In December 2016, three screenhouses (3 x 3 x 2 m, length x width x height) were set up in a greenhouse (30 x 15 x 5) at the National University of Salta, Argentina (southern hemisphere). All-purpose garden fabric (Tela antihelada, Marplast SRL, Salta, Argentina, 26.2 g/m²) was used to cover the screenhouses, allowing free passage of oxygen and light while preventing mealybugs from escaping. Planococcus ficus mealybugs were obtained from the National Agricultural Technology Institute (Instituto Nacional de Tecnología Agropecuaria, INTA) in Cafayate. Planococcus ficus was reared on sprouted potatoes in plastic containers (20 x 40 cm) covered with cotton fabric at 23 ± 1 °C and ambient light conditions. Each screenhouse was equipped with seven three-year-old vines (cv. Torrontes) in 30 x 50 cm containers, which were watered twice a week. The plants did not receive fungicide treatments or fertilisation throughout the duration of the experiment. Five months prior to the experiment, the vines had been transferred from the field to pots, and were pruned to one spur on that occasion. The three screenhouses corresponded to three treatments: 1) High density of P. ficus, 2) low density of P. ficus, and 3) a control treatment without P. ficus. Infestation levels aimed to mimic field data on low and heavily infested vineyards (Daane et al., 2007; Geiger and Daane, 2001; Walton et al., 2006). Vines were infested using leaf disk transport (Hogendorp et al., 2006; Schulze-Sylvester and Reineke, 2019). Vines were randomly assigned to the different treatments and were placed separately inside the screenhouses. In December, each plant in the high-density treatment received approximately 1000 1st instar P. ficus nymphs, and the low-density treatment was infected with approximately 100 nymphs per plant, while the control treatment plants received empty leaf disks. Infestations were repeated in early January and early February to ensure high/low infestation levels. The containers were painted with a sticky, non-toxic paint (CeroPestes Hormiga, Sanipro SRL, Buenos Aires, Argentina) to avoid mealybug tending by ants.

Between December 2016 and May 2017 (early summer to late autumn in Argentina), leaf number and leaf area were measured monthly by taking standardised photos of all leaves on all plants, resulting in seven measurements per treatment per sampling date. The surface area of each leaf was determined using the software ImageJ (National Institute of Health, Bethesda, MD, USA). The average leaf size per plant was calculated dividing the plant’s total leaf area by its leaf number. Furthermore, mealybug densities in the three treatments were determined at the end of every month (31 December to 31 May) by carrying out three-minute time counts of visible mealybugs on all plants (Geiger and Daane, 2001), resulting in seven biological replicates per treatment. These time-counts were completed by a final, destructive count of all mealybugs on five plants.

At the end of the experiment, plants were cut at the root collar. To study the biomass allocation in the experimental vines, additional morphological measurements were performed on the plant material. Three leaves from five plants per treatment were randomly sampled, measured, oven-dried for three days at 65 °C (SL30C, San Jor, Buenos Aires, Argentina) and weighed. The specific leaf mass, (leaf mass per area, LMA) calculated as leaf dry mass/leaf area, was then determined (Pérez-Harguindeguy et al., 2013). Five randomly chosen plants of each treatment were used for further analysis of total mealybug number and dry biomass weight. For this, the plant material was cleaned of honeydew and sooty mould with tap water; mealybugs were collected and counted. After that, all above-ground plant parts were oven-dried for one week at 65 °C and the dry weight was recorded for leaves (leaf biomass), stems (woody parts and petioles), and total plant parts (total biomass, all aerial plant parts). The leaf area ratio (LAR), dividing the plant's total leaf area by its weight (stem and leaves) was determined (Pérez-Harguindeguy et al., 2013).

Due to the prolonged experiment duration (6 months), the advanced plant sizes and the frequent measurements and maintenance activities, individual “bagging” of plants was not a viable option. Also, the number of available screenhouses was limited and the randomisation of plants from different treatments within the screenhouses is not advisable if cross-contamination is to be avoided. For the statistical analysis, individual vine plants were therefore considered as biological replicates.

Data were tested for normality using Shapiro-Wilks normality test. Mean mealybug densities, leaf area, average leaf size and number of leaves were tested for differences using two-way repeated-measures ANOVAs, followed by Holm-Sidak’s multiple comparison test. Deviations from sphericity were quantified and corrected using Greenhouse and Geisser’s epsilon. Treatment differences for the biomass of leaves, woody parts and total plant parts, as well as LMA and LAR, were analysed with one-way ANOVAs and subsequent Tukey tests. Mealybug three-minute counts were checked for correlation with total destructive counts using Pearson’s test. Pearson’s correlation analysis was also used to understand how total mealybug numbers were related to the biomass of leaves, stems and total biomass. All analyses were performed with GraphPad Prism version 8.00 for Windows (GraphPad Software, La Jolla, CA, USA).

