Host adaptation to potato and tomato within the US−1 clonal lineage of Phytophthora infestans in Uganda and Kenya


*To whom correspondence should be addressed (e-mail:


Twenty isolates of Phytophthora infestans from potato and twenty-two from tomato, collected in Uganda and Kenya in 1995, were compared for dilocus allozyme genotype, mitochondrial DNA (mtDNA) haplotype, mating type and restriction fragment length polymorphism (RFLP) fingerprint using probe RG57. Based on RFLP fingerprint and mtDNA haplotype, all isolates were classified in the US−1 clonal lineage. Nonetheless, isolates from potato differed from isolates from tomato in several characteristics. Isolates from potato had the 86/100 glucose-6-phosphate isomerase (Gpi) genotype, while those from tomato were 100/100, which represents a variant of US−1 that had been identified previously as US−1.7. Furthermore, while pure cultures of the pathogen were acquired from infected potato leaflets by first growing the isolates on potato tuber slices, this approach failed with infected tomato tissue because the isolates grew poorly on this medium. Tomato isolates were eventually purified using a selective medium. Six isolates from each host were compared for the diameter of lesions they produced on three tomato and three potato cultivars in one or two detached-leaf assays (four isolates from the first test were repeated in the second). On potato leaflets, isolates from potato caused larger lesions than isolates from tomato. On tomato leaflets, isolates from that host caused larger lesions than did isolates from potato, but the difference was significant in only one test. The interaction between source of inoculum (potato or tomato) and inoculated host (potato or tomato) was significant in both tests. Isolates from tomato were highly biotrophic on tomato leaflets, producing little or no necrosis during the seven days following infection, even though abundant sporulation could be seen. In contrast, isolates from potato sporulated less abundantly on tomato leaflets and produced darkly pigmented lesions that were most visible on the adaxial side of the leaflets. Nonetheless, all isolates infected and sporulated on both hosts, indicating that host adaptation is not determined by an ability to cause disease but rather by quantitative differences in pathogenic fitness. Assessment of Gpi banding patterns, mtDNA haplotype and RFLP fingerprint of 39 isolates from potato collected in Uganda and Kenya in 1997 indicated that the population had not changed on this host. The population of P. infestans from Kenya and Uganda provides an interesting model for the study of quantitative host adaptation.


Host adaptation of Phytophthora infestans has important epidemiological consequences in areas where two or more potential hosts grow in close proximity. If multiple hosts are cultivated and inoculum can pass readily among them, then disease management activities must take all hosts into consideration. In addition to this practical aspect, adaptation of P. infestans to potato (Solanum tuberosum) and tomato (Lycopersicon esculentum) is an interesting phenomenon because it appears to be caused by quantitative differences in epidemic components, rather than a fundamental ability to cause disease on either host ( Oyarzun et al., 1998 ). Elucidation of the mechanisms that govern this type of host adaptation could provide new insights into the nature of host–pathogen relationships.

One aspect of host adaptation of P. infestans that remains unclear is whether a single genotype of the pathogen has the potential for a high level of adaptation to both potato and tomato. If genotypes could become highly adapted to both hosts, it seems logical that these would have a competitive advantage over and eventually replace those genotypes that are adapted only to one host ( Turkensteen, 1973). However, dominance by genotypes adapted to both potato and tomato appears to be rare, even in areas where both crops are grown year-round and in close proximity to each other. Genetic markers indicate that distinct genotypes are generally associated with each host in Brazil ( Brommonschenkel, 1988), north-western Mexico ( Goodwin et al., 1992b ), the Philippines ( Koh et al., 1994 ), one region of the Netherlands ( Fry et al., 1991 ); France ( Lebreton & Andrivon, 1998), Ecuador ( Oyarzun et al., 1998 ), and the USA ( Legard et al., 1995 ).

