. Mice deficient in the C3 complement protein (19) were dosed with high-affinity anti-TfRA,Fc+ (Fig. 4C). Acute clinical signs and a marked reticulocyte reduction were observed in all anti–TfR-dosed C3 knockout mice, indicating that C3, and thus the complement cascade, is not required to elicit these effects when an effector-positive antibody is administered.

We next dosed C3 knockout mice with a high dose of effectorless anti-TfRD/control IgGFc− to determine whether complement mediates the residual reticulocyte reduction. Indeed, residual reticulocyte loss was rescued when both effector function and the complement cascade were eliminated by dosing C3 knockout mice with effectorless TfR bispecific antibodies at a high therapeutic dose (50 mg/kg; Fig. 4D). Thus, complement appears to play a role in the residual reticulocyte reduction after administration of effectorless TfR bispecific antibodies in mice.

Neither effector-positive anti-TfRA,Fc+ nor effectorless anti-TfRA/BACE1Fc− elicited CDC of murine bone marrow cells in an in vitro CDC assay (fig. S2B). Furthermore, when analyzing bone marrow in vivo after dosing either effector-positive (Fc+) or effector-negative (Fc) anti-TfR/BACE1, effectorless anti-TfR/BACE1 (Fc) antibodies only depleted TfR+ cells in circulation, while having no impact on bone marrow TfR+ cells (fig. S3). Therefore, we conclude that complement may mediate the residual reticulocyte reduction in vivo via opsonization of circulating reticulocytes by splenic and liver macrophages (20). Thus, any anti-TfR approach must have reduced effector function because full effector function anti-TfRD/BACE1 (Fc+) robustly depleted TfR+ cells in bone marrow (fig. S3, D and E), a risk that is likely to translate into higher species.

Anti-TfR dosing does not alter BBB permeability to antibodies

Especially relevant to this work, brain endothelial cells also express particularly high levels of TfR (4, 2123). Because high cellular TfR expression rendered reticulocytes susceptible to reduction, we addressed whether anti-TfR could disrupt brain endothelium and alter the integrity of the BBB. We first assessed BBB architectural integrity after dosing with either anti-TfRD,Fc+ or anti-TfRD/BACE1Fc− compared to control IgG at 50 mg/kg. Brains were harvested from mice 24 hours after dosing, a time point where the highest brain concentration of drug had been observed (Fig. 1D). Brain sections were stained with BBB markers, including Glut1 (Fig. 5A), Claudin-5 (Fig. 5B), and zonula occluden-1 (ZO-1) (Fig. 5C), as well as 4′,6-diamidino-2-phenylindole (DAPI) nuclear stain to assess brain endothelial integrity. Anti-TfR dosing, either with or without effector function, did not appear to alter the overall integrity of the brain endothelium as assessed by high-magnification confocal microscopy. There were no qualitative differences in the intensity or localization of BBB markers, and the nuclei of brain endothelial cells were normal in all treatment groups. Nevertheless, a more robust and quantitative measure of function was needed to conclude that the BBB remains intact, particularly to assess the possibility of a nonspecific increase in the permeability of the barrier to antibodies.

Fig. 5 Anti-TfR dosing does not alter BBB architecture and permeability to antibodies.

(A to C) Representative confocal microscopy images show immunohistochemical staining for Glut1, Claudin-5, or ZO-1 in mouse brain tissue 24 hours after a single injection (50 mg/kg) of control IgG, anti-TfRD, or anti-TfRD/BACE1. Scale bar, 20 μm. IV, intravenously. (D to F) Mean antibody concentration in WT mouse brain 24 hours after intravenous injection of either control IgG (50 mg/kg) or co-injected antibodies (25 mg/kg each) as indicated. Detection was based on either (D) a generic human Fc ELISA or (E) a specific BACE1 ectodomain ELISA (see Materials and Methods for details). (F) Quantification of Aβx–40 concentrations in WT mouse brain after intravenous injection of control IgG or co-injected antibodies (n = 6 per group). (G) Evans blue concentration in brain relative to plasma 24 hours after a single injection (50 mg/kg) of control IgG, anti-TfRD, or anti-TfRD/BACE1 (n = 6 per group). Error bars represent means ± SEM. P values were determined using two-tailed unpaired t test, compared to control IgG. *P < 0.05, **P < 0.001, ***P < 0.0001. P values for (E) were determined compared to the control IgG + anti-BACE1 group.

