Fasting in the ureotelic Lake Magadi tilapia, Alcolapia grahami, does not reduce its high metabolic demand, increasing its vulnerability to siltation events

Abstract Lake Magadi, Kenya, is one of the most extreme aquatic environments on Earth (pH~10, anoxic to hyperoxic, high temperatures). Recently, increased water demand and siltation have threatened the viable hot springs near the margins of the lake where Alcolapia grahami, the only fish surviving in the lake, live. These Lake Magadi tilapia largely depend on nitrogen-rich cyanobacteria for food and are 100% ureotelic. Their exceptionally high aerobic metabolic rate, together with their emaciated appearance, suggests that they are energy-limited. Therefore, we hypothesized that during food deprivation, Magadi tilapia would economize their energy expenditure and reduce metabolic rate, aerobic performance and urea-N excretion. Surprisingly, during a 5-day fasting period, routine metabolic rates increased and swimming performance (critical swimming speed) was not affected. Urea-N excretion remained stable despite the lack of their N-rich food source. Their nitrogen use switched to endogenous sources as liver and muscle protein levels decreased after a 5-day fast, indicating proteolysis. Additionally, fish relied on carbohydrates with lowered muscle glycogen levels, but there were no signs indicating use of lipid stores. Gene expression of gill and gut urea transporters were transiently reduced as were gill rhesus glycoprotein Rhbg and Rhcg-2. The reduction in gill glutamine synthetase expression concomitant with the reduction in Rh glycoprotein gene expression indicates reduced nitrogen/ammonia metabolism, most likely decreased protein synthesis. Additionally, fish showed reduced plasma total CO2, osmolality and Na+ (but not Cl−) levels, possibly related to reduced drinking rates and metabolic acidosis. Our work shows that Lake Magadi tilapia have the capacity to survive short periods of starvation which could occur when siltation linked to flash floods covers their main food source, but their seemingly hardwired high metabolic rates would compromise long-term survival.


Introduction
East African Rift valley alkaline soda lakes are an unusual and harsh habitat for living organisms. However, even when covered with thick soda crusts, they still contain viable niches in the lake margins, lagoons and the nearby springs that supply the lakes with volcanic groundwater (Fig. 1) (Coe 1966, Kavembe et al., 2016. One of these lakes is Lake Magadi in Kenya that has an extremely high pH (up to 10.0), extreme alkalinity (>300 mmol L −1 ), high temperature (>40 • C at some sites), high levels of reactive O 2 species (>8 μmol L −1 at some sites during daytime irradiation by high UV light levels), a salinity close to 60% seawater, and large daily fluctuations in O 2 levels ranging from anoxia or hypoxia to hyperoxia (Coe, 1966;Johansen et al., 1975., Narahara et al., 1996Johannsson et al., 2014, Wood et al., 2016. Nevertheless, it is inhabited by the Lake Magadi tilapia, Alcolapia grahami (Boulenger), the only fish species that survives there. Species that survive in extreme conditions have generally evolved special adaptations to do so, and the Lake Magadi tilapia is no exception. They are unique among teleosts in that they have a fully functional ornithine-urea cycle and are 100% ureotelic under all circumstances, an adaptation to the challenges of excreting ammonia into a pH 10 environment , Lindley et al., 1999. However, this unique fish is under threat from anthropogenic factors, and in 2006, A. grahami was placed on the IUCN Red List of Threatened Species, where it is classified as vulnerable (Bayona and Akinyi, 2006).
The tilapia live near the hot springs surrounding the lake, but the water supply from these sources is now in competition with boreholes that are being sunk around the lake to supply water to the growing local Maasai community and their cattle (Africa Geographic, 2011). Perhaps an even more severe threat is the increase in silt deposits in the lake caused by increased agricultural activity and deforestation, aggravated by road construction with inadequate culverts to drain storm water (Ministry of Environment and Forestry, 2015). Flash floods were reported to have deposited as much as 8000 t of silt after every storm into Lake Magadi (KBC Channel 1, 2015;People Daily, 2018). Such large deposits could cover the rocks encrusted with the thriving cyanobacteria Arthrospira sp., which comprise 90% of the Lake Magadi tilapia diet, which is complemented occasionally with copepods and chironomid larvae (Coe, 1966). As the fish often live in isolated springs near the edges of the soda crust, evasion possibilities are limited to absent which could leave them temporarily without access to food. Even under normal circumstances, the fish remain small and often have a very thin appearance, indicating that their energy expenditure might outweigh their energy gain.
