Watermelons employed in this study came from several sources covering two crop years. In 2007, watermelons came from a commercial field in Hinton, OK, USA. They were purposely selected because they had been graded 'culls' as the result of an anthracnose infection on their outer rinds. Other sources of watermelons included those from the 2007 crop in Terral, OK, USA, some from the 2008 crop in Louisiana, USA, and others from the 2008 crop raised at the South Central Agricultural Research Center, Lane, OK, USA. Lycopene-free juice was a waste stream from processing watermelon flesh to produce lycopene-containing chromoplasts by a procedure previously described (Fish, US Patent Appl. 60/752.279, 2005). Amino acid-free juice was prepared from the lycopene-free watermelon juice by procedures developed in this laboratory (Fish, unpublished). Watermelon juices at the various stages of processing were frozen and stored at -20°C until used. Granulated sugar and molasses employed for fermentations were purchased at a local supermarket.
Fermentations were conducted in a BF-110 benchtop modular fermentor system (New Brunswick Scientific Co., Inc., Edison, NJ, USA). The system included a 7.5 l thermostatted glass vessel, a pH/DO controller, a four-pump reagent addition module, an exhaust condenser, a DO probe, and a pH electrode. The fermentor was sanitized between fermentations by washing with detergent, rinsing with water, treating with 5% (v/v) Oxonia Active™ (Ecolab Inc., St Paul, MN, USA) for 6 h at room temperature, and rinsing thoroughly with water. Most of the experimental fermentations were conducted on a volume of 2 to 3 L of feedstock. Temperature was controlled at 32°C, and the medium stirred at 100 rpm to keep the yeast, substrate, and products evenly distributed. Media were routinely inoculated with hydrated and conditioned dried yeast to provide an initial yeast population of ~107 viable cells/ml in the fermentation medium. The pH of nearly all fermentations not pH-controlled quickly dropped to ~pH 2.8 and remained there regardless of the starting value. For those fermentations conducted at a pH other than ~3, the fermentor controller was programmed to add 1 M NaOH when needed to maintain the pH of the medium at the level desired. The starting pH of the processed watermelon juice used to prepare feedstocks was generally pH 3 so that the pH of the medium initially had to be adjusted to the value desired for the fermentation run. Fermentation runs under a specific set of conditions were repeated at least once, and in most cases, three replicate fermentations were performed. Data presented in the Figures and the text are the means of the replicates, and their standard deviations are represented by error bars in the Figures.
Yeast conditioning and nutrition
The yeast (Ethanol Red™) employed for fermentations was a commercial dried yeast (Fermentis of Lesaffre, Milwaukee, WI, USA). Sufficient active dried yeast (0.5 g/L of fermentation medium) was rehydrated with the goal of providing a starting concentration of ~107 viable cells/ml in the fermentation medium. The yeasts were rehydrated/conditioned by first suspending into water an amount of Go-Ferm™ (Lallemand, Montreal, Canada) that equaled 1.25 parts by weight of that of the dried yeast that was to be used. The Go-Ferm™ was suspended into 20 times its weight of sterile H2O at 43°C. Once the temperature of the Go-Ferm suspension reached 39 to 40°C, the dried active yeast was added and gently stirred to evenly disperse the cells. The suspension was allowed to stand for 15 to 30 min during which time the yeast slurry cooled to ~32°C. The yeast slurry was then added to the feedstock (already at 32°C) in the fermentation vessel. Nitrogen supplementation was provided, when desired, by the addition of diammonium phosphate (DAP) (Sigma-Aldrich, St Louis, MO, USA) and/or yeast extract, HY-Yest 412 (10.2% total nitrogen; 5.2% amino nitrogen) (Sigma-Aldrich, St Louis, MO, USA). The yeast nutrient complex (Fermaid K™, Lallamand, Montreal, Canada), was routinely added at 0.25 g/L to the ongoing fermentation when the starting sugar levels were 1/3 depleted. For fermentations that utilized greater than 20% (w/v) sugars, yeast hulls were added at a level of 0.2 g/L to furnish additional sterols to the yeast and to adsorb part of the yeast autotoxins that are produced during fermentation.