If TfR antibodies were to disrupt the BBB nonspecifically, a coincident increase in brain uptake of other antibodies, such as co-administered anti-BACE1, would also be expected. Five different antibody combinations, outlined in Fig. 5, were dosed and brains were harvested after 24 hours. When total IgG was measured in brain via a human Fc/Fc capture ELISA, an about fivefold increased brain uptake was seen in all groups dosed with anti-TfR or anti-TfR bispecific compared to control IgG and/or anti-BACE1 treatment (Fig. 5D). However, when a BACE1-specific antigen capture ELISA was used to measure anti-BACE1 concentrations in brain, only the co-dosed anti-TfRD/BACE1 group showed substantial brain uptake (Fig. 5E). These data correlated with the observation that only anti-TfRD/BACE1 reduced brain Aβ concentrations (Fig. 5F). The fact that the anti-TfRD (with full effector function) and anti-BACE1 co-dosed mice showed no brain uptake of anti-BACE1 and no reduction of brain Aβ concentrations allows us to conclude that anti-TfR dosing does not disrupt the BBB as a mechanism for antibody uptake in brain. This conclusion is further supported by the fact that there is no significant increase in Evan’s blue leakage when mice are dosed with anti-TfRD (with full effector function) or anti-TfRD/BACE1 (Fig. 5G).

Sustained Aβ reduction observed after multiple doses of anti-TfR/BACE1

All previous studies evaluated the impact of a single TfR/BACE1 bispecific dose on peripheral exposure, brain penetration and anti-BACE1 activity, acute clinical signs, and reticulocyte counts. To ascertain whether TfR is capable of repeatedly transporting antibody into brain, and if different and/or exacerbated effects on TfR-expressing reticulocytes would be observed with repeated anti-TfR dosing over an extended period, we administered anti-TfRD/BACE1 or control IgG to wild-type mice at a dose of 25 mg/kg once per week for a total of 4 weeks. Blood was collected for PK, PD, and hematology at 1, 4, or 7 days after the second and fourth weekly injections. In addition, total bilirubin, serum iron, and total and unsaturated iron binding capacity (a surrogate parameter for serum transferrin) in peripheral blood were evaluated on the basis of the critical role of TfR in iron homeostasis (24, 25). Microscopic analysis was also performed to evaluate for histopathology and/or abnormal iron accumulation in selected tissues known to express high amounts of TfR (heart, kidney, liver, pancreas, skeletal muscle, spleen, and bone marrow) (16).

The anti-TfRD/BACE1 plasma antibody concentration profile was similar over time after either two or four weekly doses, with a slight decrease in antibody concentration observed during week 4 (Fig. 6A). The partially attenuated plasma concentration profile suggests increased antibody clearance after repeated dosing, potentially due to development of mouse anti-drug antibodies to the administered human IgG. Similar to the plasma antibody concentrations, brain antibody concentrations were only decreased after the fourth dose administration (particularly evident on post-dose day 4), although the overall persistence of the antibodies in the brain over time mirrored that seen after the second dose, reaching similar maximal concentrations well above the anti-BACE1 IC50 (Fig. 6B). Plasma (Fig. 6C) and brain (Fig. 6D) concentrations of Aβ correlated well with the observed concentrations of anti-TfRD/BACE1 present in the plasma and brain after two or four doses, respectively.

Fig. 6 Multiple doses of anti-TfRD/BACE1 show sustained PK/PD response and do not exacerbate reticulocyte reduction.

(A to D) Quantification of (A) plasma and (B) brain antibody concentrations and (C) plasma and (D) brain Aβx–40 concentrations in WT mice after the second or fourth weekly dose (25 mg/kg) of either control IgG or anti-TfRD/BACE1 (n = 6 per group). (E) Total reticulocyte count in WT mice 1 day after the second and fourth dose, and 7 days after the fourth weekly dose of control IgG or anti-TfRD/BACE1 (25 mg/kg). Error bars represent means ± SEM. P values were determined using two-tailed unpaired t test, compared to control IgG. *P < 0.05, ***P < 0.0001.