The routine aerobic metabolic rate of the Magadi tilapia is exceptionally high (Franklin et al., 1995;Narahara et al., 1996. Even when taking the high temperatures into account, Lake Magadi tilapia show metabolic rates that are 25-60% higher than in other tilapia species at similar temperatures (Narahara et al., 1996). Indeed, the cost of merely surviving this harsh environment is high and clearly related to the continuous ionoregulatory and acid-base challenges that these fish face, forcing them to excrete nitrogen as urea. Acclimation to dilutions of Lake Magadi water reduced MO 2 by up to 60%, largely because of reduced acid-base regulatory costs (Wood et al., 2002a) which comprise 50-60% of routine metabolism (Wood et al., 2002a). By some controversial theories, urea production in the ornithine-urea cycle, through its role in HCO 3 − consumption, is linked to acid-base regulation (Meijer et al., 1990;Atkinson, 1992). Regardless, the high metabolic rate may be attributed in part to the fact that urea synthesis is costly (2.5 ATP per urea-N), and Magadi tilapia excrete enormous amounts of urea-N at typical rates of 5000-10 000 μmol urea-N.kg −1 .h −1 . Alcolapia grahami endemic to the nearby South West Hot Springs (SWHS, Figs. 1c and 2b) live at temperatures near 41 • C and show routine metabolic rates of 88 μmol.g −1 .h −1 which are similar to those of a small mammal . These are three to six times higher than the already exceptionally high values of 15-30 μmol.g −1 .h −1 seen in the more commonly studied Magadi tilapia from Fish Springs Lagoon (FSL; Fig. 2a and c) where the temperatures are between 30 and 35 • C (Wood et al., 1994(Wood et al., , 2002aFranklin et al., 1995;Narahara et al., 1996). To support these high metabolic rates, Lake Magadi tilapia are voracious feeders, continuously scraping blue-green algae from the rocks (Coe, 1966, Wood et al., 1994. Fish were even observed to jump repeatedly out of the water to bite pieces of the cyanobacterial math on a stone wall, resulting in a graze line above the water level (Narahara et al., 1996)  This was demonstrated by their stomach content containing 90% algae, with the remaining 10% consisting of dipterous larvae and copepods (Coe, 1966). These algae contain about 60% protein and thus are an energy-and nitrogen-rich food source which allows the fish to sustain their high metabolic rates. It is therefore no surprise that Lake Magadi tilapia deteriorate rapidly in captivity (Wood et al., , 1994 unless precautions are taken to minimize aggression and provide sufficient food (Narahara et al., 1996, Wood et al., 2002a. To date, most published studies therefore utilized fish that were caught only hours before experiments took place. In the only experiment comparing the physiology of fed versus fasted fish, mortality was high and only 20% of the fish survived up to the fourth day (Wood et al., 1994).
In the present study, a new attempt was made to study fasted Lake Magadi tilapia taking the necessary precautions as described by Narahara and co-workers (1996) who advised to keep densities low enough to avoid aggression and to shield the holding tanks from light. To avoid feeding on extraneous sources, extreme caution was taken to keep the tanks devoid of waste and algae, and dead fish were removed immediately. Our aims were to study metabolic rates and utilization of on-board fuels. We hypothesized that the fish would economize energy use and thus lower their routine MO 2 in the absence of food. A second hypothesis was that urea-N effluxes would be reduced when the fish were deprived of the N-rich cyanobacteria. Therefore, urea-N fluxes and plasma urea, as well as gill, gut and kidney urea transporter mRNA expression were assessed, complemented by mRNA expression of ammonia transporters and glutamine synthetase as a proxy for protein metabolism. Overall, we expected fish to be energy-limited and perform less well, and this was tested by measuring swimming capacity and plasma and tissue metabolites. Finally, gill morphology was examined as our earlier work with an Amazonian cichlid (Astronotus ocellatus) showed clear interactions between gill morphology and physiology, where fasted fish had more mitochondria-rich cells with larger crypts, indicating the increased importance of the branchial uptake route when feeding is limited (De Boeck et al., 2013).

Material and methods
All research was performed under a research and ethics clearance permit (NCST/RR1/12/1/MAS/99) from the National Commission for Science, Technology and Innovation (NACOSTI Kenya), and with the permission of the Magadi Soda Foundation. All foreign researchers were licensed by NCST and formally appointed as visiting researchers at the University of Nairobi. Collections were made under permission from the Department of Fisheries, Ministry of Livestock and Fisheries (Kenya).

