Elevated Water CO2 Can Prevent Dietary-Induced Osteomalacia in Post-Smolt Atlantic Salmon (Salmo salar, L.)

Expansion of land-based systems in fish farms elevate the content of metabolic carbon dioxide (CO2) in the water. High CO2 is suggested to increase the bone mineral content in Atlantic salmon (Salmo salar, L.). Conversely, low dietary phosphorus (P) halts bone mineralization. This study examines if high CO2 can counteract reduced bone mineralization imposed by low dietary P intake. Atlantic salmon post-seawater transfer (initial weight 207.03 g) were fed diets containing 6.3 g/kg (0.5P), 9.0 g/kg (1P), or 26.8 g/kg (3P) total P for 13 weeks. Atlantic salmon from all dietary P groups were reared in seawater which was not injected with CO2 and contained a regular CO2 level (5 mg/L) or in seawater with injected CO2 thus raising the level to 20 mg/L. Atlantic salmon were analyzed for blood chemistry, bone mineral content, vertebral centra deformities, mechanical properties, bone matrix alterations, expression of bone mineralization, and P metabolism-related genes. High CO2 and high P reduced Atlantic salmon growth and feed intake. High CO2 increased bone mineralization when dietary P was low. Atlantic salmon fed with a low P diet downregulated the fgf23 expression in bone cells indicating an increased renal phosphate reabsorption. The current results suggest that reduced dietary P could be sufficient to maintain bone mineralization under conditions of elevated CO2. This opens up a possibility for lowering the dietary P content under certain farming conditions.


Introduction
The use of increased stocking densities and land-based fish farming systems is inherently associated with the increased accumulation of metabolic carbon dioxide (CO 2 ) in the water [1,2]. The maximum, recommended level of dissolved CO 2 for Atlantic salmon ranges between 8 and 15 mg/L while 20 mg/L is also considered safe by some authors [3][4][5][6][7][8][9][10]. The CO 2 level has been reported to reach up to 40 mg/L on land-based farms [4].
Dissolved CO 2 , carbonate ion (CO 3 2− ), and bicarbonate ion (HCO 3 − ) make up the total dissolved inorganic carbon in water. Hydrolyzed carbon dioxide forms H 2 CO 3 , a weak acid that is promptly dissociated into HCO 3 − and a hydrogen ion. Bicarbonate ions bind with free calcium (Ca 2+ ) to form CaCO 3 in the process of calcification [11] whereas the hydrogen ion bonds with CO 3 2− . Thus, carbonate is the limiting factor to buffer the increase in CO 2 [12]. Following reduction in water pH, the environment becomes undersaturated with CO 3 2− . Reduced water pH leads to an increased solubility of carbonated minerals such as calcium carbonate (CaCO 3 ) [13]. Particulate CaCO 3 dissolves into Ca 2+ and CO 3 2− [14]. The buffering capacity of water is dependent on the availability of CO 3 2− and increases with a higher water alkalinity. As seawater has a higher alkalinity compared with freshwater it is a more efficient pH buffer [15,16].
Teleosts represent the most diverse vertebrate group with the ability to adapt to different and changing environmental conditions [17]. The concentration of O 2 in water is