However, although specific genotypes or lineages of P. infestans are generally associated with only one host, there are important exceptions. In a recent study in North Carolina, USA, the clonal lineage US−7 was found primarily on tomato, but US−8 was found attacking both potato and tomato ( Fraser et al., 1999 ). US−7 was found attacking both potato and tomato in the USA in 1993 ( Goodwin et al., 1998 ) and in north-western Mexico in 1991, 1992 and 1993 ( Goodwin et al., 1995b ). US−1 has been isolated from both hosts in the USA ( Legard et al., 1995 ). In the Netherlands, isolates from tomatoes in community gardens near Leiden could not be distinguished with genetic markers from isolates from commercial potato fields in the same area ( Fry et al., 1991 ).

One constraint with many pathogen population studies is that isolates with different multilocus genotypes are known to be different, but isolates with identical multilocus genotypes are not necessarily the same. In terms of tomato/potato adaptation in P. infestans, it is not known with certainty if individuals of the same multilocus genotype found on each host represent a single population with dual adaptation or two different populations.

In an attempt to measure host adaptation directly, some researchers have utilized tests of pathogenic aggressiveness, but even these have not given consistent results. Oyarzun et al. (1998) found different clonal lineages associated with potato and tomato in Ecuador. Each lineage was most aggressive on its primary host, but could cause lesions on the alternative host. In contrast, Legard et al. (1995) found that all isolates examined (regardless of origin) were aggressively pathogenic on potato, while only some were aggressively pathogenic on tomato. In a recent study done in France ( Lebreton et al., 1999 ), isolates from tomato were more aggressive on that host than were isolates from potato. On potato, however, isolates from the two species could not easily be distinguished based on lesion expansion, although potato isolates were more aggressive based on sporulation capacity and competitive fitness in the field.

A study of specific virulence of isolates of P. infestans from Uganda and Kenya indicated that there might be adaptation to potato and tomato in this pathogen population ( Kori, 1972). The objective of the research reported here was to test the hypothesis that populations of P. infestans attacking potato in this region differ from those that attack tomato. The approach was to determine the multilocus genotype of individuals collected on each host, and to compare their pathogenic capabilities on their respective and alternative hosts. Results are discussed with reference to the hypothesis that adaptation to tomato is associated with reduced pathogenic fitness on potato.

Materials and methods

Collection of isolates

One collection, in March 1995, yielded fourteen isolates from tomato and thirteen from potato in Kenya, and eight from tomato and seven from potato in Uganda ( Table 1). A second collection, in late 1997, yielded twenty isolates from Kenya and nineteen from Uganda, all from potato ( Table 1). Phytophthora infestans was isolated from potato leaflets by first growing the isolates on potato tuber slices and then transferring to agar plates ( Forbes et al., 1997 ). Phytophthora infestans from tomato grew poorly on tuber slices and was therefore isolated by placing 0·5-cm2 pieces of infected tissue on a selective medium ( Oyarzun et al., 1998 ). Isolates were maintained in culture on rye B agar ( Caten & Jinks, 1968) at 15 °C in the dark and transferred to fresh medium every six months.

Table 1.  Number and location of isolates of Phythophthora infestans collected from tomato and potato in Kenya and Uganda in 1995 and from potato in 1997
Nearest town or cityLocationNumbers of isolatesAltitudea
  • a

    Metres above sea level. Data were not taken in 1997.

  • b

    With Democratic Republic of Congo.

Kenya (1995)
 NakuruNear Maunarok 42800–2900
 NakuruEgerton College, Njoro4 2500
 NakuruNear Molo 22700
 NairobiNear city5 1900
 NairobiNaivasha, Njambini 22600
Uganda (1995)
 KisoroNear borderb332000
 KabaleNear city 32000
 KabaleKasore1 2400
 KabaleMelango3 1500
Kenya (1997)
 NairobiNjambini: Harris (2),Taraka (1), Munyaka (2),Maguga (1), Tigoni (1) 7 
 Meru/LaikipiaBurguret (2), Kahima (8), Kibiro (1), Kathuuya (1), Kiirigire (1) 13 
Uganda (1997)
 KisoroMaziba (3), Kirembe (3), Kyanankye (1) 7 
 Kabale Kalengyere (2), Murukoro (2), Kishougati (1), Butengo (1) 6 
 MbararaNyabugando (2), Ikandagiro (1), Karabo (3) 6 

Characterization of isolates

All isolates from the 1995 collection were assessed for all markers described below. All isolates from the 1997 collection were assessed for Gpi genotype and mating type. From the second collection, 36 (18 from each country) were assessed for mitochondrial DNA haplotype and 25 (13 from Kenya and 12 from Uganda) were assessed for RFLP fingerprint. Peptidase (Pep) genotype was not assessed.