Experimental animals
All Magadi tilapia, A. grahami, were collected by beach seine from Fish Spring Lagoons (FSL) at Lake Magadi, Kenya (GPS coordinates = 1 • 53 30.2 S, 36 • 18 09.9 E) in July and August 2013. Weights of fish (n = 16) used for the determination of critical swimming speeds were 5.46 ± 0.38 g. Additional fish (n = 56) were used for blood and tissue sampling with an average weight of 6.61 ± 0.23 g. Due to the size of the respirometers, fish used for respirometry (n = 16) were smaller and averaged 2.44 ± 0.15 g. All were transported together with water from the site to the nearby lab set up in a classroom of Magadi Secondary School, where they were kept outdoors in small groups (<30) in 100-L tanks equipped with standard aquarium heaters in well-aerated FSL water (T • : 30-32 • C, pH 9.80, conductivity 2.45 S/cm, salinity 16.6 ppt). Thus, temperature was slightly lower than in the field which showed diurnal temperature variations of 33-36 • C and a pH of 9.75 during that period . Therefore, we might have slightly underestimated metabolic rates. Tanks were shielded from the light by a lid, and 75% of the water was renewed daily using fresh FSL water to ensure good water quality.
For control measurements, all fish were allowed to recover from capture stress for a few hours and were always sampled within 24 h after capture. For the fasting period, fish were maintained as above and care was taken that tanks were clean at all times, any waste was syphoned off three times a day, and any fish that were dead or dying were removed. After the first 24 h, no more mortality occurred, guts were cleared of all food, and waste accumulation was minimal.

Swimming performance and respirometry
Fish were placed in four small-scale (3.9 L) Blazka-style swimming tunnels, 1 h before measurement (one fish per respirometer). Water speed was set at 10 cm/s which allowed the fish to orient themselves against the water current and swim gently. Swimming respirometers were submerged in individual water tanks containing fresh, well-aerated FSL water, and an aquarium heater maintained the temperature at 30-32 • C. A small circulation pump was used to circulate this well-aerated water into the swimming chambers during the trials.
Water velocity was changed in increments of 10 cm/s at 20min intervals, until fish became fatigued. Fish were considered fatigued once they drifted against the rear screen and would not swim for 30 s after the water velocity was temporarily lowered and then returned to the speed at which exhaustion had occurred. U crit was calculated according to the equation given by Brett (1964): where U i is the penultimate swim velocity at exhaustion, U ii is the velocity increment (10 cm/s), T i is the time elapsed at fatigue velocity, and T ii is the interval time (20 min). The absolute values (cm/s) were converted to relative swimming speeds in BL/s by dividing the absolute values by the fish fork length.
As the resolution for oxygen consumption rates (M O2 ) and urea-nitrogen excretion rates ( J urea-N ) during the swimming protocol was not high enough, separate respiration trials were performed in 530-mL amber glass bottles ('Tusker chambers', Wood et al., 1989) fitted with aeration and submerged in a water bath at 30-32 • C. Eight chambers were filled with clean well aerated water after which fish were allowed to settle for 1 h in the chamber with continuous aeration. At the start of oxygen consumption measurements, the aeration was removed, and an initial water sample was collected from each chamber. Oxygen was measured using an electrode (WTW CellOx 325, Germany) set at the appropriate salinity, and subsequently, the chambers were sealed for 1 h which was sufficient to allow oxygen levels to drop to approximately 2 mg.L −1 . At that point, oxygen level was again determined after which aeration was restored. After 5 h, a final water sample was collected to measure urea-N excretion by fish. Urea levels were determined with the diacetyl monoxime method (Price and Harrison, 1987) and multiplied by 2 to obtain urea-N.

Blood and tissue sampling and analysis
For blood and tissue sampling, fish were sedated in FSL water with clove oil (200 mg/L). Blood samples were immediately taken with a heparinized 100-μL Hamilton syringe (1000 U/mL in Cortland saline). In one subset of fish, blood glucose and lactate were immediately determined using an Accu-Check Compact Plus Meter (Roche Diagnostics, Mannheim, Germany) and a Lactate Pro Meter (ARKRAY Inc., Kyoto, Japan). Other studies have validated the use of such point-of-care portable devices for use in fish as reviewed by Stoot and co-workers (2014) (Brown et al. 2008, Serra-Llinares et al. 2012, while the Accu-Check tended to underestimate absolute values but was reliable to compare relative values (Cooke et al., 2008, Wells andPankhurst, 1999). The remaining blood was transferred to a heparinized bullet tube, centrifuged for 5 min at 5500 rpm in (NovaTecnica NT800, Sao Paulo, Brazil) and snap-frozen in liquid N 2 for later determination of plasma urea-N. After cervical section, gill, gut, liver, muscle and kidney were dissected. Liver and muscle were snap-frozen in liquid N 2 and stored in a −20 • C freezer, and gill, gut and kidney transferred to RNAlater (Sigma R0901) and kept at 4 • C for later molecular analyses.