Experimental Period
Atlantic salmon were sedated (Aqui-S vet.; MSD, Bergen, Norway) and transferred to 450 L tanks with 1 m diameter (week 0) for an acclimatization period of three weeks ( Figure 1). After three weeks of acclimatization, all fish were anaesthetized with MS-222 (Tricaine methane sulphonate; Pharmaq, Oslo, Norway) and graded to achieve an even weight within and between tanks at the start of the trial. The experimental period started at week 3 ( Figure 1). The maximum animal density in seawater was 20.6 kg/m 3 . The animals were kept at 12 • C and a water current velocity of 0.8 L kg −1 min −1 . For the P feeding experiment (weeks 3-16), semi-purified isocaloric and isonitrogenous diets were prepared to contain three different levels of total P: 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P diet (26.8 g/kg) (Skretting Aquaculture Innovation Feed Technology Plant) ( Table 1). The animals were fed three times a day. Dietary P levels were met by adjusting the supplementation of inorganic P in the form of mono-ammonium phosphate (MAP) ( Table 1). Next to mono-sodium phosphate, MAP represents a source of inorganic phosphates with the highest mineral digestibility for salmonids leading to the lowest total P losses [49][50][51]. Commercial diets use predominantly MAP as a source of dietary P, thus it was used in this experiment. The control diet (1P) was formulated to meet the requirement for total P in the early seawater stage of Atlantic salmon (8.8-11.0 g/kg total P) [52]. While the 0.5P diet was formulated without supplementation of MAP, the 3P diet was formulated to contain an excessive level of dietary P comparable to high levels of total P used in Fjelldal et al. (21.5 g/kg) [53] and in Smedley et al. (19.7 g/kg) [54]. This is to understand the effect of high dietary P on bone formation and bone mineralization in Atlantic salmon. Targeted levels of total P were met whilst keeping the remaining feed raw materials constant by  (Table 1). Cellulose is commonly used as a filler material and nutrient substitute in experimental diets with little or no nutritional value thus assuming that cellulose is inert in Atlantic salmon [55,56]. In this experiment up to 9.22% of cellulose (0.5P diet) was used to maintain a constant rate of raw ingredients while increasing the content of MAP. In a study by Kraugerud et al. [57], inclusion of 100 g/kg of cellulose in a diet for Atlantic salmon did not significantly affect macronutrient or P apparent digestibility compared to a fishmeal control diet. Similarly, Hansen and Storebakken [58] did not find any effects of increasing dietary cellulose level ranging from 0 to 150 g/kg on the digestion of main nutrients or fecal excretion of P in rainbow trout. Bromley and Adkins [59] showed that a diet containing up to 300 g/kg dietary cellulose does not have an adverse effect on the animal's nutrient intake and growth rate. To monitor feed intake, excess feed was collected once a day by feed collectors with a mesh size of <2 mm. Carbon dioxide was added to the tanks by ceramic diffusors (Gas-X, Storvik Aqua AS, Sunndalsøra, Norway). The level of CO 2 in 5CO 2 tanks was measured at the beginning and end of the trial (3-5 mg/L) with a portable CO 2 analyzer (OxyGuard, Farum, Denmark). The level of CO 2 in 20CO 2 tanks was measured once a day and averaged 22.61 ± 1.25 mg/L. The temperature and oxygen levels were monitored daily at a depth of 1 m and averaged 12 ± 0.15 • C and 102.35 ± 4.77%. The level of salinity was kept at 34 ppt. The pH measured at the end of the trial (pH meter WTW, pH3110; Sentix 41) was 7.94 in a tank without injected CO 2 and 7.21 in a tank with injected CO 2 .
0.5P diet was formulated without supplementation of MAP, the 3P diet was formulated to contain an excessive level of dietary P comparable to high levels of total P used in Fjelldal et al. (21.5 g/kg) [53] and in Smedley et al. (19.7 g/kg) [54]. This is to understand the effect of high dietary P on bone formation and bone mineralization in Atlantic salmon. Targeted levels of total P were met whilst keeping the remaining feed raw materials constant by use of cellulose as a filler (Table 1). Cellulose is commonly used as a filler material and nutrient substitute in experimental diets with little or no nutritional value thus assuming that cellulose is inert in Atlantic salmon [55,56]. In this experiment up to 9.22% of cellulose (0.5P diet) was used to maintain a constant rate of raw ingredients while increasing the content of MAP. In a study by Kraugerud et al. [57], inclusion of 100 g/kg of cellulose in a diet for Atlantic salmon did not significantly affect macronutrient or P apparent digestibility compared to a fishmeal control diet. Similarly, Hansen and Storebakken [58] did not find any effects of increasing dietary cellulose level ranging from 0 to 150 g/kg on the digestion of main nutrients or fecal excretion of P in rainbow trout. Bromley and Adkins [59] showed that a diet containing up to 300 g/kg dietary cellulose does not have an adverse effect on the animal's nutrient intake and growth rate. To monitor feed intake, excess feed was collected once a day by feed collectors with a mesh size of <2 mm. Carbon dioxide was added to the tanks by ceramic diffusors (Gas-X, Storvik Aqua AS, Sunndalsøra, Norway). The level of CO2 in 5CO2 tanks was measured at the beginning and end of the trial (3-5 mg/L) with a portable CO2 analyzer (OxyGuard, Farum, Denmark). The level of CO2 in 20CO2 tanks was measured once a day and averaged 22.61 ± 1.25 mg/L. The temperature and oxygen levels were monitored daily at a depth of 1 m and averaged 12 ± 0.15 °C and 102.35 ± 4.77%. The level of salinity was kept at 34 ppt. The pH measured at the end of the trial (pH meter WTW, pH3110; Sentix 41) was 7.94 in a tank without injected CO2 and 7.21 in a tank with injected CO2. Figure 1. Experimental set-up and analysis. Animals were acclimatized for three weeks prior to phosphorus (P) feeding. Atlantic salmon were subsequently fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO2 injected (5 mg/L) (5CO2) or with injected CO2 (20 mg/L) (20 CO2) for 13 weeks (weeks 3-16) in seawater. Each treatment was made in duplicates. Analyses, types, and number of samples taken at specified timepoints are indicated.  Experimental set-up and analysis. Animals were acclimatized for three weeks prior to phosphorus (P) feeding. Atlantic salmon were subsequently fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) (5CO 2 ) or with injected CO 2 (20 mg/L) (20 CO 2 ) for 13 weeks (weeks 3-16) in seawater. Each treatment was made in duplicates. Analyses, types, and number of samples taken at specified time-points are indicated.

Sampling and Growth Measurements
The animals' fork length, weight, and condition factor were measured (15 animals/tank, 30 animals/group) (at week 3, 8, and 16 ( Figure 1)). Fulton's condition factor (CF) was calculated according to the following equation: where W is weight (g) and L is fork length (cm) [61]. The specific growth rate (SGR) was calculated as follows: where W week3 and W week16 represent the average animal weight for a given period and days represent the number of feeding days for that period [4]. The feed conversion ratio (FCR) was calculated on a dry matter basis (FCR = total feed consumed/weight gain) [60]. Feed intake (g of feed day −1 ) was recorded daily based on the quantity of recovered feed subtracted from the total daily ration.

Blood Analysis
Blood was collected at weeks 3, 8, and 16 after a 24 h starvation period by puncture of the ventral caudal peduncle in 4 mL lithium heparinized Vacuette tubes (Greiner Bio-One GmbH, Kremsmünster, Austria) and placed on ice. Whole blood samples were analyzed by an iSTAT hand-held Portable Clinical Analyzer with an i-STAT G3+ cartridge (Abaxis Europe GmbH, Griesheim, Germany) on a VetScan iSTAT1 handheld blood analyzer (Abaxis Europe GmbH, Griesheim, Germany) at weeks 3 and 8. The remaining whole blood samples were centrifuged at 4000× g (4 • C) for 7 min with a VWR Mega star 600R centrifuge (Darmstadt, Germany), the plasma supernatant was carefully collected, and stored at −80 • C. Plasma was chemically analyzed on a Konelab 30i chemistry analyzer (Thermo Fisher Scientific, Basel, Switzerland) at weeks 3 and 16 according to the manufacturer's instructions. For the iSTAT analysis, blood was collected from 24 animals at week 3, and from 5 animals/ tank (10 animals/ group) at week 8 ( Figure 1). A volume of 120 µL of whole blood was analyzed for pH, PCO 2 (partial pressure of CO 2 , i.e., the amount of CO 2 dissolved in blood), PO 2 (partial pressure of O 2 ), bicarbonate ions, TCO 2 (total CO 2 ), and sO 2 (O 2 saturation, i.e., the fraction of hemoglobin able to bind to O 2 ).

