1Biology Department, IENS, Lancaster University, LA1 4YQ, UK; 2Department of Aghcultural Botany, University of Reading, Reading, RG6 6AU, UK

Background and objectives
Field observations show that concurrent infection by more than one fungal pathogen is rather common. For example, in natural populations of the perennial weed Rumex obtusifolius (broad-leaved dock) leaves are often observed with concurrent infections of rust (Uromyces rumicis), the hemibiotroph Venturia rumicis and the necrotoph Ramularia rubella. There appear to be few investigations of such multiple infections. We have analysed the frequency distribution of foliar pathogens on R. ;obtusifolius in field experiments to consider whether infection by different fungi was mutually independent.

Materials and methods
R. ;obtusifolius plants, 4 weeks old, were grown in the field either in spring (late May to mid-July) or autumn (mid-September to early November). Plants were grown at three levels of added N (1, 1 0, or 25 ;mmol/l N added every 4 ;days as a solution of NaNO3). No artificial inoculation was used. Plants became naturally infected with R. ;rubella and V. ;rumicis in spring, and with all three pathogens in autumn. 'Source' plants infected with R. ;rubella (spring) and U. ;rumicis (autumn) were interspersed with experimental plants to provide uniform inoculum. In the spring experiment, the number of lesions per leaf of R. ;rubella and V.nbsp;rumicis were counted, and the number of leaves with no infection, or infected by one or both species were tallied for each N treatment. The probability of any leaf being infected by R. ;rumicis (P(R)) and that for infection by V. ;rumicis (P(V)) were then calculated. If it is assumed that the probability of infection by the two pathogens is mutually independent, the expected number of leaves with none, one or two fungi can be calculated:
Expected frequency of leaves with V. ;rumicis alone=P(V).[1-P(R)]
Expected frequency of leaves with R. ;rubella alone=P(R).[1-P(V)]
Expected frequency of leaves with both fungi=P(R).P(V)
Expected frequency of leaves with no fungi=[1-P(R)].[1-P(V)].
In the autumn experiment the same calculations were made but including a new term P(U), the probability of U. ;rumicis infection. Observed and expected values were compared using chi2.

Results and conclusions
In spring, approximately 25% of leaves of plants grown at low N were not infected by either pathogen, 5% were infected by R. ;rumicis alone, 40% by V. ;rumicis alone and 27% by both fungi. In this case, the frequency of dual infections was predictable from the frequency of single infection. Increasing N had little effect on the percentage of uninfected leaves or those infected by R. ;rubellaalone, but single infection by V. ;rumicis became less frequent and dual infection more frequent (20% and 49%, respectively, at 25 ;mmol/l-1 N). At high N, the frequency of dual infection was significantly greater than that calculated from the probability of single infections (39%).

In autumn, approximately 17% of leaves in plants grown at low N were not infected by any of the three pathogens, 5% were infected by only one pathogen, 38% by two and 42% by all three. The frequency distribution of infections was significantly different from that calculated from the frequencies of single infections: triple infection was more frequent and single-species infection less frequent than expected. V. ;rumicis lesion density was significantly higher in leaves coinfected with U. rumicis than in leaves with V. ;rumicis alone.

In both experiments infection by fungal pathogens were clumped, with multiple infections being more common than expected. The mechanisms underlying these nonrandom distributions remain far from clear. Whatever the underlying mechanism(s), such clumped pathogen distribution could have implications for both competition between pathogens and the physiological responses of the host to infection, for example in the scope of compensatory responses in uninfected leaves.