No exacerbation of reticulocyte toxicity was observed after multidose anti-TfRD/BACE1 administration. In Fig. 6E and consistent with results from the single-dose experiments, reticulocyte count was decreased 1 day after the second dose of anti-TfRD/BACE1. However, 1 day after the fourth dose, reticulocyte levels showed a slight improvement toward control levels and were completely recovered to levels modestly above baseline counts by 7 days after the fourth dose. There was no evidence of sustained decreased mature red cell mass or changes in serum iron parameters (fig. S4), or any histopathology findings or altered stainable iron levels in any tissues evaluated after 4 weeks of dosing.

The anti-TfRD/BACE1 brain uptake data (Fig. 6B) and the correlated reduction in brain Aβ levels (Fig. 6D) address a key question about TfR trafficking at the BBB, namely, whether TfR transcytosis can be used repeatedly in a multiple-dose paradigm to increase antibody levels in brain. These data lay the foundation for an anti-TfR bispecific approach to be investigated in nonhuman primates in preparation for clinical validation of these BBB-crossing antibodies.

Comparison of circulating TfR+ cells in blood from mouse, monkey, and human

Observing that TfR+ cells are specifically susceptible in circulation, but not in bone marrow, after administration of effectorless anti-TfR/BACE1 to mice (fig. S3), we explored the number of circulating TfR+ cells in monkey and human blood compared to mouse (Fig. 7). Consistent with previous reports (24), we saw a difference in the number of TfR+ cells in monkey (0.2%) and human (0%) when compared to mouse (1.1%). These data suggest some differences in the site of reticulocyte maturation (Fig. 2A) in mice relative to monkeys and humans (for example, blood versus bone marrow). That is, immature reticulocytes are largely retained in the bone marrow of primates. We therefore speculate that primates will be less susceptible to the residual decreases in circulating reticulocytes seen with treatment of effectorless TfR bispecific antibodies.

Fig. 7 Comparison of circulating TfR+ cells in blood of mouse, monkey, and human.

(A to C) TfR+ reticulocyte populations in blood were labeled and quantified using flow cytometry for mouse (A), monkey (B), and human (C), respectively (n = 5 per species). Cells were identified with an anti-TfR antibody and an erythrocyte lineage marker appropriate for each species: (A) Ter119 expression; (B) gating based on cell size and lack of CD41a (platelet marker) and CD45 (pan-leukocyte marker) expression; and (C) glycophorin A expression. TfR+ populations were then quantitated as a percentage of blood erythroid cells gated on the isotype control (mean ± SD).


The idea to use endogenous receptors that undergo receptor-mediated transcytosis to carry therapeutics across the BBB is not new (3, 2527); however, uncertainties around the therapeutic potential of this approach remain. Clinical proof of concept demonstrating technical success of enhanced antibody uptake in the human brain remains a long-term goal that will require rigorous drug development to understand the therapeutic properties and intrinsic liabilities of BBB-targeted therapies.

Ensuring adequate drug exposure (PK) and associated drug activity (PD) in the brain is an essential step toward this goal. Most antibody-based approaches have relied on tertiary efficacy endpoints, for example, plaque reduction or behavioral endpoints after chronic dosing to show signs of drug activity. This approach poses limitations for engineering the desired properties of a therapeutic antibody, because it is difficult to fine-tune an optimal dosing regimen based on results from chronic dosing studies with limited dose levels. To overcome this obstacle, we focused on developing antibodies to the enzyme BACE1 (14), an aspartyl protease that constitutively contributes to Aβ production in both the periphery and the brain (28). BACE1 is a long sought after target for the treatment of Alzheimer’s disease and is an ideal brain antibody target for understanding BBB penetration, because we have recently shown that drug exposure in both the periphery and the brain tightly correlates with inhibition of BACE1 proteolytic activity as measured by reduction in Aβ (14). By combining an antibody against BACE1 with an antibody directed against TfR in a bispecific antibody format (anti-TfR/BACE1), we were able to demonstrate that improved brain uptake correlates with better Aβ reduction (13), yet numerous questions remain to be addressed in preparation for clinical testing.