Blood Plasma Analysis
Samples for blood plasma were collected from 10 animals/tank (20 animals/group) ( Figure 1). The concentration of total and non-bound Ca 2+ , Cl − , inorganic P (Pi), pH, Zn 2+ , ALP, and bicarbonate ions were measured by the Skretting Aquaculture Innovation Laboratory (Stavanger, Norway). The non-bound Ca 2+ , Cl − , and Pi were determined as described in [39]. The total Ca 2+ and pH was measured by ion selective electrode (art. #981772, #981597, respectively). Photometry at 560 nm was applied for quantitative determination of Zn 2+ and the use of 5 Br-PAPS. Photometry at 405 nm was used for quantitative determination of bicarbonate ions and ALP levels (art. #981889, #981833, respectively). Alkaline phosphatase activity (ALP) was determined in three animals/tank (six animals/group) using p-nitrophenyl-phosphate (4-NPP) as a substrate in a 2-amino-2methyl-1-propanol (AMP) phosphate-accepting buffer. The ALP activity was expressed as enzyme units (U/L). The activity of bone/liver/kidney ALP (b/l/k ALP) was determined with the application of 10 mmol levamisole (VWR Chemicals) to inhibit the non-intestinal form of the enzyme [62,63]. The measured activity of intestinal ALP was then deducted from total plasma ALP to determine b/l/k ALP.

Radiography
Animals were filleted on both sides and placed on a 10 MP digital X-ray tablet and radiographed by a portable X-ray unit Gierth TR 90/30 peak (Riesa, Germany). Radiographs were taken at 80 cm distance between the X-ray source and the tablet at 40 kV, 0.2 mA, and 0.2 s exposure time. Radiographs were analyzed as Dicom files using RadiAnt DICOM viewer (Medixant, Poznań, Poland). A sample of 24 animals was X-rayed at week 3 to check for deformities at the start of the trial. At week 16, 15 animals/tank (30 animals/group) were X-rayed ( Figure 1).

Vertebral Centra Deformity Diagnosis
Deformities were analyzed based on digital X-ray images. Vertebral column regions were identified following ( Figure 2A) [64]. Types of vertebral deformities were recorded based on the types described by [65]. The fusion of one or two vertebrae to the base of the skull (occipital region) or one or two vertebrae to the urostyle was not included in the deformity count. These fusions are common, non-pathological phenomena in wild and farmed teleosts [66][67][68][69][70].

Whole Blood Analysis
For the iSTAT analysis, blood was collected from 24 animals at week 3, and from 5 animals/ tank (10 animals/ group) at week 8 ( Figure 1). A volume of 120 µL of whole blood was analyzed for pH, PCO2 (partial pressure of CO2, i.e., the amount of CO2 dissolved in blood), PO2 (partial pressure of O2), bicarbonate ions, TCO2 (total CO2), and sO2 (O2 saturation, i.e., the fraction of hemoglobin able to bind to O2).

Blood Plasma Analysis
Samples for blood plasma were collected from 10 animals/tank (20 animals/group) ( Figure 1). The concentration of total and non-bound Ca 2+ , Cl − , inorganic P (Pi), pH, Zn 2+ , ALP, and bicarbonate ions were measured by the Skretting Aquaculture Innovation Laboratory (Stavanger, Norway). The non-bound Ca 2+ , Cl − , and Pi were determined as described in [39]. The total Ca 2+ and pH was measured by ion selective electrode (art. #981772, #981597, respectively). Photometry at 560 nm was applied for quantitative determination of Zn 2+ and the use of 5 Br-PAPS. Photometry at 405 nm was used for quantitative determination of bicarbonate ions and ALP levels (art. #981889, #981833, respectively). Alkaline phosphatase activity (ALP) was determined in three animals/tank (six animals/group) using p-nitrophenyl-phosphate (4-NPP) as a substrate in a 2-amino-2-methyl-1-propanol (AMP) phosphate-accepting buffer. The ALP activity was expressed as enzyme units (U/L). The activity of bone/liver/kidney ALP (b/l/k ALP) was determined with the application of 10 mmol levamisole (VWR Chemicals) to inhibit the non-intestinal form of the enzyme [62,63]. The measured activity of intestinal ALP was then deducted from total plasma ALP to determine b/l/k ALP.

Radiography
Animals were filleted on both sides and placed on a 10 MP digital X-ray tablet and radiographed by a portable X-ray unit Gierth TR 90/30 peak (Riesa, Germany). Radiographs were taken at 80 cm distance between the X-ray source and the tablet at 40 kV, 0.2 mA, and 0.2 s exposure time. Radiographs were analyzed as Dicom files using RadiAnt DICOM viewer (Medixant, Poznań, Poland). A sample of 24 animals was X-rayed at week 3 to check for deformities at the start of the trial. At week 16, 15 animals/tank (30 animals/group) were X-rayed ( Figure 1).

Vertebral Centra Deformity Diagnosis
Deformities were analyzed based on digital X-ray images. Vertebral column regions were identified following ( Figure 2A) [64]. Types of vertebral deformities were recorded based on the types described by [65]. The fusion of one or two vertebrae to the base of the skull (occipital region) or one or two vertebrae to the urostyle was not included in the deformity count. These fusions are common, non-pathological phenomena in wild and farmed teleosts [66][67][68][69][70].

Vertebral Body and Scale Morphology
For the histological and whole mount staining analyses, vertebral columns were fixed in 10% buffered formalin for at least 72 h, rinsed with tap water three times, and left in tap water overnight while the water exchange was achieved using a dripping tap. Samples were then transferred stepwise to 70% ethanol until further analysis. Scales (~15 scales/animal) were collected from the area between the adipose fin and the lateral line [71] and stored in acetone (VWR, Darmstadt, Germany) at −20 • C.