RFLP fingerprints were obtained using the moderately repetitive probe RG57 ( Goodwin et al., 1992a ). DNA was extracted as described previously ( Ordoñez et al., 2000 ). Two micrograms of DNA from each isolate were digested with EcoRI for 24 h, then underwent electrophoresis on a 0·7 or 0·8% agarose gel (56 V, 20 mA) for 24–45 h in 1 × TBE. Hybridization and detection were done using the nonradioactive kits ECL (Amersham, Inc., Buckinghamshire, UK) for the 1995 sample and Renaissance (NEN Life Science Products, Boston, USA) for the 1997 sample, according to the manufacturers' instructions.

Mitochondrial DNA (mtDNA) haplotypes were determined by amplification of DNA of individual isolates using primers designed for specific regions of the mitochondrial genome of P. infestans ( Griffith & Shaw, 1998). Digestion of the amplified regions with CfoI, MspI and EcoRI restriction enzymes yielded band patterns by which the isolates could be classified into one of four different haplotypes: Ia, Ib, IIa and IIb ( Carter et al., 1990 ; Griffith & Shaw, 1998).

Isozyme electrophoresis was conducted for the enzyme glucose-6-phosphate isomerase (Gpi) on cellulose acetate ( Goodwin et al., 1995a ) for the 1997 collection, and for Gpi and peptidase (Pep) on cellulose acetate, polyacrylamide gels or potato starch gels for the 1995 collection. Polyacrylamide gel electrophoresis (PAGE) was done using 1-mm-thick 7·5% gels with 25 m m Tris and 0·19 m glycine at pH 8·5 as separating gel and electrode buffer. Bands were clearer when a 1-cm stacking gel (2·5% acrylamide, 0·06 m Tris-HCl, pH 6·7) was used ( Davis, 1964). PAGE gels were run with a constant current of 5 mA for 1 h, which was then increased to 10 mA. Voltage rose continually throughout the run, from about 50 V to 280 V. PAGE electrophoresis was terminated when the bromophenol blue dye reached the bottom of the gel, a distance of about 16 cm. Ten isolates from the 1995 collection were also assessed for Gpi genotype on potato starch gels ( Spielman, 1991) to confirm the 100/100 banding pattern. To date it has been possible to separate the 100/100 banding pattern from the 90/100 banding pattern only on starch gels. These two patterns were identical on cellulose acetate or PAGE. Allozyme genotypes (inferred from banding pattern phenotypes on cellulose acetate, PAGE and starch) were scored as described previously ( Tooley et al., 1985 ), and represent the mobilities of the enzyme alleles relative to an allele designated as 100. Isolates with known alleles from the Cornell University collection were used for comparison.

All isolates were paired with others of known A1 and A2 mating type on clarified rye A medium ( Caten & Jinks, 1968) or clarified V8 juice medium. The presence or absence of oospores was recorded after 15 days of incubation at 18 °C in the dark.

Two inoculation tests (hereafter referred to as aggressiveness assays) were done on detached leaflets of potato and tomato to compare lesion size and symptoms caused by a subset of isolates from 1995. Since isolates had been maintained in culture for approximately three years prior to testing for lesion development, they were screened in a preliminary assay. Many isolates (especially those from potato) had lost the ability to infect either host, causing small water-soaked lesions with little sporulation or no lesions at all. Therefore, the first criterion for selection of isolates was that they caused sporulating lesions on both hosts. Based on this preliminary assessment, six isolates from each host were chosen for aggressiveness assays.

