Of the traditional medicine of China, the most famous proverb is that, "If it's a drug, the toxicity would hold the room with the thirty percent possibility." [1] When a herbalist writes a prescription, he still finds a solution in order to counteract the toxicity more or less. This solution means that the patient must drink at least two kinds of drugs, [2] which one or more of them will decrease the discomfort from the other staples. Now we know that the discomfort even the lethal mainly initiate from the drug-induced liver toxicity.

The tissue model dominates the reliability of drug evaluation in modern biomedicine. By studying cell responses on a lab table, the developers predict possible drug metabolism in human tissue. Obviously, there are some major deficiencies with these indirect methods. One comes from preclinical animal test. This popular standard is inadequate to predict drug responses of human because of species-specific variation. The other one is due to very limited samples of human tissues. Except we can tackle moral and ethical problems on cloning, [3] the models of human tissue are forced to consider a scale of a few cells. Just how reliable is this evaluation? Before we have an answer, central challenges are still around scale-up, and it's not surprising that a TV news calls for "urgent attention" to a drug withdrawal. However, for the patients, it is just not a withdrawal.

The liver has a key role in drug metabolism, therefore, the liver toxicity as well as the models of human liver will still be intense focus. Recently, there is a more accurate liver model with improved testing for a wide range of chemical compounds, reported Sangeeta N Bhatia and Salman R Khetani in the January issue of Nature Biotechnology in 2008. [4] Their strategy is based on micropatterned cocultures of human liver cells and mouse fibroblasts (Hepatocyte-fibroblast cocultures). The cocultures within the microscale holes of a honeycomb-like host, which made by a soft lithography process, effectively developed liver subunits that maintained phenotypic functions for 4-6 weeks. The system is compatible with robotic fluid handling. More importantly, because "tissue function depends on hierarchial structures extending from single cells (~ 10 micrometers) to functional subunits (100 micrometers - 1 millimeter) that coordinate," their method can be used to study cell responses within higher-order processes, that is, the model performs better.

Put simply, thought the authors, "This approach should be useful for high-throughput drug screening aimed at reducing costs, increasing the likelihood of clinical success and limiting human exposure to unsafe drugs." It's really glad for me to read this argument. I agree, because our livers have been overloaded much. ♦

* Lin PU is in the Physics Department of Nanjing University, Nanjing 210093, CHINA.
♦ Received 20080124. TXT online 20080125.