Whole-Mount Alizarin Red S Staining
Whole-mount Alizarin Red S staining for mineralized bone was performed on the caudal vertebral region (Figure 2A) in three samples/tank (six samples/group) (Figure 1) according to [38]. Vertebral bodies were analyzed for structure and, based on the extent of mineralization, vertebral bodies were characterized as low-mineralized or mineralized.
Scales from five fish/tank (10 fish/group) ( Figure 1) were bleached in 3% H 2 O 2 and 1% KOH for 4 h and stained with 0.5% Alizarin Red S solution for 15 min, rinsed in dH 2 O, and gradually transferred into 100% glycerol through a 1% KOH-glycerol series, as described in [41]. Scale length and length of the non-mineralized part of the scale were measured ( Figure 3A). Observations on the vertebral column and scales were made using oblique illumination on a Zeiss Axio Zoom V16 Fluorescence Stereo Zoom Microscope equipped with a 5MP CCD camera for imaging.

Vertebral Body and Scale Morphology
For the histological and whole mount staining analyses, vertebral columns were fixed in 10% buffered formalin for at least 72 h, rinsed with tap water three times, and left in tap water overnight while the water exchange was achieved using a dripping tap. Samples were then transferred stepwise to 70% ethanol until further analysis. Scales (~15 scales/animal) were collected from the area between the adipose fin and the lateral line [71] and stored in acetone (VWR, Darmstadt, Germany) at −20 °C.

Whole-Mount Alizarin Red S Staining
Whole-mount Alizarin Red S staining for mineralized bone was performed on the caudal vertebral region (Figure 2A) in three samples/tank (six samples/group) ( Figure 1) according to [38]. Vertebral bodies were analyzed for structure and, based on the extent of mineralization, vertebral bodies were characterized as low-mineralized or mineralized.
Scales from five fish/tank (10 fish/group) ( Figure 1) were bleached in 3% H2O2 and 1% KOH for 4 h and stained with 0.5% Alizarin Red S solution for 15 min, rinsed in dH2O, and gradually transferred into 100% glycerol through a 1% KOH-glycerol series, as described in [41]. Scale length and length of the non-mineralized part of the scale were measured ( Figure 3A). Observations on the vertebral column and scales were made using oblique illumination on a Zeiss Axio Zoom V16 Fluorescence Stereo Zoom Microscope equipped with a 5MP CCD camera for imaging.

Histology
Vertebral bodies no. 32-35 and 36-39 were dissected for serial sectioning of previously non-demineralized or demineralized samples (two samples/group) ( Figure 1). For embedding in glycol metacrylate (GMA), samples were treated as described in [38]. Sections were stained with toluidine blue, a metachromatic stain which visualizes skeletal and cartilaginous tissues, cellular structures, and collagen fibers. von Kossa/Van Gieson was used to visualize phosphate in the mineralized bone and type I collagen bone matrix [72]. Sections were cover slipped with DPX (Sigma-Aldrich, Overijse, Belgium) and observed using a Zeiss Axio Imager-Z compound microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) equipped with an Axiocam 503 color camera.

Mineral Analysis of Vertebral Column
Vertebral columns were stained with Alizarin Red S to determine the extent of mineralization prior to bone mineral analysis. Individual Alizarin Red S stained vertebral columns of animals fed the 0.5P and 1P diet were analyzed for bone mineral content to determine if the content differs between animals reared in 5 mg/L CO 2 and 20 mg/L CO 2 . One sample/tank (two samples/group) were analyzed ( Figure 1). Samples were gradually rehydrated stepwise to dH 2 O through glycerol series (3:1, 1:1, 1:3), 1 h per step, analysis samples were then rinsed with distilled water (dH 2 O). Lipids were removed in a 1:1 (v/v) solution of acetone and methanol (2 × 8 h). Samples were dried for 24 h at 105 • C [40]. Samples were processed and analyzed according to [38]. Samples were analyzed by Masterlab Analytical Services (Nutreco, Boxmeer, The Netherlands). Only 0.5P/20 animals characterized as mineralized were analyzed in 0.5P/20 mg/L group. Animals fed 3P were not analyzed as the analysis of X-ray images, whole mount Alizarin Red S stained specimens, and non-demineralized histological section in the current study and in other studies [39][40][41] showed no differences between the mineral content and extent of bone mineralization between Atlantic salmon fed regular and high P in freshwater and seawater.

Vertebral Centrum Compression Tests
The compression tests were carried out at the Institute of Marine Research, Matre Research Station, Matredal, Norway on three vertebrae/ sample in three animals/tank (six samples/group) ( Figure 1). Vertebra no. 33-35 from the transitional region ( Figure 2A), a region with higher stiffness and yield load [48,73], were dissected. Neural and hemal arches and the remains of notochord tissue were removed and vertebrae were stored in saline solution (6 ppt), at 4 • C. Vertebral centra were measured for anterior-posterior length, dorsal-ventral length, and lateral length diameters to the nearest 0.01 mm using a slide caliper. Load deformity testing was measured with a texture analyzer (Model: TA-HD plus Texture Analyzer; Stable Micro Systems Ltd., Surrey, UK) by compressing a single vertebral centrum along its anterior-posterior axis with a steadily advancing piston (0.01 mm/s). The test was terminated at 35% compression. The load deformity data were used to calculate the mechanical properties adjusted for vertebra size/dimensions [74]. The Texture Exponent Software Version 6 (Stable Micro Systems Ltd., Surrey, UK) was used for calculations of values. Size/dimension-adjusted mechanical properties (modulus of elasticity, yield point, failure point, and toughness) were calculated using a modified version of [74], where the yield point was determined by a 0.2% offset line [75], and the failure point was determined by a 5% offset line according to [76].