No significant differences between treatments were found for the leaf area, leaf size and leaf number per plant during the course of the experiment at any time (Figure 1). Not surprisingly, the leaf area and number per plant increased in all three treatments over the duration of the experiment (Figure 1A, C; F = 57.96; df = 2.055, 36.98; P < 0.0001 and F = 94.29; df = 1.818, 32.72; P < 0.0001 respectively). Meanwhile average leaf size per plant remained constant (high-density treatment) or was reduced to a constant level after January (low-density and control treatment) (Figure 1B; F = 28.06; df = 1.673, 30.11; P < 0.0001). One-way ANOVAs of the aboveground leaf area ratio (LAR) and the leaf mass per area (LMA) showed significant differences between treatments (Table 1; F = 9.11; df = 2, 12; P = 0.004). The LAR was highest for high-density plants, while the LMA showed the lowest leaf mass per area in the high-density treatment (Table 1; F = 5.60; df = 2, 12; P = 0.02).

Table

A) Leaf area, B) leaf size, and C) leaf number of grapevine plants infested with different P. ficus densities (high, low, control without P. ficus); n = 7 plants, results are expressed as mean values ± SD)

The density-dependent effects of P. ficus on its host plants have never been considered so far. In congruence with our hypothesis, the present data show a negative relationship between biomass and mealybug numbers. This relationship is significant for the total biomass and the leaf biomass, while the stems show a slight, but not significant, negative trend. The density-dependence of the total biomass reduction is mainly due to the effects on the leaf biomass, while stem biomass decreased equally for both treatments. This suggests that stems are more sensitive to mealybug feeding than leaves. Another explanation could be that the plant redirects stored resources from the stems to maintain leaves (source tissue), but eventually, when a certain infestation threshold is exceeded, leaf biomass also declines. Comparing our results with the few available studies involving sap-feeder densities, the citrus mealybug, P. citri, showed a density-dependent effect on the above-ground biomass of the shrub, C. odorifera, while the roots did not seem to be affected (Mills, 1984). Frosted scales showed a density-dependent decrease in internodes in grapevine, while effects on the chlorophyll concentration and branch length were not density-related (Simbiken et al., 2015). Scale insects also showed strong density-dependent effects on the root, stem and leaf biomass in eucalyptus plants (Vranjic and Ash, 1997).

Mealybug populations are very dynamic and infestation levels can increase dramatically within weeks (Becerra et al., 2006; Geiger and Daane, 2001). Spring populations are typically considered low to moderate when timed counts detect < 10 adult mealybugs, and high when > 10 mealybugs are found (Walton et al., 2006). In autumn, Charles (1981) counted up to 1800 individuals per leaf which corresponded to three generational cycles. In Argentina, six generations of mealybugs are common (Becerra et al., 2006), hence infestation levels might be even higher. Mealybug numbers obtained in the present study resemble infestation levels reported for vineyards in California and Auckland, New Zealand (Charles, 1981; Daane et al., 2007; Geiger and Daane, 2001; Walton et al., 2006) and Salta (Schulze-Sylvester, unpublished data).

Typically, P. ficus overwinters in small numbers in refuge areas under bark and on the roots and follow the plant's resources during spring-summer by moving from roots to shoots to leaves, and eventually to fruit clusters (Daane et al., 2012). In local vineyards in Salta, Argentina, mealybugs are usually not detected in the canopy until February (personal observation).

Experimental plants were visibly infested on stems, petioles and leaves with different mealybug densities around mid-February, and time counts had already revealed the first statistically relevant differences between mealybug densities in January. After a steep population growth between February and April, growth rates slowed down, suggesting a density-dependent restriction of mealybug population growth due to limited resources or crowding effects (e.g., reducing size and fecundity, and increasing mortality) (Washburn et al., 1985). Consequently, the per-capita impact of P. ficus on plant biomass declined with increasing density. Similar to the study by Geiger and Daane (2001), our results show good correlations between the three-minute counts and the destructive sampling method in the low-density treatment, but not in the high-density treatment. It is possible that there is a maximum number of individuals that can be counted in three minutes and that we reached this limit. We suggest extending the interval for time counts to 5 minutes, especially when mealybug densities are high.