In a second subset of fish, blood was immediately centrifuged and plasma frozen as above for later determination of electrolytes and osmolality. Second and third gill arches of these fish were dissected, quickly rinsed and immediately fixed in cold Karnovsky's fixative (Karnovsky, 1965) and kept cold (4 • C) for morphology. Plasma [Na + ], [Cl − ] and [K + ] were analysed using an AVL 9180 Electrolyte Analyzer (AVL, Roche Diagnostics, Belgium), and osmolality using an Advanced TM Micro Osmometer (Model 3300, Advanced Instruments, USA). On a final set of remaining fish from the 5-day fasted group and some control fish, plasma total CO 2 was measured in duplicate on 50-μL plasma samples using a Corning model 965 CO 2 analyser (Lowell, MA, USA).
Frozen liver and muscle samples were later homogenized in four parts of deionized water using a Sonicator (Sonozap, 25 kHz Battery operated ultrasonic processor). They were analysed spectrophotometrically (Elx800 Universal Microplate Reader, Bio-Tek Instruments Inc.) for protein content by Bradford's method (Bradford, 1976), glycogen content using anthrone reagent (Roe and Dailey, 1966) and lipid content according to Bligh and Dyer (1959). Approximate energy contents of the different energy stores were calculated assuming an energetic content of 17.1 kJ per gram of carbohydrates, 38.9 kJ per gram of lipids and 17.6 kJ per gram of protein (Randall et al., 1997).

Gene expression analysis
After return of the samples (stored in RNAlater) to the University of Antwerp in Antwerp, Belgium, total RNA was isolated from gill, gut and kidney samples using TRIzol (Invitrogen, Merelbeke, Belgium) according to the manufacturer's instructions. The extracted RNA samples were DNAse-treated to avoid genomic DNA contamination. The quantity of the RNA was evaluated using NanoDrop spectrophotometry (Nan-oDrop Technologies, Wilmington, DE, USA). The integrity (quality) was checked by QIAxcel Advanced system (Qiagen, Hilden, Germany) and the purity by measuring the OD260/OD280 absorption ratio (>1.95).
For quantitative real-time PCR (qPCR), a starting amount of 1 μg RNA was transcribed into first-strand cDNA using the RevertAid H minus First-Strand cDNA synthesis Kit (Fermentas, Cambridge, UK). Expression of mRNA in the gills, gut and kidney of Magadi tilapia subjected to fasting was compared with that in control fish by qPCR using the specific primers listed in Table 1. The primer sequences were adopted from Wood et al. (2013). qPCR analyses were performed on an Mx3000P QPCR System (Agilent Technologies, Belgium). Reactions containing 5 μL of 5× diluted cDNA, 10 pmol each of forward and reverse primers, 0.4 μL ROX dye (1:500 dilution) and 10 μL Brilliant II SYBR Green qPCR (Agilent) were performed in a four-step experimental run protocol: a denaturation program (10 min at 95 • C), an amplification and quantification program repeated 40 times (30 s at 95 • C, 50 s at 55 • C, 40 s at 72 • C), a melting curve program (55-95 • C with a heating rate of 0.10 • C/s and a continuous fluorescence measurement) and finally a cooling step. Melt curve analyses of the target genes and reference genes resulted in single products with specific melting temperatures. In addition, 'no-template' controls (i.e. with water sample) for each set of genes were also run to ensure no contamination of reagents or primer-dimer formation. Data were extrapolated from standard curves generated by serial dilution of one randomly selected control sample. β-Actin mRNA expression remained stable across treatments and was used for normalization.

Gill morphology
Gills of a total of 12 fish, 4 controls (fed) and 8 from the experiment (4 fasting for 3 days and 4 fasting for 5 days) were examined by scanning electron microscopy (SEM) and light microscopy (LM). Fixed gills were transported to San Diego State University (SDSU), CA, USA. At SDSU, each gill arch was cut into two equal pieces which contained 10-12 filaments in both anterior and posterior rows for SEM and LM studies. All samples were rinsed in phosphate buffered saline (PBS), postfixed in 1% osmium tetroxide for 1 h and dehydrated in ascending concentrations of ethanol from 30 to 100%. Samples that were visualized using SEM study were critical-point dried, mounted on the stubs, sputter-coated with gold and examined with a Quanta 450 scanning electron microscope (FEI) (Hillsboro, OR, USA) at the accelerating voltage of 3-5 kV. Samples that were examined using LM after dehydration in ethanol were exposed through a graded acetone series to 100% acetone and embedded in EPON epoxy resin (Hexion Inc., Columbus, Ohio). Longitudinal semi-thin sections (1 μm) were made parallel to the long axis of the filaments and were cut with an ultramicrotome (Leica EM microtome, Bannockburn, IL, USA). Then, sections were mounted on glass slides, stained with 0.5% methylene blue and examined in a Nikon Eclipse E200 microscope (Melville, NY, USA) and used for morphometric analysis. The data were based upon 40 measurements per each characteristic (10 randomly selected filaments in 4 fish per condition). The following parameters were measured to estimate the volume of interlamellar cell mass (  (determined by LM) and d-distance between two adjacent lamellae (determined by LM). The volume of ILCM was calculated according to Sollid et al. (2003) as follows: V = dhL.

