Abasic (AP) sites are common, potentially mutagenic DNA damages that are attacked by AP endonucleases. The biological roles ofthese enzymes in metazoans have not been tested. We have cloned the human cDNA (APE) that encodes the main nuclear AP endonuclease. The predicted Ape protein, which contains likely nuclear transport signals, is a member of a family of DNA repair enzymes that ludes two bacterial AP endonucleases (ExoA protein of Streptococcus pneumonias and exonuclease HI of Escherichia coa) and Rrpl protein ofDrosophila melanogaster. Purified Ape protein lacks the 3'-exonuclease activity against undamaged DNA that is found in the bacterial and Drosophila enzymes, but the lack of obvious amino acid chanes to account for this difference suggests that the various enzyme functions evolved by fine tuning a conserved active site. Expression of the active human enzyme in AP endonuclease-deficient E. cofl conferred significant resistance to kiling by the DNA-alkylating agent methyl methanesulfonate. The APE cDNA provides a molecular tool for analyzing the role of this central enzyme in maintaining genetic stability in humans.A prominent insult to cellular DNA is the continuous loss of bases, either through spontaneous reactions such as hydrolytic depurination (1) and free-radical attack (2) or by the action of DNA glycosylases that remove various altered bases (3). The resulting abasic (AP) sites can block the progress of the DNA replication apparatus and cause mutations in Escherichia coli (4). Evidently, AP sites must be corrected to restore genetic integrity. The major enzymes initiating this repair process, AP endonucleases, have been identified, and their in vivo roles in correcting alkylationinduced AP sites and other DNA damages have been confirmed for bacteria and yeast (4-7). Such molecular information has been lacking for metazoans and particularly for human cells.The major human AP endonucleases have been purified from various sources (8, 9). These enzymes are modest in size (Mr 37,000), act efficiently in the absence of other proteins, and lack activity against undamaged DNA (8, 9). Like the microbial AP endonucleases, the human enzymes are multifunctional activities that not only attack AP sites but also remove fragments of deoxyribose from the 3' termini ofDNA strand breaks produced by free-radical attack. For the main AP endonuclease of HeLa cells, this 3'-repair activity is relatively weak ["1% of the AP-cleaving activity (10)], in contrast to exonuclease III and endonuclease IV of E. coli and Apnl of Saccharomyces cerevisiae, in which the 3'-repair and AP-cleaving activities are about equal (11,12). However, all of these enzymes cleave DNA AP sites in the same way, as hydrolytic (so-called class II) AP endonucleases that incise the phosphodiester just 5' to the site to generate a normal 3'-terminal 3'-hydroxyl nucleotide and a 5'-terminal deoxyribose-5-phosphate (13). A second class II AP endonuclease is also present in HeLa cells in lesser amounts, and this enzyme has about equal levels ...

Using visual behavioral screens in Drosophila, we identified multiple alleles of N-cadherin. Removal of N-cadherin selectively from photoreceptor neurons (R cells) causes deficits in specific visual behaviors that correlate with disruptions in R cell connectivity. These defects include disruptions in the pattern of neuronal connections made by all three classes of R cells (R1-R6, R7, and R8). N-cadherin is expressed in both R cell axons and their targets. By inducing mitotic recombination in a subclass of eye progenitors, we generated mutant R7 axons surrounded by largely wild-type R cell axons and a wild-type target. R7 axons lacking N-cadherin mistarget to the R8 recipient layer. We consider the implications of these findings in the context of the proposed role for cadherins in target specificity.

Mammalian cells were investigated for enzymes that help correct oxidative damages in DNA. We focused on 3'-repair diesterases, which process DNA ends at oxidative strand breaks by removing 3'-blocking fragments of deoxyribose that prevent DNA repair synthesis. Two enzymes were found in a variety of mouse, bovine and human tissues and cultured cells. The two activities were purified to differing degrees from HeLa cells. One enzyme had the properties of the known HeLa AP endonuclease (Mr approximately 38,000, with identical substrate specificity and reaction requirements, and cross-reactivity with anti-HeLa AP endonuclease antiserum) and is presumed identical to that protein. The second activity did not interact with anti-HeLa AP endonuclease antibodies and had relatively less AP endonuclease activity. This second enzyme may have been detected in other studies but never characterized. In addition to the 3'-repair diesterase and AP endonuclease, this partially purified preparation also harbored DNA 3'-phosphatase and 3'-deoxyribose diesterase activities. It is unknown whether all activities detected in the second preparation are due to a single protein, although activity against undamaged DNA was not detected. The in vivo roles of these two widely distributed 3'-repair diesterase/AP endonucleases have not been determined, but with the characterizations presented here such questions may now be focused.

Different classes of photoreceptor neurons (R cells) in the Drosophila compound eye connect to specific targets in the optic lobe. Using a behavioral screen, we identified LAR, a receptor tyrosine phosphatase, as being required for R cell target specificity. In LAR mutant mosaic eyes, R1-R6 cells target to the lamina correctly, but fail to choose the correct pattern of target neurons. Although mutant R7 axons initially project to the correct layer of the medulla, they retract into inappropriate layers. Using single cell mosaics, we demonstrate that LAR controls targeting of R1-R6 and R7 in a cell-autonomous fashion. The phenotypes of LAR mutant R cells are strikingly similar to those seen in N-cadherin mutants.