Here, we show that optimization of the TfR binding affinity in the bispecific TfR/BACE1 can enhance the duration of Aβ reduction by increasing exposure in both the periphery and the brain. We also show that the TfR pathway is robust and can be used repeatedly to enhance antibody uptake in brain. These insightful PK/PD and multidose data lay the foundation for engineering monkey/human cross-reactive antibodies for clinical testing, and also illustrate the relationship between TfR binding properties and associated activity of the BACE1 targeting arm. The PK/PD relationship showing that Aβ is robustly reduced if drug levels in brain are above the in vivo IC50 for BACE1 inhibition implies that an increase in anti-BACE1 potency could further reduce the amount of drug needed to effectively inhibit BACE1 in vivo. Therefore, a general principle for BBB targeting TfR bispecific therapies can be inferred: Reduced affinity to TfR improves duration of exposure and thus brain uptake, whereas enhanced affinity to the therapeutic targeting arm can reduce dose levels and frequency.

Using TfR as a means of drug transport across the BBB has received significant attention over the past 2 decades (57, 911). Several reports have shown limited brain uptake with TfR antibodies as a result of sequestration in brain endothelium (11, 12). We have observed similar findings with high-affinity bivalent anti–TfR-targeting antibodies but have shown that reducing affinity to TfR can improve uptake and biodistribution as previously addressed (13). Although significant effort has been placed on exploiting TfR for BBB transport, limited information is available on the safety of TfR-targeted therapies. Nevertheless, a recent study reported no adverse findings after chronic dosing of an anti-TfR bivalent antibody/glial cell line–derived neurotrophic factor (GDNF) fusion protein dosed at 2 mg/kg twice a week for 12 weeks in mice (29). A possible explanation of why acute clinical signs were not observed in this study is that the GDNF/Fc fusion may have interfered with Fc effector function. Nevertheless, reticulocyte count may be reduced in the absence of effector function, a possibility that was not investigated.

In contrast, we show that a single dose of a TfR bivalent antibody at doses as low as 1 mg/kg can cause reversible but severe acute clinical signs as well as profound reticulocyte reduction in mice. These observations, which were not noted initially in anti–TfR/BACE1-dosed mice, resulted in a series of experiments aimed at understanding the mechanism of toxicity. We show that antibody effector function is both necessary and sufficient to cause the acute clinical signs observed after a single anti-TfR dose and that effector function is sufficient but not necessary to drive reticulocyte reduction. Through genetic deletion of C3, we conclude that complement mediates a residual reticulocyte reduction observed after high doses of effectorless anti-TfR/BACE1 in mice. We also show that reducing either dose or affinity can further mitigate the loss of circulating reticulocytes. Finally, a multiple-dose study with anti-TfR/BACE1 ultimately showed a regenerative response of reticulocytes and no other adverse effects, including no changes in iron metabolism or in bone marrow histopathology.

Although our data provide important insight into obstacles that must be overcome to ultimately succeed in the clinical setting, several important issues remain, some of which have been previously highlighted (30). Here, we show that anti-TfRD/BACE1 (25 mg/kg) robustly reduces brain Aβ concentrations, although the effects are transient and directly correlate with brain exposure. As previously mentioned, improving the BACE1 potency will reduce the dose needed and/or improve the duration of effects. This will likely need to be accomplished before advancing to the clinic but is dependent on scaling to primates. Furthermore, a dose (25 mg/kg) of effectorless anti-TfRD/BACE1 shows a reduction in circulating reticulocytes. Although reversible, this observation will likely be dose-limiting. There are at least two possibilities to consider that will allow a dose that shows robust efficacy without associated safety findings. First, as we have shown, primates have substantially fewer TfR+ circulating cells in the blood. Thus, if an antibody has reduced Fc receptor binding, and as a result does not alter bone marrow TfR+ cells, it is likely that primates will be less susceptible to the residual decreases seen in mice treated with effectorless anti-TfR/BACE1. Second, we show that C3 mediates the residual decrease in circulating reticulocytes; therefore, further engineering of the TfR bispecific antibodies might be necessary to eliminate the predicted complement-mediated clearance of TfR+ reticulocytes from blood. Ultimately, the anti-TfR platform must be tested in primates to better understand drug exposure and potential safety considerations. However, the TfR antibodies used in this study are mouse-specific; thus, engineering non-Tf/TfR blocking primate antibodies with various affinities is the first step in this process.

The impact of these findings on the next steps in developing therapies that use the TfR pathway for drug transport is evident. The fact that our observations are consistent with known TfR biology also sheds light on the importance of identifying potential liabilities for other receptor-mediated transcytosis pathways. Although these findings raise important concerns, we provide a potential path forward to engineering safer and more effective TfR-based therapeutics to cross the BBB.