Statistics
All data have been presented as mean values ± standard error (SEM). The normality of the data was checked using the Kolmogorov-Smirnov test and data log-transformed as needed.
For comparisons among different experimental groups, a oneway analysis of variance (ANOVA) was performed followed by Fisher's least significant difference (LSD) test. Student's unpaired two-tailed t-test was used for single comparisons for the swimming performance and respirometry data. The data were analysed using GraphPad Prism 7.00. A probability level of 0.05 was used for rejection of the null hypothesis.

Results
Magadi tilapia survived well during the fasting period after an initial typical mortality of around 10-20% during the first night after capture. In general, we did not observe any further mortality or any significant reduction in weight, hepatosomatic index or Fulton condition index (results not shown). Figure 3 shows the typically high routine metabolic rates starting at 13.8 ± 0.8 μmol.g −1 .h −1 in control fish, which significantly increased to 19.4 ± 0.5 μmol.g −1 .h −1 when fish were fasted for 5 days. Despite withdrawal from their N-rich food source, urea-N excretion rates (J urea-N ) did not change and remained just below 5 μmol.g −1 .h −1 . As a consequence of the increased metabolic rate, the nitrogen excretion to oxygen uptake ratio (NQ) decreased with fasting. Also, swimming performance, measured as critical swimming speed U crit , tended to increase, albeit not significantly.
Plasma urea-N remained very stable throughout the fasting period (Fig. 4). A non-significant trend of decreasing plasma glucose levels could be seen, but plasma lactate showed a substantial drop from 11 to 4.5 mM which was complete by 3 days of fasting. Also liver and muscle protein levels were reduced by approximately 15% after 5 days of fasting (Fig. 5). While liver glycogen levels remained stable, muscle glycogen decreased by approximately 30%. In contrast, liver lipid concentrations increased after 5 days of fasting, while no changes were seen in muscle lipid levels. When calculating energy content of the tissues (protein: 17.6 kJ.g −1 , carbohydrates: 17.1 kJ.g −1 and lipids: 38.9 kJ.g −1 ), lipid was the main energy source present in liver tissue while protein was the main source in muscle tissue.
Plasma Na + and osmolality quickly decreased by Day 3 of fasting (Fig. 6). Plasma Na + dropped from 195 to about 180 mM by Day 3 of fasting, while osmolality dropped from 447 to 377 mOsm, deviating further from Fish Springs Lagoon water, which had Na + levels of 392 mM and a total osmolality of 513 mOsm during our sampling period . Plasma Cl − and K + remained unchanged. Total CO 2 in plasma was only measured before and after 5 days of fasting but decreased by 66% from 11.9 ± 1.2 to 4.0 ± 0.4 mM, again substantially lower than the 165 mM measured in Lake Magadi water .
Gene expression results of key transporters in the gills show clear changes in the mRNA expression of rhesus glycoproteins (Table 2). Rhbg mRNA expression decreased by about 85% at Day 3 of fasting with some recovery thereafter, while the expression of Rhcg-2 exhibited a similar reduction and remained depressed on both Day 3 and Day 5 of fasting. Also, the urea transporter showed a 35% transient reduction at Day 3 of fasting. No change was seen in Na + /K + -ATPase mRNA expression. Glutamine synthetase showed a 50% reduction in mRNA expression rates at both Days 3 and 5 of fasting. In the gut (Table 2), there were no significant changes in either of the Rh glycoproteins mRNA expression levels, but there was a 75% drop in urea transporter mRNA expression on Day 3 of fasting. The 30% drop in glutamine synthetase mRNA expression was not significant. No significant differences were seen in the expression of any of these genes in the kidney either.
Gross morphology of the gills did not show any changes over the fasting period, and there were no changes in lamellar dimensions and interlamellar cell mass at all (Fig. 7, Table 3). Although there were no differences in appearance and numbers of apical crypts of ionocytes, there was a clear difference in the appearance of the pavement cells. They showed micro-ridges organized into a concentric pattern with a flat and smooth central area in control fish, and this evolved to pavement cells with a highly expanded flat central area  Table 2: Relative expression of mRNA of rhesus glycoproteins Rhbg and Rhcg-2, urea transporter (UT), Na + /K + -ATPase (NKA) and glutamine synthetase (GS) in the gills of Magadi tilapia before and during fasting; values are expressed as means and SEM; asterisks indicate a significant difference between the fasted fish (n = 6-9) and the control fish (n = 7-8) ( * P < 0.05, * * P < 0.01)

Rhbg
Rhcg-2 UT NKA CS  surrounded by few marginal micro-ridges in the fasted fish resulting in a reduced surface area (Fig. 7).