Gene Expression Analysis
Bone tissue of the vertebral centra (the rectangle in Figure 2B) was collected from the anterior transitional region (below the dorsal fin) (Figure 2A) (one vertebra/animal, five vertebrae/tank, and ten vertebrae/animal group) (Figure 1). Samples were cleaned of muscle, kidney tissue, blood vessels, spinal cord, notochord tissue, and stored in RNAlater, at −80 • C. For RNA extraction, 26-32 mg of vertebral tissue from a single vertebra sample was homogenized in 350 µL of RTL lysis buffer (RNeasy Mini Kit, Qiagen, Hilden, Germany) with 5 mm stainless steel beads in a TissueLyser (Qiagen, Hilden, Germany) (2 cycles of 2 min, 25 Hz). A total of 20 µL of RNA was isolated with the RNAeasy Fibrous Tissue Mini Kit (Qiagen) on a QIAcube workstation. DNase treatment was applied (RNAase free DNase set (250), Qiagen, Hilden, Germany). RNA quality and concentration were determined by a NanoDrop spectrophotometer (Thermo Fisher Scientific Inc, Verona, WI, USA). The RNA was reverse transcribed in a thermocycler (T100 Thermal Cycler; BIO-RAD) with the QuantiTect Reverse Transcription kit (Qiagen, Hilden, Germany) Real-Time PCR was performed with SYBRgreen (PowerUp SYBR Green Master Mix; AppliedBiosystems by Thermo Fisher Scientific, Vilnius, Lithuania) on a Quant Studio 5 thermocycler (Thermo Fisher Scientific) which was programmed to: 50 • C for 2 min, 95 • C for 2 min (with a temperature ramp of 2.74 • C/s), 40 cycles consisting of 95 • C for 1s, and 60 • C for 30 s; this was followed by 95 • C for 1 s, 60 • C for 20s, and 95 • C for 1 s with a temperature ramp of 0.15 • C/s for melt-curve analysis. Samples were run in duplicates. Systematic negative controls containing no cDNA (NTC, no template control) were included. No amplification occurred in NTC. The analysis of the melt-curve showed a single peak for all reactions and an additional peak for several samples analyzed for the expression of enpp1. These PCR products were subsequently analyzed by gel electrophoresis. No additional products and no primer-dimers were detected in these products. The primers used to amplify the transcripts of fgf23, entpd5, and enpp1 were designed by PRIMER3 (NCBI) ( Table 2). Primer pairs used to amplify transcripts of bgp, col1a1a, and opn were obtained from [77] (Table 2). Target specificity was checked in silico using Blast (NCBI). Only primer pairs with no unintended targets were selected. The relative gene expression was calculated by the ∆∆Ct method [78] using β-actin and ef1a as the reference genes [79,80]. Reference genes were selected using RefFinder [81]. To determine the efficiency of each primer pair standard curves were performed in triplicates using a series of 5-fold cDNA dilutions (1:5, 1:25, 1:125, 1:625, 1:3125).
The target genes analyzed in the present study were selected based on their function of the respective proteins in bone mineralization, bone formation, and P metabolism. Type I collagen is the most prominent collagen in teleost skeleton [82]. Based on a study in zebrafish collagen in teleosts consists of three pro-alpha chains α1(I), α3(I), and α2(I) coded by col1a1a, col1a1b, and col1a2, respectively [82]. col1a1a thus represents a specific marker for collagen type I. Osteocalcin/ Bone gla protein (BGP) and osteopontin (OPN) represent abundant non-collagenous proteins in bone. The exact function of BGP is still being discussed. Bgp is synthetized by osteoblasts and can thus be used as a reliable marker for bone matrix formation [83]. OPN is supposed to inhibit bone mineralization and has also been suggested to be required for osteoclastic bone resorption [84,85]. Ectonucleotide pyrophosphatase phosphodiesterase 1 (enpp1) and ectonucleoside triphosphate/diphosphohydrolase 5 (entpd5) are reciprocal regulators of bone mineralization by determining the level of pyrophosphate, the most potent bone mineralization inhibitor [86]. ENPP1 is a positive regulator of pyrophosphate synthesis [87]. ENTPD5 hydrolysis extracellular nucleotide diphosphates into nucleotide monophosphates and inorganic phosphate and hence represents a positive regulator of bone mineralization [86,88]. Fibroblast growth factor 23 (fgf23) is a hormone produced by osteoblasts and osteocytes. Increased levels of FGF23 suppress the renal phosphate reabsorption [89].

Statistical Analyses
All data are given as mean ± standard deviation. Data for gene expression are given as mean ± standard error. The data were arranged by a two factorial arrangement, with two levels of CO 2 (5CO 2 and 20CO 2 ) and three levels of dietary P (0.5P, 1P, and 3P) and analyzed with a general linear model (GLM). Variances among groups were analyzed using one-way ANOVA followed by Tukey's post hoc test (Bonferroni corrected) or Kruskal-Wallis test with pairwise post hoc comparisons (Bonferroni corrected). The Shapiro-Wilk test was used to test the normal distribution of the variances. Values for week 3 were not included in the statistical analysis. The SPSS software was used for the statistical analysis (Version 27, IBM Corp., Armonk, NY, USA). Individual animals were considered as statistical replicates. For SGR, FCR, feed intake, and total feed eaten tanks were considered as statistical replicates and variances among groups were therefore not analyzed. The level of statistical significance was set at p ≤ 0.05.

Diets
The current semi-purified diets were formulated to differ in the content of P. The proximate analysis showed that the targeted total and estimated available P in the diets were reached ( Table 1). Contents of moisture, crude protein, and crude fat were comparable. Ash and total iron (Fe) content in the diets increased gradually with the increased content of MAP (Table 1).

