Sampling strategyA hierarchical sampling strategy was employed to collect samples (stem cuttings) from the different locations. Wild material was collected from seven tropical natural forests: (i) Budongo forest; (ii) Itwara Central Forest Reserve; (iii) Kalangala (Lutoboka central forest reserve); (iv) Kibale forest national park; (v) Mabira forest reserve; (vi) Malabigambo forest and (vii) Zoka forest. In each location (forest) except Kalangala, samples were collected from five sub-sites that were separated by distances of at least 5 km. From each sub-site, five healthy C. canephora trees were identified from which we collected stem cuttings. Since C. canephora is an allogamous species, each sampled plant was considered to be genetically unique and therefore, each sampled tree was regarded as a distinct genotype in this study. The assumption that each sampled tree is a unique genotype was confirmed by genetic analysis in Kiwuka et al., (2021). Contrary to other locations, the Kalangala site comprised remnants of natural forest systems and secondary forests regenerated from formerly cultivated coffee fields, and therefore, the coffee populations in Kalangala were considered wild or feral depending on where they were collected from. Samples that were collected from natural forest fragments were regarded as wild while samples from collected abandoned cultivation fields were considered as feral. The cultivated samples were collected from two germplasm field collections of the Ugandan National Agricultural Research Organization (NARO): National Coffee Research Institute Kituza and from the National Agricultural Research Laboratories at Kawanda. The cultivated genotypes were selected on the basis of their historical and passport data with the aim of representing the total range of traditional and commercially cultivated C. canephora diversity, including the two predominant forms found in Uganda: Erecta, or upright forms, and Nganda, or spreading forms (Thomas, 1935) and the six elite clones, namely: KW13, KW14, KW15, KW16, KW18 and KW19 (details can be found in (Kiwuka et al. , 2021)).
Stem cutting establishment
All the collected stem cuttings were rooted in a screen house at the National Agricultural Research Laboratories (NARL), Kawanda at 0º 25’ N, 32 º 32’ E, 1195 m a.s.l., starting on 30th May 2015. The establishment of the material from stem cutting followed a tested protocol by the National Coffee Research Institute (NaCORI, unpublished). The collected stem cuttings were cut into 7 cm inter-nodal wood cuttings with one pair of leaves. A total of 7,419 inter-nodal cuttings, for all the collected genotypes (230) were planted in poly-pots and placed in transparent plastic cages for root establishment. The number of cuttings per genotype ranged from 7-99 the median being 33. The poly-pots had a diameter of 5 cm and a height of 7 cm and were filled with a mixture of topsoil, sand and manure in a ratio of 3:2:2 by volume. Before planting, each stem cutting was dipped in rooting hormone (Seradix ’2ʹ, 0.8% w.w, IBA, Twiga Chemicals Industries, Nairobi, Kenya) to boost their rooting potential. After seven months, the young plants that had grown from the cuttings were hardened off and, transferred into 10 L pots. The potting medium comprised of black loamy forest soil, lake sand and decomposed cattle manure in the ratio of 3:1:1, with a volumetric water content of 30 % (± 0.22) at field capacity and 6 % (± 0.16) at permanent wilting point respectively (See details of the chemical and physical properties of the potting medium in Appendix Data File A.1.). Ten grams of an inorganic compound fertilizer comprising: 25% nitrogen, 5% phosphorous, 5% potassium and 5% of sulphur of the total weight of the elements in the fertilizer was added per pot. Pots were optimally irrigated for six months before starting the experimental treatments.
Experimental design
Out of the 230 collected genotypes, 148 produced sufficient number (≥5) of properly rooted plantlets to start the experiment with. From October 10th to 15th 2016, 16 months after re-planting the stem cuttings, 1184 rooted plants were arranged into a split-plot design; with two watering regimes (ample vs restricted-water) as the main factor and the different C. canephora genotypes as the sub-factors. Plants were grown in a ‘rain out’ screen house (40 m by 6.5 m) that was blocked into four sections, based on the variation in radiation that was visually assessed (148 remaining genotypes x 4 blocks (with each split into two) x 2 water regimes (ample and restricted).
To establish ample vs restricted-water availability treatments, we assessed the potting medium’s properties, e.g. water content at field capacity, permanent wilting point and the daily evapotranspiration rates within the screen house by weighing over time a selection of 10 pots. Soil water loss was also estimated from monitoring soil moisture content in pots using a soil moisture sensor (Trime-Pico 64/32, HD2 IMKO Micromodultechnik, Ettlingen, Germany). The ample-water treatment was set at 25 v% which was about 80% of soil moisture content at field capacity, while the drought-stressed regime (restricted-water) was sustained at 10 v% soil moisture at the permanent wilting point.