In the first aggressiveness assay, five isolates from potato and five from tomato were inoculated on three potato cultivars (Yungay, Cruza−148 and Chata Blanca) and three tomato cultivars (Caribe, Flora Dade and FMX-93). All potato and tomato cultivars are free of known genes for qualitative resistance. In the second test, four isolates were repeated from the first test (two from each host) and two isolates (one from potato and one from tomato) were assessed for the first time. The same potato and tomato cultivars were used in the second test. For both assays, plants were grown in a greenhouse in 20-cm plastic pots filled with a pasteurized mixture of equal parts of soil from earthworm culture, field soil and granular pumice.

For both assays, fully expanded leaflets were chosen from plants that were between six weeks old and initiation of flowering. Tomato leaflets were sometimes used after flower initiation but flower buds were removed. Leaflets were placed in Petri dishes as described previously ( Forbes et al., 1997 ) and inoculated by placing one 10-µL drop of inoculum near the midrib of each leaflet, thus producing one lesion per leaflet. Petri dishes were incubated at 15 ± 2 °C with 14 h per day of fluorescent light (2900 lux). After seven days of incubation, lesion diameter was measured with a ruler parallel to the leaflet midrib. Lesion growth was generally restricted by the width of the leaflet in the dimension perpendicular to the midrib but lesion expansion continued parallel to the midrib throughout the duration of the assay.

Four leaflets, which were distributed randomly in two Petri dishes, represented each host genotype by isolate combination. The experimental unit for this analysis was the average diameter of the two lesions in each Petri dish. Therefore, there were two data values for each host genotype by isolate combination. Since these values are really subsamples of each combination they were considered pseudo-replicates ( Hurlbert, 1984). The variance among pseudo-replicates was not used in tests of statistical significance. The statistical model, including appropriate F-tests, was described in an earlier report ( Oyarzun et al., 1998 ).


All isolates collected in 1995 and 1997 belong to the US−1 clonal lineage, based on two or more of the following markers: RFLP fingerprint ( Fig. 1), 92/100 Pep genotype, Ib mtDNA haplotype and A1 mating type ( Goodwin et al., 1994 ; Griffith & Shaw, 1998). Not all isolates from the 1997 collection were tested for all markers; however, these isolates were all 86/100 for Gpi. More than half of the 1997 isolates were assessed for RFLP fingerprint and mtDNA haplotype and all these were typically US−1. Among clonal populations of P. infestans, the 86/100 Gpi genotype is primarily associated with the US−1 clonal lineage. These observations are consistent with the hypothesis that all isolates from the 1997 collection belong to US−1.

Figure 1.

RFLP fingerprints of three isolates of Phytophthora infestans from potato (P) and three isolates from tomato (T) collected in Kenya and Uganda in 1995. All are typical US−1 fingerprints. Numbers on the right represent conventional numbering of bands ( Goodwin et al., 1992a ). These are the same isolates shown in Fig. 2.

Isolates from potato from the 1995 collection had the 86/100 Gpi genotype, but all isolates from tomato produced a single band at the 100 migration distance on cellulose acetate ( Fig. 2) or polyacrylamide. Using these two electrophoresis techniques, a single band was produced at the 100 migration distance for both the 100/100 and 90/100 Gpi genotypes. Therefore, 10 isolates from tomatoes collected in 1995 were tested with potato starch gel electrophoresis, which does separate the 90 and 100 Gpi alleles. All 10 isolates were 100/100 for Gpi, thus confirming that the tomato-adapted population of P. infestans collected in 1995 was a variant of US−1, which is homozygous for Gpi. This variant has been identified as US−1.7 ( Goodwin & Drenth, 1997).

Figure 2.

Glucose-6-phosphate isomerase banding pattern on cellulose acetate for three isolates of Phytophthora infestans from potato (P) and three isolates from tomato (T) collected in Kenya and Uganda in 1995. The three isolates from potato are 86/100 genotype; the isolates from tomato are 100/100. These are the same isolates shown in Fig. 1.