Growth Data
The data on fork length, weight, and condition factor at week 3 are listed in Table 3. Values for fork length, weight, and condition factor at weeks 8 and 16 and data on SGR, FCR, total amount of eaten feed, and feed intake at week 16 with significant explanatory variables for these parameters can be found in Table 3. Total feed intake, total feed eaten, and all growth-related parameters except FCR were reduced in the 3P animal group and in all dietary P animals reared in 20 mg/L CO 2 ( Table 3). The FCR was elevated in animals from all dietary P groups reared in 20 mg/L CO 2 (Table 3). High CO 2 and interaction between high CO 2 and dietary P (0.5P and 3P) resulted in a reduced fork length and weight in week 8, and weight and condition factor in week 16 (Table 3). Table 3. Growth data of Atlantic salmon measured at weeks 3, 8, and 16. Animals were fed diets containing different levels of total P 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L) from week 3 till week 16. The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. For SGR, FCR, feed intake, and total feed eaten, tanks were considered as statistical replicates and variances among groups were therefore not analyzed. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated. X-ray images only show the mineralized parts of the vertebral bodies. Vertebral body endplates can be recognized as typical X-shaped structures with radiolucent spaces between the mineralized vertebral centra (Figure 2A). The radiolucent spaces (R) between the vertebral centra were significantly extended in 0.5P/5CO 2 (Table 4). Vertebral length + R and fork length of animals used for vertebral body measurements was significantly reduced in 0.5P/20CO 2 . Vertebral length was significantly reduced in 0.5P, irrespective of the CO 2 level (Table 4). A low-mineralized phenotype characterized by reduced vertebral length and vertebral height (radiodense area) and extended radiolucent space between the mineralized parts of the vertebral bodies was found in all 0.5P/5CO 2 animals. At week 16, 15 out of 30 animals fed the 0.5P diet and reared in 20 mg/L CO 2 were mineralized, while the remaining 15 animals were low-mineralized ( Figure 4). Further measurement of the low-mineralized and mineralized 0.5P/20CO 2 animals revealed that the extent of R was significantly reduced in the mineralized 0.5P/20CO 2 group compared with the 0.5P/5CO 2 and low-mineralized 0.5P/20CO 2 group (Figure 4). Measurements of the mineralized part of the vertebral centra showed no differences between the 1P and 3P diet groups irrespective of the CO 2 level (Figure 4). Table 4. Vertebral centra measurements. The analysis was performed on X-ray images of the vertebral column in animals fed diets containing different levels of total P 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) which were reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). Fork length of animals used for this analysis and vertebral (V) length + radiolucent space (R) were reduced in 0.5P/20CO 2 animals. The extent of vertebral length was reduced in 0.5P animals irrespective of the CO 2 treatment. Radiolucent space was significantly increased in 0.5P/5CO 2 animals. The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated.

Mineralization Pattern
Alizarin Red S stained specimens and non-decalcified parasagittal histological sections from 0.5P/5CO 2 and from low-mineralized 0.5P/20CO 2 animals demonstrated extended areas of non-mineralized bone matrix at the edges of the anterior part of scales ( Figures 3A and 5), bone trabeculae, vertebral body endplates, and zygapophyses ( Figure 6A-A',D-D',H,I, Table 5). The areas of low-mineralized bone at zygapophyses and vertebral body endplates were smaller in mineralized 0.5P/20CO 2 ( Figure 6E tions from 0.5P/5CO2 and from low-mineralized 0.5P/20CO2 animals demonstrated extended areas of non-mineralized bone matrix at the edges of the anterior part of scales ( Figures 3A and 5), bone trabeculae, vertebral body endplates, and zygapophyses ( Figure  6A-A',D-D', Figure 6H,I, Table 5). The areas of low-mineralized bone at zygapophyses and vertebral body endplates were smaller in mineralized 0.5P/20CO2 ( Figure 6E-E',H,I) but remained significantly larger compared with the animals of the 1P/5CO2, 1P/20CO2, 3P/5CO2, and 3P/20CO2 groups ( Figure 6B-C   and N which include measurement of 0.5P animals reared in 20 mg/L CO2 which showed either low-mineralized or mineralized bone phenotype. While the scale length of 0.5P/20CO2 mineralized animals is comparable to 0.5P/20CO2 low-mineralized (B), 0.5P/20CO2 mineralized animals show a trend to reduce the extension of N in the anterior part of scales (C). Significant differences in measurements are denoted by a different letter.  Table shows bone mineral content analysis. The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO2, diet (D), and their interaction are indicated. There was a trend of an increased ash, Ca, and P (J) in the vertebral column of animals reared in high CO2 and fed increased dietary P content.  Table shows bone mineral content analysis. The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated. There was a trend of an increased ash, Ca, and P (J) in the vertebral column of animals reared in high CO 2 and fed increased dietary P content. Table 5. Extent of non-mineralized bone in zygapophyses. The analysis was performed on whole mount Alizarin Red S stained vertebral columns of animals fed diets containing different levels of total P 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). The extent of non-mineralized bone on both pre-zygapophyses and post-zygapophyses was significantly increased in 0.5P/5CO 2 animals. The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated.

Mineral Analysis
The values for ash, Ca, and P analysis on Alizarin Red S stained samples revealed a gradual increase in bone mineral content from the lowest values in 0.5P/5CO 2 animals, intermediate values in 0.5P/20CO 2 , and the highest values in 1P/5CO 2 and 1P/20CO 2 samples ( Figure 6J). The content of ash in the vertebral column was significantly affected by different dietary P content and by high water CO 2 . The Ca:P ratio remains comparable among all animal groups ( Figure 6J).

Vertebral Centra Deformities
Deformities were not observed in the samples collected at week 3. There were no differences in the vertebral centra deformity prevalence among diet groups, neither was there an effect of elevated CO 2 on the increased rate of deformities at week 16. Mild deformities were found in five 0.5P/5CO 2 , two 1P/5CO 2 , four 3P/5CO 2 , four 0.5P/20CO 2 , four 1P/20CO 2 , and two animals in the 3P/20CO 2 animals. Regions of the vertebral column affected with deformities were the abdominal and transitional regions (Figure 2A for regions) (17 deformed animals), and the caudal region (8 deformed animals). The average number of deformed vertebrae per deformed animal was 3.53 ± 3.1. Deformity types were compression-and fusion-related: compression with fusion, one-sided compression, fusion centra, complete fusions, and homogenous compression. Hyper-dense vertebrae, characterized by an increased radiodensity on X-ray images, were found in five animals and one animal had a vertical shift with deformed intervertebral joints.

