The company that received federal approval last November for the first field test of a live genetically engineered microbe had already tested it on trees in the open air as early as last February, company officials acknowledged yesterday.

Advanced Genetics Sciences Inc. of Oakland, Calif., said it believed that the experiment was safe and within government guidelines. "We conducted a series of pathogenicity studies on trees contained in an asphalt area" on the company's roof, said John Bedbrook, vice president and director of research. " . . . We followed all the guidelines to the letter."

An Environmental Protection Agency official, who asked not to be identified, called the report of the outdoor experiment "a real shock." A spokesman said the agency would investigate.

Advanced Genetics, which had sought permission to use the microbe on strawberries, tested it on more than 45 trees -- apple, almond, peach, cherry and pear -- to see if any would be injured. The roof is ringed by a four-foot wall but has no barriers to prevent the microbes from being picked up by insects and birds.

The results of the open-air tests were unclear, a source familiar with the experiment said.

Bedbrook said a syringe was used to inject the genetically altered microbes so "the organism was physically contained within the tree itself." He said the syringe also is "containment."

But, according to the source, the trees spilled sap where the syringe was inserted, because the injection point remained open for some time. Such leakage, the source said, might attract insects that could pick up the bacteria and spread it for miles.

Advanced Genetics, after a long public debate last year, obtained EPA approval to test the laboratory-altered microbe. It hoped that applying it to strawberry plants would protect them from frost.

But the experiment, approved also by the state of California, has been held up by local officials in the Salinas region, where the test was to be carried out on a one-fifth acre strawberry patch. The location, officials said, is too near other crops and populated areas.

In seeking EPA approval, the company tested the microbe's viability, its ability to spread to other plants, insects and animals, and its toxicity to other plants.

Those tests, the company said, were carried out in its greenhouse in Oakland. Bedbrook said the greenhouse test followed federal guidelines.

But, the source said, the greenhouse was not sealed against outside air. A ventilation system pumped in a stream of air and let it out through windows in the roof, the source said.

Since the bacteria probably floated in the air near the plants they were sprayed on, it is likely that they were allowed to escape when windows were open, the source said.

Milton Schroth, a professor of plant pathology at the University of California and a former AGS employe, said he doesn't think present federal regulations governing such tests make sense. He said the "whole notion of containment in a greenhouse is foolish."

He said that in addition to the problem posed by windows, greenhouse workers could brush against plants and carry bacteria outside on their clothing. "And what about the water used to water the plants?" he wondered. "Where does that go?"

Despite his concerns, Schroth and other scientists said they think the microbe being tested is safe, since it closely resembles one that is common in nature.

Others, led by social activist Jeremy Rifkin and his Foundation on Economic Trends, have said that not enough testing has been done to assure that the microbe won't have adverse affects on other plants, such as triggering disease, or on the weather, since it prevents normal formation of rain and snow in the atmosphere.

Rifkin has sued the EPA for approving the experiment without more complete testing. A court hearing is due this week.