Erythropoietin gene therapy leads to autoimmune anemia in macaques

Clinical administration of therapeutic proteins, such as insulin for diabetes, requires repeated parenteral administrations. Problems of protein delivery could be overcome through the use of gene transfer with vectors capable of stably expressing the therapeutic protein. The most advanced application of this concept is the treatment of hemophilia B with adeno-associated virus (AAV) vectors expressing factor IX. Intramuscular administration of this kind of vector in patients with hemophilia B looks promising, though expression levels were low.¹ 

We evaluated the systemic delivery of the hormone erythropoietin (EPO) in cynomolgus macaques following intramuscular injection of AAV vectors. Previous studies in rodents injected with AAV vectors expressing EPO demonstrated stable expression of the hormone and persistent elevations in hematocrit levels.² 

The EPO cDNA derived from a rhesus macaque³  was cloned into a plasmid vector that contained a promoter from cytomegalovirus flanked by inverted tandem repeats (ITRs) from AAV serotype 2. AAV2CMVrhEPO vectors were pseudopackaged with capsids from AAV serotypes 1, 5, 7, and 8 and were administered to the vastus lateralis muscles of 2 cynomolgus macaques (AAV2/1: 17106 and 17108; AAV2/5: 17111 and 17112; AAV2/7: 17146 and 17147; AAV2/8: 17148 and 17152) at a dose of 1 × 10¹³ genome copies per kilogram of animal weight. The animals were treated and cared for at a contract, nonhuman primate facility (LABS of Virginia, Yemassee, SC) during the study, following a study protocol approved by the Environment Health and Safety Office, the Institutional Biosafety Committee, and the Institutional Animal Care and Use Committees (IACUCs) of the University of Pennsylvania and LABS of Virginia. Blood samples were taken by venipuncture of the saphenous vein. Therapeutic phlebotomy was performed for any animal with a hematocrit greater than 65%. When the hematocrit dropped to less than 10%, animals were humanely killed with atropine and ketamine. A full macroscopic evaluation was performed at necropsy, followed by histopathologic analysis of tissues.

The presence of anti-recombinant human EPO (anti-rhEPO) antibodies in the animal sera was evaluated using 2 different assays. One was designed to measure the binding of the antibody to EPO in serial dilutions of the animal sera. Such binding resulted in interference with the detection of known concentrations of EPO in a standard enzyme-linked immunosorbent assay (ELISA). The other was designed to examine the ability of the macaque serum to neutralize and abolish EPO function in supporting the growth of an EPO-dependent human cell line, HCD57 (+). Briefly, HCD57 (+) cells were cultured for 4 days in growth medium containing EPO, harvested, and washed twice with phosphate-buffered saline (PBS). The cells were then seeded in 96-well plates (1 × 10⁵ cells/well) in the growth medium containing 50 mIU EPO/mL, which is the minimum quantity of EPO that can support the proliferation of HCD57 (+) cells. Equal volumes of macaque serum were serially diluted (1:2, 1:10, 1:100, and 1:1000) in the growth medium, added to each well, and cultured for 48 hours. Cell Titer 96 Aq One Solution (Promega, Madison, WI) was added to each well and incubated for 90 minutes. Plates were read at A490 nm and A650 nm as reference wavelengths.

The EPO cDNA was subjected to sequence analysis, as were the EPO genes from genomic DNA isolated from 8 cynomolgus macaques (including the 4 that developed anemia in this study 17108,17111,17112, and 17148) and 3 rhesus macaques that had developed anemia after AAVEPO gene transfer to the lung. EPO coding sequences in genomic DNA were characterized by polymerase chain reaction (PCR) amplifying the exons, followed by direct sequence analysis of the amplified fragment or cloned fragment. Predicted amino acid sequences of the vector, 8 cynomolgus EPO genes, and 3 rhesus EPO genes were identical. This predicted amino acid sequence (called Penn) differed from the published rhesus EPO sequence (called pubR)⁴  at positions 17 (Penn-Leu and pubR-Val), 33 (Penn-Ile and pubR-Val) and 81 (Penn-Met and pubR-Ile) and from the published cynomolgus sequences (called pubC)⁵  at position 17 (Penn-Leu and pubC-Val). In each case, sequence was confirmed by the full characterization of each strand.