The interaction between the host and the source of the pathogen (potato or tomato) was highly significant (P < 0·0001) for both detached-leaf assays ( Table 2), and was evident by visual examination of plotted means of lesion length ( Fig. 3). On potato leaflets, isolates from potato caused larger lesions than did isolates from tomato. On tomato leaflets, isolates from this host caused the largest lesions, although the difference was only significant in one test, based on overlap of 95% error bars ( Fig. 3).

Table 2.  Analyses of variance from two experiments that tested effects of origin (potato or tomato) of isolates of Phytophthora infestans and inoculated host species (potato or tomato) on diameter of lesions (cm) in a detached-leaf inoculation assay involving isolates collected in Kenya and Uganda in 1995
Sourced.f. aMean squareF value bP > F
  • a

    Degrees of freedom.

  • b NT, not tested. Main effects of isolate origin and host species were not tested because of their highly significant interaction. This interaction, designated O –H in the table, was tested using the mean square for the interaction between individual isolates and plants, which is designated I O –P H in the table. This interaction was also used to test the main effects of isolate embedded in origin, IO, and plant embedded in host, P H.

  • c

    Based on variance among Petri dishes, which are pseudo-replications of the experiment.

First experiment
 Isolate origin (O)13·71NT 
 Host species (H)117·23NT 
 Isolate embedded in origin (IO) 83·379·360·0001
 Plant embedded in host (PH) 414·4340·000·0001
 Residual error c600·25  
Second experiment
 Isolate origin (O)11·52NT 
 Host species (H)142·9NT 
 Isolate embedded in origin (IO) 45·4311·270·0001
 Plant embedded in host (PH) 43·276·790·0001
 Residual error c1080·38  
Figure 3.

Histograms of two aggressiveness tests based on mean lesion length on leaflets of potato and tomato caused by isolates of Phytophthora infestans collected from potato (hatched bars) and tomato (open bars) in Kenya and Uganda in 1995. For the first test (a), five isolates each from tomato and potato were assessed. For the second test (b), four isolates from the first tests were repeated (two from each host) and two new isolates were assessed (one from each host). Error bars represent 95% confidence intervals.

Symptom expression on tomato leaflets was also different for the two populations of P. infestans studied here. Isolates from tomato were highly biotrophic on tomato, producing little or no necrosis on tomato leaflets during the seven days following infection, even though abundant sporulation could be seen after four or five days. In contrast, isolates from potato sporulated less abundantly on tomato and induced dark pigmentation that was most visible on the adaxial side of the leaflets ( Fig. 4). This characteristic was relatively stable and was generally sufficient to determine whether an isolate belonged to the potato or tomato population of the pathogen. Differential symptom expression was not apparent on potato leaflets.

Figure 4.

Differential symptom expression on tomato leaflets of a tomato-adapted isolate (left) and potato-adapted isolate (right) of Phytophthora infestans.


The data obtained in this study strongly support the hypothesis that the populations of P. infestans attacking potato and tomato in Kenya and Uganda belong to the US−1 clonal lineage. All isolates collected in 1995, regardless of source, had an RFLP fingerprint, Pep genotype, mtDNA haplotype and mating type characteristic of the US−1 clonal lineage. If other clonal lineages or genotypes resulting from sexual recombination are present, they must exist in low frequencies in the areas sampled, or are restricted to areas not sampled. The principal potato regions and important tomato regions were sampled in both countries; therefore, it appears highly probable that US−1 is the unique or at least dominant pathogen population in this region.

The data also clearly show that tomato and potato are attacked by two separate, host-adapted populations of P. infestans in Kenya and Uganda. Potato and tomato populations of P. infestans could be distinguished based on isolation characteristics, Gpi genotype, differential lesion development on potato and tomato, and symptoms on tomato. Based on symptoms and lesion expansion, it is concluded that host adaptation is quantitative rather than qualitative, because isolates were generally more aggressive on their original hosts but also pathogenic on their alternative host. This is similar to a situation recently described in Ecuador ( Oyarzun et al., 1998 ), except that in Ecuador the host-adapted populations belong to different clonal lineages. Also, in Ecuador, Uganda and Kenya, host adaptation is very strong in the field, although only small differences are detected in the detached-leaf assays. Therefore, although lesion expansion (as measured in the detached-leaf assay) is correlated with host adaptation, other factors (e.g. sporulation, infection efficiency) are probably also involved in the field.