Materials and Methods

Antibody and bispecific antibody production

Affinity engineered variants with lower affinity for TfR were generated through single alanine mutations in one of the complementary-determining regions. Mutants were screened for an appropriate apparent affinity using an anti-TfR competition ELISA. TfRA,D and E IgG competed against biotinylated-TfRA for binding to murine TfR with IC50s of 1.1 ± 0.1 nM, 58.5 ± 8.3 nM, and 5.0 ± 1.2 μM, respectively (mean ± SD).

TfR/BACE1 bispecific antibodies were produced from selected affinity engineered variants using a “knobs into holes” approach as described previously (13, 31). The TfR arm was expressed as a half antibody “hole” in Chinese hamster ovary (CHO) cells with the mutation N297G to prevent glycosylation; the BACE1 arm was expressed as a half antibody “knob” in Escherichia coli as previously described (13, 31). The bispecific arms were individually purified and assembled in vitro. Using the anti-TfR competition ELISA, bispecific TfRA, D and E/BACE1 could compete against biotinylated TfRA for binding to murine TfR with IC50s of 18.0 ± 0.5 nM, 588 ± 26 nM, and 105 ± 140 μM, respectively (mean ± SD). Because of the lack of glycosylation, these bispecific variants have a severely reduced capacity to interact with FcγRs. To further eliminate FcγR interaction, we also introduced the mutation D265A.

Bispecific anti-TfRD/BACE1 was also generated with a wild-type Fc having full effector function and normal glycosylation. TfR (hole) and BACE1 (knob) half antibodies were expressed separately in CHO and annealed in vitro as previously described (31).

Antibody affinity measurements

The anti-TfR competition ELISA was performed in MaxiSorp plates (Nunc) coated with purified muTfR-His (2.5 μg/ml) in phosphate-buffered saline (PBS) at 4°C overnight. Plates were washed with PBS/0.05% Tween 20 and blocked with SuperBlock blocking buffer in PBS (Thermo Scientific). A 1:3 serial titrated bivalent IgG (anti-TfRA, anti-TfRD, and anti-TfRE) or bispecific antibody (anti-TfRA/BACE1, anti-TfRD/BACE1, or anti-TfRE/BACE1) was combined with 1 nM biotinylated anti-TfRA and added to the plate for 1 hour at room temperature. Plates were washed with PBS/0.05% Tween 20, and horseradish peroxidase (HRP)–streptavidin (SouthernBiotech) was added to the plate and incubated for 1 hour at room temperature. Plates were washed with PBS/0.05% Tween 20, and biotinylated anti-TfRA bound to the plate was detected with 3,3′,5,5′-tetramethyl benzidine (TMB) substrate (BioFX Laboratories).

Measuring antibody concentrations in mouse brain and plasma

The animals’ care was in accordance with Genentech Institutional Animal Care and Use Committee guidelines. Wild-type female C57B/6 mice ages 6 to 8 weeks were used for all studies unless otherwise indicated. Fcγ−/− mice were purchased from Taconic (B6.129P2-Fcer1gtm1Rav N12), and C3−/− mice were previously described (19). Mice were intravenously injected with either anti-TfR variants, control IgG, or anti-TfR/BACE1 (50 mg/kg). Total injection volume did not exceed 250 μl, and antibodies were diluted in Dulbecco’s PBS (D-PBS) when necessary (Invitrogen). After the indicated time, mice were perfused with D-PBS at a rate of 2 ml/min for 8 min. Brains were extracted, and the cortex and hippocampus were isolated and homogenized in 1% NP-40 (Calbiochem) in PBS containing Complete Mini EDTA-free protease inhibitor cocktail tablets (Roche Diagnostics). Homogenized brain samples were rotated at 4°C for 1 hour before spinning at 14,000 rpm for 20 min. Supernatant was isolated for brain antibody measurement. Whole blood was collected before perfusion in EDTA microtainer tubes (BD Diagnostics), allowed to sit for 30 min at room temperature, and spun down at 5000g for 10 min. The top layer of plasma was transferred to new tubes for antibody and mouse Aβx–40 measurements.

