We have isolated a cDNA from a mouse brain library that encodes a protein whose predicted amino acid sequence is 42% identical and 64% similar to that of the amylold (3 protein precursor (APP). This 653-amino acid protein, which we have termed the amyloid precursor-like protein (APLP), appears to be similar to APP in overall structure as well as amino acid sequence. The amino acid homologies are concentrated within three distinct regions ofthe two proteins where the identities are 47%, 54%, and 56%. The APLP cDNA hybridizes to two messages of approximately 2.4 and 1.6 kilobases that are present in mouse brain and neuroblastoma cells. Polyclonal antibodies raised against a peptide derived from the C terminus of APLP stain the cytoplasm in a pattern reminiscent of Golgi sai . In addition to APP, APLP also displays significant homology to the Drosophila APP-like protein APPL and a rat testes APP-like protein. (11-14). Therefore, the normal constitutive processing ofAPP precludes the formation of fA4.In the course of cloning the human APP gene, other cross-hybridizing cDNAs were also obtained, leading to the suggestion that APP is a member of a larger gene family (4). We have now identified a mouse brain cDNA that encodes a protein whose predicted amino acid sequence is 42% identical and 64% similar to APP.L This amyloid precursor-like protein (APLP) is clearly not the mouse homologue of human APP, which has previously been cloned and found to be 96.8% identical to human APP (15). APLP contains three particularly conserved regions of homology with APP and is encoded by a 2.4-kilobase (kb) message that is present in mouse brain and neuroblastoma cells. Antibodies raised against a synthetic peptide derived from the C terminus of APLP stain the cytoplasm of neuroblastoma cells in a pattern similar to that obtained with an antibody to a known Golgi protein.The cDNA described here, together with the gene encoding APP and two other genes-one from Drosophila (16) and one from rat (17)-form a gene family producing proteins that are highly conserved and may therefore share common functions and, perhaps, undergo similar processing.MATERIALS AND METHODS Cells. Neuroblastoma NB2A cells were maintained as previously described (18 (23,24). Production ofAntisera to an APLP Peptide. A 20-mg sample of the peptide CQQLRELQRH

Tissue-specific gene delivery is an important aspect of many gene therapy applications. The experiments reported here constitute the first successful demonstration that cell-specific entry can be obtained by screening a random library of retroviral envelope proteins produced from a mammalian cell system. The library consisted of 10(6) different subgroup A feline leukemia virus envelope protein variants with 10 randomly substituted amino acids in the receptor-determining region. Selecting the library for fully functional envelope proteins able to mediate stable gene transfer resulted in the identification of a single envelope protein variant (EF). Subsequent examination of the host range of EF revealed that it was highly specific for D17 canine osteosarcoma cells. This was in contrast to the host ranges of the parental subgroup A and closely related subgroup C envelope proteins. Interference assays on D17 cells further indicated that receptor usage by EF was also altered compared with the A and C envelope proteins. The EF envelope protein thus isolated should be useful for studying gene therapy treatments of osteosarcoma in a large-animal model.

An important requirement for many gene therapy applications is to direct therapeutic genes specifically to target cells. Here we describe an improved vector targeting method that does not depend on the use of a known cell-targeting ligand. It entails screening a library of constitutively produced retroviruses with random amino acid substitutions in the cell-targeting region of the envelope proteins for their ability to mediate gene delivery to a target cell. By screening such a library on the ras-transformed 143B human cell line, we have isolated an envelope protein that preferentially targets 143B cells and 293T cells expressing the SV40 T antigen via a novel, unidentified receptor. Furthermore, retroviruses expressing the library-derived envelope protein can be concentrated by centrifugation. This is the first demonstration of a novel concept in vector targeting: the selection of productive retroviral entry via an alternate receptor with modified cellular tropism in the absence of a known cell-targeting moiety. The method is, in principle, applicable even to cells that have not been well characterized, and therefore potentially suitable for targeting many diverse cell types.

No abstract

(11)(12)(13)] viral receptors. With these evolutionary observations, it is of great interest as to whether a given retrovirus must remain restricted to use receptors within the broad class of the original receptor or can be altered to use receptors beyond these protein family boundaries.A variety of approaches have been developed for altering the receptor usage of viral delivery systems. Receptor specificity and binding of the viral Envelope (Env) is mapped to the N-terminal half of the surface (SU) protein through the close interaction of two highly variable regions, VR1 (VRA) and VR2 (VRB) (14-18). Exchange of a defined VR1 segment between the FeLV A and C subgroups altered the viral host range (19). In our studies, codons for 10/11 amino acids within this VR1 FeLV Env receptor-binding region were randomized within a FeLV A/C Env chimera backbone (20). Viral particles bearing this library of Ͼ1 million viral VR1 Env variants are selected for productive infection (20)(21)(22). This library displays the targeting peptide within the context of the authentic Env backbone, eliminating the problem that preidentified phage display peptides inserted into alternative contexts do not maintain their original properties (23). Alternative approaches have involved the addition of targeting domains (24, 25), peptides (23, 26), or single-chain antibodies (27) within the framework of viral Env or capsids (28-30) or through heterologous intermediates (31). Successful retargeting for Sindbis Env pseudotyped onto lentiviral particles has been achieved with the insertion of the Protein A ZZ domain and sandwiching targeting antibodies to the modified particles (32) or through coexpression of Env proteins and surfacebinding molecules (33).Screening the FeLV random Env library on feline AH927 cells identified the A5 Env isolate, which also efficiently infected human 293T embryonic kidney cells (21). Superinfection interference studies indicated that the receptor for A5 was outside the receptor family used by FeLV A, B, and C and MuLV 4070A viruses (21). The A5 Env therefore provided an excellent system to examine the range of proteins capable of serving as receptors for retroviral Env libraries. In this study, the receptor for the A5 Env isolate was identified as the SLC35F2 protein, a transporter protein of unknown function, through screening a feline AH927 cDNA library. ResultsDirected evolution allows for the selection of proteins with new properties. Using this approach, randomization of the FeLV Env receptor-binding domain, followed by selection for productive infection, identified Env isolates with receptor usage outside the interference groups of the known FeLV A, B, and C viruses (21,22). This result implies that the selection of a receptor protein can occur through screening a single round of viral infection. To definitively prove that alternative viral receptors can be selected, the receptor protein for the A5 Env isolate was identified through cDNA screening. The authors declare no conflict of interest.Abbreviations: Env,...