Adaptation of P. infestans to potato and tomato has been detected in other locations ( Brommonschenkel, 1988; Fry et al., 1991 ; Goodwin et al., 1992b ; Koh et al., 1994 ; Legard & Fry, 1996; Oyarzun et al., 1998 ; Fraser et al., 1999 ; Lebreton et al., 1999 ). In those cases, the two populations generally belonged to different clonal lineages, or were different genotypes from areas where sexual recombination occurs, but there were exceptions. Fraser et al. (1999) found that tomato was principally attacked by US−7, but they also found US−8 attacking both tomato and potato. Legard et al. (1995) found both tomato-aggressive (referred to here as tomato-adapted) and tomato-nonaggressive isolates within the US−1 clonal lineage. US−1, however, is now rare in the USA ( Goodwin et al., 1998 ). Sub-Saharan Africa seems to be the only area studied to date where both tomato-adapted and potato-adapted populations are made up exclusively of the US−1 lineage. It is hypothesized, however, that this situation probably exists in areas where US−1 is still the unique or dominant lineage on potato, such as Chile (unpublished data), Australia (unpublished data) and the Philippines ( Koh et al., 1994 ).

Although the present study demonstrates a common origin for the two host-adapted populations of P. infestans in Kenya and Uganda, it does not prove that these populations developed sympatrically . The population attacking tomato could have been introduced from another location where it evolved independently of potato late blight. The authors believe, however, that it is more probable that the tomato-adapted population of P. infestans in Kenya and Uganda evolved from the potato-adapted population and was not introduced. The principal reason for this is the apparent scarcity of transport mechanisms by which a tomato-adapted population could have been introduced. The authors are unaware of tomato plantlets or fruits being imported into this region, and commercial tomato seed is not considered to be a transport mechanism of P. infestans ( Vartanian & Endo, 1985). Furthermore, the disease developed on tomatoes soon after the pathogen was introduced on potato ( Cox & Large, 1960), so if two separate introductions were made, they would have occurred almost at the same time – which seems unlikely.

If one accepts the arguments given above, then the data support the hypothesis that aggressiveness of P. infestans on tomato in Kenya and Uganda is associated with reduced pathogenic fitness (lesion development) on potato. The tomato-adapted isolates were less aggressive on potato than were potato-adapted isolates. Tomato-adapted isolates also did not grow well in potato tubers, as evidenced by the difficulty in isolating them with tuber slices. Tomato-adapted isolates (US−1) in Ecuador ( Oyarzun et al., 1998 ) were less aggressive on potato than were potato-adapted isolates (EC−1). Differential adaptation was demonstrated in a field experiment in the USA ( Legard & Fry, 1996) and in Europe ( Lebreton et al., 1999 ). These data are also consistent with the hypothesis that adaptation to tomato is associated with a loss of pathogenic fitness on potato.

Nonetheless, other data would suggest that adaptation to tomato is not always associated with reduced fitness on potato. Lebreton et al. (1999) found that pathogenic adaptation alone could not explain the diversity of genotypes of P. infestans attacking potato in France. Some isolates from potato caused smaller lesions on that host than did isolates from tomato. The authors hypothesized that lesion development may not be the best measure of adaptation in this case. Survival ability (overwintering) or random genetic drift may contribute to the persistence of weakly pathogenic genotypes in the potato population. In a study conducted recently in the south-eastern USA, the US−8 lineage was found on both potato and tomato ( Fraser et al., 1999 ). Since aggressiveness tests were not done, it is not known if US−8 is highly pathogenic on both hosts, or if host-specific adaptation has developed within the US−8 lineage. Legard et al. (1995) reported that all isolates they examined were aggressively pathogenic on potato.