Discussion
Due to the extremely demanding conditions they live in, Lake Magadi tilapia have some of the highest metabolic rates seen in fish. To sustain such high metabolic rates, a continuous energy supply is of vital importance (Wood et al., 1994). Recently, the lake has been subjected to siltation and flash floods due to human activities, which could endanger the growth of the cyanobacteria Arthrospira sp., the main food source of the tilapia. Therefore, our first aim was to determine if A. grahami would adjust their metabolic rates to a lower level when devoid of food. In previous studies at comparable temperatures, Lake Magadi tilapia showed their typical high routine metabolic rates with oxygen consumption rates between 14 and 21 μmol.g −1 .h −1 (Narahara et al., 1996;Wood et al., 1994Wood et al., , 2002aWood et al., , 2016. Our control values were in the lower range of these values, indicating that the fish had sufficiently recovered from capture. Surprisingly, and in contrast to our first hypothesis, no metabolic depression occurred, and instead, routine MO 2 and swimming perfor-mance of A. grahami actually increased with 5 days of fasting indicating that aerobic metabolism was not compromised. In a previous study, attempting to measure the effects of fasting, a similar increase in routine metabolic rate was seen after 4 days (Wood et al., 1994). However, in that study, they could not exclude that the fish had fed on cyanobacteria growing on the walls of the holding tanks and on dead fish, which is why we took extreme care to avoid this by cleaning tanks daily, syphoning off any waste and removing any fish that were dead or dying three times a day. Clearly, the Lake Magadi tilapia do not economize their energy use to compensate for the lack of energy supply, instead maintaining a high metabolism, which could compromise their chances of survival when food becomes scarce.
Different species seem to respond to fasting in different ways. For example, in two closely related cyprinids, the goldfish Carassius auratus and the common carp Cyprinus carpio, goldfish sacrificed feeding metabolism to support swimming with similar MO 2 and U crit between fasted and fed fish (Liew et al., 2012), whereas in common carp, fasting reduced MO 2 by approximately 35% when measured during routine swimming and aerobic exhaustion during a U crit test (Liew et al., 2012;Shrivastava et al., 2017)  fasting juvenile rainbow trout (Oncorhynchus mykiss) and largemouth bass (Micropterus salmoides) showed a 30 and 41% lower MO 2 respectively, compared to the control group (Alsop and Wood, 1997;Gingerich et al., 2010). In contrast, a study on another tropical cichlid, the oscar A. ocellatus, also found a steep increase in oxygen consumption rates with fasting (De Boeck et al., 2013) as was observed in the present study with A. grahami. In the oscars, this was mainly attributed to behavioural differences in the respirometers, with fed fish being much more passive compared to starved fish. However, due to the submerged amber respirometers in the present study, behavioural differences were not assessed. Interestingly, both the oscar and the Lake Magadi tilapia live in environments that are challenging for ion-and acid-base regulations. Oscars live in the Amazon basin, with warm, acidic and extremely soft, ion-poor water which can become hypoxic or even anoxic on the floodplains where these fish breed. In contrast, Lake Magadi tilapia live in even warmer, but extremely alkaline and ion-rich water, which can also become hypoxic and anoxic at night when the blue-green algae respire. Although the two environments are different, they both pose large demands on ion-and acid-base regulations, the cost of which for Lake Magadi tilapia has been estimated at up to 60% (Wood et al., 2002a). Therefore, metabolic depression might not be a valid option for these animals, and they vigorously look for food when deprived access to it.
Although metabolic depression did not occur, switches in fuel source were observed. There was a clear reduction in plasma lactate, due to reduced anaerobic metabolism and/or recycling of lactate to glucose to sustain increased use of carbohydrates. We cannot exclude that some of this reduction in lactate levels was due to attenuation of capture stress rather than cessation of feeding, but the increased use of carbohydrates seems to be supported by the accompanying reduction in muscle glycogen and a trend towards reduced plasma glucose levels. The reduced NQ, which is a function of nitrogen excretion divided by oxygen consumption rates, indicates that nitrogen-containing substrates become less important on a relative basis in the whole energy budget, however not on an absolute basis. In absolute numbers, urea-N excretion remained high at approximately 5 μmol-N.g −1 .h −1 or 2.5 μmol.g −1 .h −1 of urea. With a cost of 5 ATP per urea synthetized in the ornithine-urea cycle (Kirschner, 1993) and an estimated yield of 30 ATP per glucose using 6 O 2 (or a P/O ratio of 2.5) in aerobic respiration, this means that 2.5 μmol.g −1 .h −1 of O 2 is used for urea synthesis or 18-13% of the measured aerobic respiration rates (14-19 μmol.g −1 .h −1 ) of the Lake Magadi tilapia was designated towards urea production. Since nitrogen catabolism as such remained stable during the 5-day fasting period despite the increase in metabolic rate, it was now being fuelled by endogenous protein sources instead of the N-rich cyanobacteria as shown by the reduced protein levels in liver and muscle.