Total antibody concentrations in mouse plasma and brain samples were measured with a generic human Fc ELISA. Nunc 384-well MaxiSorp immunoplates were coated with F(ab′)2 fragment of donkey anti-human IgG and Fc fragment–specific polyclonal antibody (Jackson ImmunoResearch) overnight at 4°C. Plates were blocked with PBS and 0.5% bovine serum albumin (BSA) for 1 hour at 25°C. Each antibody (control IgG, anti-TfR variants, anti-BACE1, and anti-TfR/BACE1 bispecific) was used as a standard to quantify respective antibody concentrations. For mice dosed with a combination of two antibodies, equal mixture of the two antibodies was used as the standard. Plates were washed with PBS and 0.05% Tween 20 using a microplate washer (Bio-Tek Instruments Inc.), and standards and samples diluted in PBS containing 0.5% BSA, 0.35 M NaCl, 0.25% CHAPS, 5 mM EDTA, 0.05% Tween 20, and 15 ppm (parts per million) Proclin were added for 2 hours at 25°C. Bound antibody was detected with HRP-conjugated F(ab′)2 goat anti-human IgG and Fc-specific polyclonal antibody (Jackson ImmunoResearch) and developed with TMB (KPL Inc.), and absorbance (A) was measured at 450 nm on a Multiskan Ascent reader (Thermo Scientific). Concentrations were determined from the standard curve with a four-parameter nonlinear regression program. The assay had lower limit of quantification (LLOQ) values of 3.12 ng/ml in serum and 12.81 ng/g in brain. Statistical analysis of differences between experimental groups was performed with two-tailed unpaired t test.

For the BBB permeability study, brain antibody concentrations were also measured with a specific BACE1 ectodomain ELISA following similar procedures described above. The BACE1 extracellular domain was used as coat, and the HRP-conjugated F(ab′)2 goat anti-human IgG and Fc specific polyclonal antibody were used as the detection. This assay had LLOQ values of about 2.56 ng/g for anti-BACE1 and 12.8 ng/g for anti-TfRD/BACE1.

Measuring mouse Aβx–40 in brain and plasma

Mice were intravenously injected with the indicated antibody (5, 25, or 50 mg/kg) and taken down at indicated time points from 1 to 10 days after the injection. Antibody dosing and perfusions were performed in wild-type mice as stated above. For Aβx–40 measurements, hemi-brains were homogenized in 5 M guanidine hydrochloride buffer, and samples were rotated for 3 hours at room temperature before diluting (1:10) in 0.25% casein and 5 mM EDTA (pH 8.0) in PBS containing freshly added aprotinin (20 mg/ml) and leupeptin (10 mg/ml). Diluted homogenates were spun at 14,000 rpm for 20 min, and supernatants were isolated for Aβx–40 measurement. Plasma was prepared as described above. The concentrations of total mouse Aβx–40 in plasma and brain were determined with a sandwich ELISA following similar procedures described above. Rabbit polyclonal antibody specific for the C terminus of Aβx–40 (Millipore) was coated onto plates, and biotinylated anti-mouse Aβ monoclonal antibody M3.2 (Covance) was used for detection. The assay had LLOQ values of 1.96 pg/ml in plasma and 39.1 pg/g in brain. Statistical analysis of differences between experimental groups was performed with two-tailed unpaired t test.

Hematology and chemistry analysis

Red cell and reticulocyte counts and indices were determined on KEDTA blood with the Sysmex XT-2000iV (Sysmex). The Sysmex detects and classifies reticulocytes and the immature reticulocyte fraction (sum of high and middle/intermediate fluorescent reticulocytes) by flow cytometry with a fluorescent polymethine dye to bind cellular RNA and cell light scatter characteristics. Direct bilirubin, serum iron, and total and unsaturated iron binding capacity were determined on serum by colorimetric assays on the Integra 400 (Roche).