Yeast and bacterial counts
The density of viable yeast cells was routinely estimated by determining the absorbance at 420 nm of a diluted sample of fermentation medium. This relationship had been determined earlier by sampling, diluting, plating, and counting colony-forming units from fermentations that were in the log phase of growth. The number of colony-forming yeast cells per ml of media was then correlated with the absorbance at 420 nm of a sample from the fermentation medium measured at the same time as the sample was removed for plating. Microbial counts were taken from media after 48 h of fermentation. Dilutions of the fermentation medium were plated onto nutrient agar that contained 10 μg/ml of cyclohexamide (Sigma, St Louis, MO, USA) to inhibit yeast growth.
Determination of ethanol and CO2 production rates
The rates of ethanol production at various pH values were estimated by plotting the ethanol levels versus fermentation times between 12 and 28 h and fitting the data with a linear least squares equation. The slope of the line was then used as a measure of the maximal rate of ethanol production. Although it is recognized that this type of direct plot is not truly linear, data in this region approach linearity (0.990 <R
2 < 0.999), and the slope of a linear fit to the data approaches the slope of a tangent to a point in this region.
The rate of CO2 evolution from the fermentor was estimated by counting the rate of bubbles released into the water trap per unit time. The conversion from number of bubbles per min to liters of CO2 per h was possible after first determining the number of CO2 bubbles from the water trap that it took to yield a liter of CO2 gas by displacement of water from an inverted water-filled graduated cylinder.
High performance liquid chromatography (HPLC) was carried out on a Varian ProStar ternary solvent system equipped with an autosampler and diode array and RI detectors. Quantitative carbohydrate profiles of the feedstocks were obtained with a 250 mm × 4 mm 5 μm Luna™ amino column (Phenomenex, Torrance, CA, USA). The sugars, glucose, fructose, and sucrose, were eluted with an isocratic system of 80% acetonitrile/20% H2O at a flow rate of 1 ml/min and a column temperature of 35°C. Fermentation substrates, glucose and fructose, and fermentation products, including ethanol, acetate, glycerol, citrate, and lactate, were separated and quantified on an Aminex™ HPX-87H 300 mm × 7.8 mm column (BioRad, Hercules, CA, USA). Components were eluted with an isocratic system of 5 mM H2SO4 in H2O. The flow rate was 0.6 ml/min, and the column temperature was maintained at 50°C. Three standard solutions containing fructose, glucose, and ethanol at different concentration levels were run each time a fermentation analysis was performed in order to confirm the fidelity of the respective calibration curves. In order to quantify fermentable sugars with the Aminex™ column during fermentation runs, samples that contained sucrose had to be pre-treated with invertase before column chromatography. This was necessary because the acid conditions of the 5 mM H2SO4 eluting solvent catalyzed hydrolysis of the glucose-fructose glycosidic bond during time on the column and created a reaction zone so that none of the three sugars could be quantified. Pre-column hydrolysis of sucrose with invertase was performed by incubating 100 μl of sample (up to 40% sucrose) with 100 μl of yeast invertase (Sigma, St Louis, MO, USA) at ~100 U/ml in 0.05 M sodium acetate buffer, pH 4.6, at 37°C for 30 min. By the end of the incubation, all sucrose had been hydrolyzed to glucose plus fructose, and these two hexose sugars could be separated and quantified by HPLC analysis on the Aminex™ column. Samples for analysis during fermentation were collected via the sterile sampler on the fermentor. Two 1 ml aliquots of the fermentation broth were centrifuged for 2 min at 10,000 × g on a microfuge (Eppendorf, Hamburg, Germany) to pellet all insoluble material. After invertase treatment and appropriate dilution (~50-fold), samples for HPLC analysis were filtered through 17 mm disposable nylon filters of 0.45 μm pore size before being placed in the autosampler. A sample loop of 100 μl was employed in the autosampler. Amino nitrogen was estimated in feedstocks by the ninhydrin method . The concentrations of L-citrulline in the respective watermelon juices were estimated by separating the juice amino acids by thin layer chromatography  and comparing the intensity of each juice's L-citrulline spot, after staining with ninhydrin, to those of authentic L-citrulline co-developed on the chromatogram at known concentrations. The concentration of total nitrogen in a watermelon juice was estimated by measuring its amino nitrogen and adjusting for the additional nitrogen furnished by its L-citrulline content. This then assumed that all other amino acids in the watermelon juice each contained only one nitrogen. So while the estimate for amino nitrogen is accurate, the estimated numbers for watermelon juice total nitrogen may be as much as 5 to 15% lower than the true value since there was no quantitative measure of those amino acids that contained more than one nitrogen.