Vertebral Body Morphology and Structure
Based on whole mount Alizarin Red S stained specimens and demineralized and nondemineralized serial histological sections, animals showed well-developed vertebral bodies in all diet groups ( Figure 6A'-G' and Figure 7). Osteoblasts were in active spindle-shaped form typical for salmon (black arrows in Figure 7A-C,G-I), collagen type I fiber bundles (white arrows in Figure 7D-F,J-L) and their extension in bone (Sharpey's fibers) were present (arrowheads in Figure 7A-C,G-I). The vertebral centra of animals fed the 0.5P diet developed a characteristic ectopic cartilage at the interface between the non-mineralized vertebral body endplates and the trabecular bone region [40], which were present at the lateral side of the vertebral centra and absent from the midline of vertebral centra (white rectangle in Figure 7M,N).

Blood Analysis
Whole blood analysis (iSTAT) from weeks 3 and 8 can be found in Table 6. The blood pH, PCO 2 , bicarbonate ions, and TCO 2 were significantly increased in animals reared in 20 mg/L CO 2 compared with animals reared in 5 mg/L CO 2 ( Table 6). The value for PCO 2 was significantly increased by the 3P diet in animals reared in high CO 2 compared with 0.5P/20CO 2 animals ( Table 6). Table 6. Whole blood analysis (iSTAT). Listed values for pH, PCO 2 (partial pressure of CO 2 ), bicarbonate ions, TCO 2 (total CO 2 ), PO 2 (partial pressure of O 2 ), and sO 2 (O 2 saturation) of Atlantic salmon at weeks 3 and 8 fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated.

CO 2 5 mg/L 20 mg/L p-Value
Week Blood plasma analysis (Konelab) for ALP, pH, bicarbonate ions, total and non-bound Ca 2+ , Pi, and Zn 2+ from weeks 3 and 16 can be found in Table 7. At week 16, plasma ALP increased with increasing dietary P and high 20 mg/L CO 2 (Table 7). Plasma pH, bicarbonate ions, and total and non-bound Ca 2+ were increased in 20 mg/L CO 2 groups. Plasma Cl − was significantly reduced by 20 mg/L CO 2 (Table 7). Plasma Pi showed a two-fold reduction in 0.5P/5CO 2 and 0.5P/20CO 2 relative to the other groups (Table 7). Plasma Zn 2+ was significantly reduced by the 3P diet and 20 mg/L CO 2 (Table 7). Table 7. Blood plasma analysis (Konelab). Listed values for different plasma alkaline phosphatase (ALP), bone/liver/kidney ALP (b/l/k ALP), pH, bicarbonate ions, total and non-bound Ca 2+ , Pi, Cl − , and Zn 2+ of Atlantic salmon at weeks 3 and 16 fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated.

CO 2 5 mg/L 20 mg/L p-Value
Week 3 Week

Mechanical Properties of Vertebral Centra
The mechanical property analysis of the vertebral centra with associated significant differences between the treatments can be found in Table 8. Values for modulus of elasticity and failure point were significantly reduced by the 0.5P diet. Values for yield point and toughness in animals fed the 0.5P/5CO 2 diet were significantly reduced compared with 3P/5CO 2 and 3P/20CO 2 (Table 8). Table 8. Mechanical properties of vertebral centra. Analyzed mechanical properties were corrected for size of the vertebral centra. Three animals per tank (six/group) were measured at week 16. Animals were fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). The statistically significant differences among groups (within a row) are indicated by different lowercase superscript letters. If none of the letters are the same, differences are significant. The significant values (p) for CO 2 , diet (D), and their interaction are indicated.

Gene Expression
Gene expression analysis showed a significant downregulation of fgf23 expression in the 0.5P animals (Figure 8). A trend of downregulation of fgf23 gene expression was observed in the 0.5P/20CO 2 compared with 0.5P/5CO 2 animals (Figure 8). Expression of the remaining genes associated with bone formation and bone mineralization (bgp, col1a1a, enpp1, entpd5, and opn) remained unaffected by the different levels of dietary P or level of CO 2 (Figure 8).

Discussion
This study on Atlantic salmon analyzed the combined effects of dietary P and levels of dissolved CO2 in water on the mineralization of the vertebral column. The main findings that will be discussed below are: (i) an elevated level of CO2 in the water partly com- Figure 8. Gene expression. Quantitative PCR of gene expression in the vertebral centra of animals fed different levels of total phosphorus (P) 0.5P (6.3 g/kg), 1P (9.0 g/kg), and 3P (26.8 g/kg) and reared in conditions with no CO 2 injected (5 mg/L) or with injected CO 2 (20 mg/L). Graphs show expression of bgp (bone gla protein), col1a1a (collagen type 1 alpha 1a), enpp1 (endonucleotide pyrophosphatase/phosphodiesterase 1), entpd5 (ectonucloside triphosphate disphosphohydrolase 5), fgf23 (fibroblast growth factor 23), and opn (osteopontin). fgf23 expression levels were significantly downregulated in 0.5P/5CO 2 animals compared with the other groups. Results are expressed as mean +/− standard error. Different letters denote significant differences among the groups.

Discussion
This study on Atlantic salmon analyzed the combined effects of dietary P and levels of dissolved CO 2 in water on the mineralization of the vertebral column. The main findings that will be discussed below are: (i) an elevated level of CO 2 in the water partly compensates the effect of a P deficient diet (0.5P), (ii) high dietary P and increased CO 2 level, as single factors and in combination, reduce growth (SGR), (iii) independent from the water CO 2 content, an increased level of dietary P does not improve bone mineralization. In addition, no effects of different dietary P levels on the occurrence of skeletal deformities are observed.