Plants in the ample-water treatment received on average 1000 ml of water per watering interval, which was, on average, once a week. Plants in the restricted-water treatment were subjected to gradually increasing severity of drought stress and the basic regime was that on average, plants received 300 ml per week for the first month, 300 ml per fortnight for the following month, a onetime 300 ml water gift in the third month and finally a month without water. To minimize the potential plant-size drought bias i.e., the fact that larger plants consume more water and are therefore exposed on average to drier conditions, the following procedure was used: in the initial experimental phase, a sub-set of plants (54 plants; selected to represent the architectural [number of leaves, number of primary branches, number of suckers and leaf area], variation across the experiment) were monitored to determine their soil water content (both gravimetrically and with the soil moisture probe) every week and their corresponding number of leaves, number of primary branches, the number of suckers and leaf area were non-destructively estimated. Leaf area of fully expanded leaves was estimated from leaf length and width using the linear model (area per leaf = leaf length x leaf width x k (k=correction factor = 0.66)) of Schmildt et al., (2015). These data yielded a correlation between leaf area and water loss and the relation was used as a guide to determine the frequency of watering for every plant based on its leaf area. This procedure ensured that size-dependent effects on the actual soil moisture experienced by plants were minimized. At the end of the experiment, it appeared that the amount of water supplied (W (ml)) could be linearly related to the leaf area (L (cm2)) to each plant was described by the formula: W = 1479 + 0.178 L ), p = 0.000 and R2 = 0.27.
The experimental treatment period lasted four months (from plant age 20 months to 24 months; age zero is when the stem cuttings were planted to root). Data on temperature and relative humidity in the screen-house were recorded by sensors with data logging (Tinytag logger Plus 2 Dual Channel Temperature/Relative Humidity, TGP-4500, Gemini data loggers Ltd., Chichester, Chichester West Sussex, UK) on an hourly basis. The average daily temperatures and relative humidity of the screen-house throughout the experimental treatment period were: 23.1 º C (± 4.3) and 83.1 % (± 18.0) respectively while average daily vapour pressure deficit estimates were 0.49 (± 0.15).
Data collection
Data were collected at three stages: (i) at the start of the treatment phase; (ii) during the treatment phase and at (iii) at the end of the treatment phase (Appendix Table A.2.). At the start of the treatment phase on 25th May 2017 (plant age 20 months) several non-destructive measurements were done to provide a baseline for later size increment measurements: plant height, number of nodes, number of leaves (fully grown and proportion/fractions from estimated full size of developing ones), length and width of fully expanded leaves and stem diameter at 5 cm from the base. After these measurements, the youngest fully expanded leaf pair was marked, to establish a recognition point for measuring new growth. The second data collection stage (at the point when 10 % of the plants subjected to drought treatment started to exhibit leaf wilting (scored visually), was taken 21-24 June 2017. The final measurement occasion, at the end of the treatment phase, was conducted at 12-26 September 2017, with measured traits as listed in Appendix Table A.2.
Methods to measure plant properties
Plant height was measured using a meter ruler from the base (point of origin from the cutting) to the last node. To estimate area per leaf and subsequently the total leaf area, we used the same model as that used for determining leaf area in relation to the watering regimesi.e. we measured length and width and then used the linear model (leaf length x leaf width x k (correction factor)) of (Schmildtet al. , 2015) on all fully unfolded leaves and obtained a correction factor (k of 0.66) that was used on all measured leaves. Leaf area on the main stem was measured in this way for all plants. But due to the necessity to reduce the workload, the leaf area of primaries and suckers was measured using the aforementioned linear model, but only for all plants in one block. For every genotype the leaf area of primaries and suckers in blocks two, three and four were estimated from the ratio of leaf fresh weight to leaf area, generated from the measured plants in block 1. At the end of the experiment, for each plant, leaves were separated into leaves from the main stem, primaries and suckers. To obtain total leaf fresh weight (TLFW) and total leaf dry weight (TLDW), the fresh weight of all leaves was estimated by weighing fresh leaves while leaf dry weights were measured after oven drying (70 °C to a constant weight).
Specific leaf area (SLA) was estimated as the ratio of leaf area and leaf dry weight accumulated within the experimental treatment phase. The roots of each plant were harvested and cleaned under running water and on a wire mesh. Using the water displacement method, fine roots (excluding the taproot with a diameter larger than 3 mm) were dipped in a measuring cylinder to estimate their root volume. The root volume and total leaf area (TL) of each plant were used to estimate the root volume to leaf area ratio (RL). Four growth-related traits were used to characterize the genotype responses to drought stress. These were relative growth rate in leaf area (RGRA, see below for how this was calculated), the total number of leaves (TNL) total leaf area (TL[cm2]), total leaf dry weight (TLDW [g]), specific leaf area (SLA[cm2 g-1]) and root volume to leaf area ratio (RL [cm3cm-2]). Note that all traits, except root volume, refer to growth during the experimental period, excluding the biomass at the start of treatments. RGRA was used to assess in more detail the cultivation status, location and genotype response to restricted and ample availability of water. Relative growth rates were used for two reasons: (i) to reduce confounding effects of initial plant size and (ii) we dealt with very young plants for which the assumption of them being in exponential growth phase was reasonable. We focused on area, dry mass and number of leaves because of practical reasons (measurable non-destructively; base measurements of biomass were not available) and because leaf area determines light interception capacity, photosynthesis and subsequent growth (Poorter and Remkes, 1990; Weraduwage et al. , 2015) (and in coffee fast vegetative growth are typically associated with high yields (Cilas et al. , 2006)).
RGRA was calculated as