Immunohistochemistry of BBB markers

Wild-type mice were intravenously injected with control IgG, anti-TfRD, or anti-TfRD/BACE1 (50 mg/kg), followed by D-PBS perfusion after dosing. Brains were drop-fixed in 4% paraformaldehyde (PFA) overnight at 4°C, followed by 30% sucrose overnight at 4°C. Brain tissue was sectioned at 30-μm thickness on a sliding microtome, blocked for 1 to 3 hours in 5% BSA/0.3% Triton, incubated with primary antibodies in 1% BSA/0.3% Triton overnight at 4°C, washed three times in PBS/0.1% Triton, incubated with 1:200 secondary antibodies in 1% BSA/0.3% Triton for 2 hours at room temperature, washed twice in PBS/0.1% Triton, and washed once in 1:5000 DAPI in PBS. Mounted slides were subsequently analyzed by Leica fluorescence microscopy. The following antibodies were used for immunohistochemistry: anti-Glut1 (Novus Biologicals, NB300-666), anti–ZO-1 (Invitrogen, 402200), and anti–Claudin-5 (Invitrogen, 35-2500).

Evans blue leakage assay

Wild-type mice were intravenously injected with control IgG, anti-TfRD, or anti-TfRD/BACE1 (50 mg/kg). Twenty-four hours after dose, 100 μl of 0.5% Evans blue (diluted in H2O) was intravenously injected in the animals and allowed to circulate for 30 min before PBS perfusion. Whole blood was collected before perfusion in EDTA microtainer tubes (BD Diagnostics) and spun down at 2000g for 3 min. Brain tissue was extracted and dehydrated at 50°C for 48 hours, followed by addition of 1 ml of N,N′-dimethylformamide to the dried brain tissue for 72 hours for dye extraction at 50°C. The content of Evans blue was measured by spectrometer at A620 and A740. The value of A620A740, normalized to plasma Evans blue content, was used as Evans blue content from brain tissue.

Flow cytometric comparison of TfR-positive blood cells across species

Mouse, cynomolgus monkey, and human whole blood samples were essentially stained with the same methods. Each whole blood sample was diluted 1:100 with PBS. Aliquots (100 μl) of the diluted blood were then blocked with normal human serum (30 μl per aliquot) on ice for 20 min. Afterward, the aliquots were incubated for 30 min at 3° to 8°C in the dark with the appropriate antibodies to stain for RBCs and TfR or Her2 (isotype control). Next, the samples were washed twice with PBS to remove unbound antibodies and reconstituted with 1% PFA. Samples were acquired on a BD FACSCanto II with logarithmic scaling for forward and side scatter. Mouse RBCs were determined from Ter119+ cells; human RBCs were glycophorin A (CD235a+) cells; there were no commercially available antibodies that specifically stained cyno RBCs. Therefore, negative gating was used to select out platelets and leukocytes to designate cyno “RBCs” as CD41a and CD45.

Supplementary Materials

Materials and Methods

Fig. S1. Anti-TfRA/BACE1 and anti-TfRD/BACE1 dose-response PK/PD.

Fig. S2. In vitro ADCC and CDC characterization of TfR antibodies.

Fig. S3. Effector-positive, but not effectorless, anti-TfR/BACE1 reduces mouse erythroid progenitor cells in bone marrow.

Fig. S4. No exacerbation of hematologic toxicity or impact on serum iron parameters after 4 weeks of dosing with anti-TfRD/BACE1.

References and Notes

  1. Acknowledgments: We thank M. Schweiger and M. Kenrick for the initial observations of acute clinical signs; M. McDowell for coordination of clinical pathology and hematology studies; V. Quarmby for guidance on in vitro assays; J. M. Elliott and S. Lee for the antibody generation; and J. Maloney and J. Atwal for designing and cloning numerous constructs and for coordination of numerous study designs and material generation, respectively. Author contributions: R.J.W., M.S.D., J.A.C., and Y.J.Y. designed the project. R.J.W., J.A.C., Y.J.Y., W.J.M., H.S., K.S.-L., J.M.T., Z.L., M.B., and R.N.F. designed, performed, oversaw, and analyzed various in vivo experiments. M.S.D. and Y.Z. designed and constructed the antibodies and performed in vitro binding experiments. R.K.T. and J.A.E. purified and refolded the bispecific antibodies and antigens. K.H., W.L., and Y. Lu ran PK/PD assays. K.G. and S.P. provided modeling and analysis of PK/PD data. B.A.O., Q.N., and Y. Lin designed, conducted, and analyzed various in vitro experiments. R.J.W., J.A.C., M.S.D., and Y.J.Y. wrote the manuscript. Competing interests: All authors are paid employees of Genentech Inc. Genentech has filed patent applications related to this work on blood-brain barrier transport using transferrin receptor antibodies.
View Abstract

Navigate This Article