NQ was very high under control conditions and exceeded the already unusually high NQ of 0.20-0.27 from earlier measurements (Wood et al., 1994(Wood et al., , 2002a. With a value of 0.35, it exceeded the maximal theoretical value for 100% protein oxidation via the citric acid cycle by deamination and amino acid oxidation under aerobic circumstances, which was estimated at 0.27 to 0.33 for fish tissue (Kutty, 1972;Kutty and Peer Mohamed, 1975;Kutty, 1978;Lyndon et al., 1992;De Boeck et al., 1995, 2000. However, glucogenic amino acids from protein breakdown such as alanine can also be converted to pyruvate and as such enter gluconeogenesis or be converted to lactate in anaerobic pathways. As indicated by the high plasma lactate levels, the high NQ suggests that  the anaerobic component of metabolism in A. grahami was important. Such high plasma lactate levels have been observed before . However, after 5 days of fasting, when aerobic metabolism increased, lactate levels were reduced, and NQ dropped to 0.25 which is within the aerobic range. It is interesting to note that, in humans, Sola-Penna (2008) has suggested that lactate should not be regarded as simply an anaerobic metabolite but should be considered as a regulatory molecule that modulates the integration of metabolism and is therefore, together with metabolic acidosis, linked to protein catabolism.
At the average routine MO 2 measured in our study (16.6 μmol/g/h), and assuming a linear increase over time, this metabolic rate corresponded to an average energy use of 0.9 kJ per day for a 5 g fish. When looking at the whole organism ( Fig. 8) with the liver contributing about 1% to the total weight (Wood et al., 2002a and our own measurements), and muscle tissue estimated to contribute 50% of the total weight, protein was the most important energy source and contributed about 80% to the total available energy. From our measurements of liver and muscle energy stores, we calculated an energy loss of only 2.1 kJ. Assuming that during the first day, food was still available in the intestinal tract to sustain metabolism, this 'budget' lacks 1.5 kJ of consumed energy which cannot be explained by our data. A possible source could be visceral fat, which we did not assess. Of the 2.1-kJ energy drop that we did observe, protein represented 82%, followed by glycogen with a 13% energy loss and a nonsignificant loss of 5% for lipids. This is unusual, as fasting fish primarily utilize glycogen as an immediate energy source, followed by lipid stores and only then is protein mobilized (Alliot et al., 1984;Echevarría et al., 1997;Pastoureaud, 1991). A previous study on common carp and goldfish using a 7-day fasting period also showed that the former primarily relied on liver and muscle glycogen, while the latter mainly used protein (Liew et al., 2012). In that study, protein stores were also reduced in common carp, although to a lesser extent, and lipid stores remained stable. Also, others have suggested that protein is the main fuel used by fish (Van Den Thillart, 1986;Jobling, 1994).
Lake Magadi tilapia are different in their nitrogen metabolism compared to other freshwater teleosts, being a 100% ureotelic. Plasma urea concentration and urea-N excretion were not affected by the 5-day deprivation of the N-rich food source, a finding that supports rejection of our second hypothesis that a reduction in urea-N excretion would be seen when nitrogen input ceased. Urea transporters were not directly quantified by Western blot or immunohistochemistry but expression of branchial and intestinal urea transporter mRNA was only transiently downregulated at Day 3, and they were back to control levels at Day 5. So nitrogen/protein metabolism remained important and seems to be hardwired, as is also seen in ureotelic elasmobranchs (Kajimura et al., 2008), even though some extra energy was supplied by carbohydrates. Notably,

13
similar hardwiring (continued production of urea-N and no ammonia) was seen when Magadi tilapia were adapted to circumneutral fresh water, where excretion of ammonia would become possible (Wood et al., 2002a). However, the gene expression of the ammonia transporting Rh proteins and glutamine synthetase were reduced in the gills. Despite their ureotelic excretion mechanisms, the high metabolic rates and nitrogen dependence of Lake Magadi tilapia likely produce large amounts of ammonia internally, necessitating cellular ammonia handling. Tissue ammonia levels are comparable to those of other teleosts, and the presence of Rh protein in Magadi tilapia gill tissue has been confirmed both by gene expression studies and by immunohistochemistry . Expression of Rh glycoprotein mRNA responds to ammonia exposure confirming their role in ammonia transport in these fish . Glutamine synthetase is especially involved in N-sequestration and amino acid synthesis rather than protein catabolism, or in detoxifying ammonia. The simultaneous reduction in gene expression in both glutamine synthetase and Rh protein suggests that plasma and/or tissue ammonia levels might have been reduced during fasting. Regrettably, our attempts to measure plasma ammonia failed, and ammonia metabolism in these fish is an area of research that certainly warrants further research. On a final note, Rh glycoproteins have also been suggested to facilitate CO 2 transport (Perry et al., 2010) and in mammals even HCO 3 − transport (Huang and Ye, 2010) which is interesting because the Magadi tilapia must actively excrete HCO 3 − across their gills (Wood et al., 2012), and plasma total CO 2 concentration (or [HCO 3 − ]) was decreased during fasting.