AAV vectors were constructed containing a macaque-derived EPO cDNA expressed from a cytomegalovirus (CMV) promoter. Pseudotypes of this vector were created by packaging the constructs with capsids from AAV serotypes known to confer high-level transduction in skeletal muscle, including AAV1, AAV5, AAV7, and AAV8.⁶  Each vector was injected into the vastus lateralis muscle of 2 cynomolgus macaques. Shortly after gene transfer, supraphysiologic serum levels of EPO, derived from the transgene, were detected, resulting in polycythemia (Figure 1). In most cases, the animals required repeated therapeutic phlebotomies to maintain nontoxic hematocrits (less than 65%). EPO levels in both animals injected with the AAV2/5 vector dropped sharply to undetectable levels within 3 weeks of gene transfer, and severe anemia ensued with hematocrits of 10% at the time of necropsy. Similar outcomes were observed in single animals injected with the AAV2/1 and AAV2/8 vectors, though the decreases in serum EPO and hematocrit were delayed. This phenomenon is not unique to muscle; similar findings were noted in 3 of 3 rhesus macaques that received AAV vectors expressing EPO into the lung (data not shown).

A number of potential mechanisms were considered in evaluating the cause of anemia in these animals. No laboratory or clinical evidence showed hemolysis or hemorrhage (data not shown). The total absence of reticulocytes in peripheral blood in animals with severe anemia, but no leukopenia or thrombocytopenia, suggested a selective defect in erythropoiesis. We speculated that the anemia might have been caused by an autoimmune response to EPO protein, leading to the production of antibodies that interfered with EPO function. This was tested in 2 ways. First, sera were mixed with purified human EPO and were analyzed with a commercially available ELISA. The second test measured the ability of animal sera to reduce the biological activity of EPO. Sera in serial dilutions were added to the culture of an EPO-dependent cell line in the presence of a known quantity of EPO, followed by an evaluation of the degree of growth inhibition after the addition of the sera. In each case, the apparent decrease in serum EPO, as seen in the animals in vivo, correlated with the appearance of a factor in the serum that interfered with EPO detection by ELISA and EPO activity in the in vitro proliferation assay (Figure 2).

Sequences of the transgene and endogenous EPO open-reading frames of all affected animals were analyzed and were identical to those of normal animals (see “Study design”). This indicates that the genetic intervention broke tolerance to a self-antigen, resulting in autoimmune anemia.

This is the first example of the inadvertent development of an autoimmune disease in nonhuman primates as a result of somatic gene transfer of a gene expressing a self-antigen. EPO protein replacement has been extraordinarily successful in treating several forms of anemia. There have been rare case reports of pure red cell aplasia following EPO protein therapy caused by antibodies against EPO,⁷  especially in patients with chronic renal failure.⁸  The frequency with which we observed EPO-induced anemia in nonhuman primates after AAV gene transfer raised concern. Mechanisms that predispose to this toxicity are unclear. This phenomenon is not restricted to a specific vector serotype because it was observed in 3 of 4 serotypes tested, nor is it unique to intramuscular administration given that it was observed in rhesus macaques administered EPO expressing AAV into the lung as well. Furthermore, this occurrence is not species-specific because it was observed in rhesus and cynomolgus macaques. The anemia was severe enough to require euthanasia, though the level of interfering antibody began to abate at the time of necropsy in 2 of the animals. In the companion paper,⁹  Moullier's group describes similar findings in cynomolgus macaques injected with AAV expressing EPO from a tetracycline (TET)-inducible promoter.

The development of a clinically significant autoimmune response to EPO in macaques under a variety of experimental conditions should be considered in assessing the safety of clinical gene therapy. This type of complication has not characterized AAV gene transfer using other transgenes, which may be caused by unique problems with EPO. Alternatively, this may reflect the fact that autoimmune responses to other transgenes might not have been thoroughly assessed and, if present, might not have been as easily detected. Even with EPO gene transfer, this complication is sporadic, as evidenced by the fact that in other studies we have achieved long-term expression of EPO through an AAV vector expressed from a regulated promoter in skeletal muscle of 21 rhesus macaques (V.M. Rivera, G. Gao, R.L. Grant, et al., manuscript submitted, November 2003). A better understanding of the factors that predispose to transgene-driven autoimmunity would help to assess the potential risks of this complication.