Lesion development studies of the kind conducted here and by others on detached leaflets ( Legard et al., 1995 ; Oyarzun et al., 1998 ; Lebreton et al., 1999 ) could involve artefacts that might affect the interpretation of results. A small number of isolates were tested in this study because all were first screened for aggressiveness and it was found that a large number of those from Kenya and Uganda (especially those from potato) did not cause normal lesions on either potato or tomato. Instead, these isolates caused small, water-soaked lesions with little sporulation, or caused no lesions at all. It is not known why so many of these isolates lost aggressiveness after culturing, but the authors have observed this phenomenon before with P. infestans, although to a lesser extent. Loss of aggressiveness of isolates of P. infestans after culturing has been reported by other authors ( Gallegly, 1968; Miller et al., 1998 ; Ordoñez et al., 1998 ).

Selection of parameters used to measure lesion development could also affect the interpretation of detached leaf assays. Lebreton et al. (1999) measured lesion size as the proportion of the half leaflet showing late blight symptoms. In the present study, and an earlier one done in the same laboratory ( Oyarzun et al., 1998 ), lesion size based on the limits of sporulation was measured. This is an important distinction, because, as noted above, host adaptation in this system appears to be manifested in both the size and type of lesion. The potato-adapted isolates studied caused darkly pigmented lesions on tomato, which might be classified as a partially necrotrophic reaction. The lesions that tomato-adapted isolates caused on tomato were typical of a biotrophic reaction and were not easy to see until sporulation was abundant, after four or five days. This might explain why Lebreton et al. (1999) found that isolates from potato caused the largest lesions on tomato leaflets after three days of incubation, while isolates from tomato caused the largest lesions after four or more days of incubation.

Symptom expression in this system leads to interesting speculation about the nature of host adaptation. The strong pigmentation induced on tomato leaflets by potato-adapted isolates might indicate that a pathogen product, present only in potato-adapted isolates, is being recognized by the host. Loss of the product may lead to tomato adaptation, but apparently at a cost to the pathogen, since all the tomato-adapted isolates tested were less aggressive on potato, although sample size for the detached leaf assays was small.

The hypothesis of host-specific aggressiveness could be better tested by conducting aggressiveness studies with the assessment of more components (e.g. sporulation, infection efficiency) and with a larger sample size, in countries where US−1 is still the dominant lineage on both hosts. Field experiments should be done to test competitive fitness on potato and tomato, since additional factors in the field could be very important. Legard & Fry (1996) and Lebreton et al. (1999) performed field experiments to compare isolates from potato and tomato that could be differentiated by allozyme analysis. The same approach could be used in Uganda and Kenya, where host-adapted populations differ at the Gpi locus. A simple test using cellulose acetate electrophoresis ( Goodwin et al., 1995a ) could be used to rapidly monitor populations in field studies.

The time at which the change of the Gpi locus occurred in the Kenya/Uganda population of P. infestans remains unclear, as do the factors that may have led to the apparent universality of 100/100 Gpi genotype within the tomato-adapted population. Change from 86/100 to 100/100 within the US−1 clonal lineage has been reported previously ( Goodwin et al., 1994 ; Forbes et al., 1998 ), and this variant of US−1 was named US−1.7 ( Goodwin & Drenth, 1997). The US−1.7 genotype is represented in a global marker database for P. infestans by one isolate from Brazil, which was isolated from tomato ( Forbes et al., 1998 ). However, US−1 (Gpi = 86/100; Pep = 92/100) and US−1·1 (Gpi = 86/100; Pep = 100/100) have also been isolated from tomato in Brazil ( Forbes et al., 1998 ). Tomato adaptation has been associated with many different genotypes of P. infestans worldwide ( Legard et al., 1995 ; Oyarzun et al., 1998 ; Fraser et al., 1999 ; Lebreton et al., 1999 ). Therefore, it would appear that tomato adaptation has evolved independently in different parts of the world.