In previous studies, urea-N excretion did not increase with increasing MO 2 when Magadi tilapia were subjected to either exhaustive exercise  or aerobic steadystate exercise . It was suggested that this could have been due to an inhibition of ureagenesis resulting from low HCO 3 − availability during metabolic acidosis (Wood et al., 1994), or to an inhibition of urea synthesis to reduce costs during aerobic exercise . Our data also indicate that these fish depend on a stable urea-N excretion rate. They support the view by Wood et al. (1994Wood et al. ( , 2002a that urea-N excretion is not linked to acid-base regulation and is regulated independently. This was evidenced by plasma total CO 2 concentration (i.e. [HCO 3 − ]) which decreased substantially after fasting independent of the stable urea excretion rates.
Together with the reduced plasma total CO 2 , we observed reductions in plasma Na + and osmolality, but not in plasma Cl − . This is surprising, as net transepithelial electrochemical gradients at the gills drive Na + and HCO 3 − inwards while Cl − experiences a small outward pressure, at least in recently fed fish (Wood et al., 1994(Wood et al., , 2012. Normally, this entry of Na + and basic equivalents occurs either via passive diffusion across the gills or as a result of drinking. Indeed, Lake Magadi tilapia resemble marine fishes in many aspects. They show marine-type ionocytes including accessory cells (Maina, 1990;Laurent et al., 1995;Wood et al., 2002b) and imbibe the surrounding water at rates of about 8 mL.kg −1 .h −1 (Maloiy et al., 1978;Wood et al. 2002b). When they are not feeding, i.e. at night, the stomach is bypassed and contains very little water (Bergman et al., 2003). Absolute differences in drinking rates were not reported in this study, but it is possible that drinking and/or absorption rates were lower in the fish with empty stomachs. This would greatly reduce the ion-and baseload that the fish experienced and could explain the lowered plasma Na + and total CO 2 levels. A simultaneous reduction in Na + and HCO 3 − without change in Cl − is consistent with metabolic acidosis, in accord with strong ion difference theory Stewart (1983), ensuring electroneutrality in the plasma.
Although respiration rates were high and urea-N excretion remained stable, we cannot exclude reduced inwards diffusion of Na + and base at the gills. In control fish, PVCs were ornamented with micro-ridges organized into a concentric pattern with a flat and smooth central area as described before for PVCs located on the filament (Maina, 1990;Laurent et al., 1995;Wood et al., 2002b). In fasted fish, PVCs showed a highly expanded flat central area surrounding by a few marginal micro-ridges. This would reduce the exchange surface area for diffusion of base and Na + into the fish. If this reduction in micro-ridges on the PVC surface would be caused by swelling of the PVC as a consequence of the lowered osmolality in the plasma, diffusion distance would be increased as well, lowering diffusion even further. Ionocytes that are linked to base excretion rather than uptake (Laurent et al., 1995) remained unchanged.
In conclusion, A. grahami responded remarkably well to a 5-day fasting period, which is reassuring from a conservation point of view as it shows that they have at least some resilience towards short periods of food deprivation. In contrast to our original hypothesis, aerobic metabolism and performance went up, and urea-N excretion remained stable. Protein remained the main energy source, now originating from proteolysis in both liver and muscle, which was supplemented by muscle glycogenolysis. In contrast to these shortterm effects, the sustained high metabolism and preference for proteolysis are worrying in case longer periods of low food availability would occur. The apparent inability of Magadi tilapia to lower their metabolic rates and nitrogen metabolism necessitate a continuous energy supply in the long run. At the moment, possible mitigating measures that are being considered to avoid siltation and flash floods include practicing safe farming by constructing terraces, the diversions of rivers, building of check dams in order to ease velocity of water, and the construction of water catchment areas upstream (Ministry of Environment and Forestry, 2015). Even though these efforts are driven by the need to keep the soda crust devoid of impurities so as to allow its continued commercial exploitation, they would also protect the exceptional Lake Magadi tilapia. Other threats however remain, extracting 14 water from the lagoons and aquifers through boreholes will influence the size, and water composition of these peripheral pools as the evaporating water will not be replenished sufficiently. Therefore, a more integrated water management plan is essential.