Elevated CO 2 Partly Compensates for Low Dietary P
The increase in bone mineralization is usually considered to be advantageous for Atlantic salmon and thus the dietary P content in commercial feeds is often elevated [53,90,91]. Increased levels of dietary P above the recommended 10.0 g/kg total P (1% total P) have been suggested as enhancing bone mineralization and preventing development of deformities in farmed Atlantic salmon [36,90]. Unexpectedly, in this study the Atlantic salmon fed a 0.5P diet showed comparable bone mineralization with animals fed a 1P and 3P diet when reared in water with 20 mg/L CO 2 levels. In 15 out of 30 animals fed a 0.5P diet the typical low-mineralized phenotype (osteomalacia), characterized in Atlantic salmon and zebrafish by the development of extensive areas of non-mineralized bone [39][40][41][42]92], was completely absent. X-ray images of 0.5P/20CO 2 showed reduced radiolucent areas between the radiodense parts of the vertebral centra. Analysis of whole mount Alizarin Red S specimens of vertebral bodies, scales, and serial histological sections revealed a thin layer of non-mineralized bone in the 0.5P/20CO 2 group somewhat comparable to the thickness of the osteoid in the fully-mineralized 1P and 3P animals. The remaining 15 animals in 0.5P/20CO 2 showed an intermediate level of the osteoid thickness. In contrast, all 30 animals fed the 0.5P diet and reared in 5 mg/L CO 2 had a typical low-mineralized phenotype with extended areas of non-mineralized bone matrix at the edges of the vertebral bodies and scales. Possible explanations for these curious observations are discussed in the following sections and include (i) a systemic response to respiratory acidosis which likely facilitates increased bone mineralization in animals reared in high CO 2 , (ii) enhanced renal phosphate reabsorption under P-deficiency, (iii) and a reduced P requirement due to reduced growth.

Can A Systemic Response to Respiratory Acidosis Help Mineralize A Non-Mineralized Bone?
Hypercapnia leads to reduced systemic pH (respiratory acidosis) and increases the level of plasma PCO 2 , bicarbonate ions, and Ca 2+ , while plasma Cl − is reduced [26,93]. Respiratory acidosis compensation is achieved faster and to a higher degree (94-96%) in water with high levels of Ca 2+ and bicarbonate ions (seawater and hard freshwater) compared with compensation in soft freshwater under low Ca 2+ conditions (83-88%) [27]. Reduced water pH increases the CaCO 3 solubility and increases the availability of free Ca 2+ [13,14]. High ambient bicarbonate ions stimulate an influx of Ca 2+ via Cl − outflux exchange and further aids in pH stabilization [27]. Animals reared in 20 mg/L CO 2 in the current study showed a 2.5-fold increase in bicarbonate ions, and 2-fold increase in PCO 2 and TCO 2 . While the plasma showed a reduced Cl − content, plasma Ca 2+ increased. The systemic response remarkably led to an increased plasma pH in animals reared in 20 mg/L CO 2 compared with animals in 5 mg/L CO 2 water. The pH overcompensation (=high pH) was therefore achieved by the increased influx of Ca 2+ , bicarbonate ions, and outflux of Cl − . Such an increased influx of free Ca 2+ could moreover represent a source of minerals which assisted in the bone mineralization of 0.5P animals reared in 20 mg/L CO 2 . It is possible that the increased Ca 2+ uptake even increased the intestinal absorption of dietary P. Indeed, the extended areas of non-mineralized bone in 0.5P animals are promptly mineralized when Ca and P become available [38][39][40].

Increased Renal Phosphate Reabsorption Driven by the Downregulation of fgf23
Enhanced bone P content can potentially be achieved through increased intestinal absorption from the diet while enhanced renal reabsorption increases phosphate retention. Fibroblast growth factor (fgf23), a hormone synthetized by osteoblasts and osteocytes, suppresses renal phosphate reabsorption [89]. The downregulation of fgf23 in 0.5P animals suggests that the phosphate is retained rather than excreted. Animals fed a 0.5P diet and reared in high CO 2 showed further downregulation of fgf23 which suggests enhanced renal phosphate reabsorption. In rainbow trout (Oncorhynchus mykiss), deficiency in dietary P increased the expression of sodium-dependent phosphate (Na + Pi-IIb) co-transporters responsible for active intestinal phosphate absorption [94,95].

Growth Rate and Mineralization
Bone is first laid down as a non-mineralized matrix and only later is it mineralized [96][97][98]. Sustained swimming induces bone mineralization in Atlantic salmon [99][100][101]. In contrast, dietary P exceeding the animals' requirements has no effect on bone mineralization [39][40][41]. In humans (Homo sapiens), the rapid growth of children in puberty results in the development of under-mineralized bone for a transient period of time [102]. Even if the growth is increased, the rate of mineralization remains the same. Mineralization of the fast forming bone matrix during puberty therefore takes a relatively longer time. It is thus expected that minerals are integrated into the bone matrix at a constant rate under regular growth and mechanical load. This indicates that the reduced growth of 0.5P/20CO 2 animals observed in this study would allow more time for mineral integration into the bone matrix, possibly facilitating bone mineralization even under a P-deficient diet. Comparable to the current study, Gil-Martens et al. [4] observed reduced weight and SGR, and enhanced bone mineral content in animals reared in 35 mg/L CO 2 . Increased bone mineral content in white grouper reared in 28.5 mg/L CO 2 was associated with a reduced fork length in a study by [45]. Fjelldal et al. [48,73] showed that rapid growth correlates with a reduced mineral content and yield-load, and vertebral centra stiffness in fast growing under-yearling post-smolt Atlantic salmon.
However, apart from the reduced growth, animals reared in 20 mg/L CO 2 in the current study also reduced their feed intake. The content of dietary P would thus be equally reduced in animals with a lower appetite. Interestingly, Fivelstad et al. [9] did not record a reduced fork length and weight among Atlantic salmon with increased bone mineral content reared in 16 mg/L or 24 mg/L CO 2 compared with animals reared in 6 mg/L CO 2 . Thus, the reduced growth is less likely to be responsible for the observed increased bone mineralization in 0.5P/20CO 2 Atlantic salmon. The differences in vertebra morphology and mineralization between individuals within the same group of animals, as observed in the current 0.5P/20CO 2 group, have been previously reported in Atlantic salmon reared at presumably 'friendly' water CO 2 levels [43]. In that study, the authors suggested, based on a growth analysis of individuals with different mineralization phenotypes, that reduced appetite in combination with sustained growth in length may cause a low vertebra mineral content in Atlantic salmon post-smolts. In the present study, there is a trend towards an increased mineralization when post-smolts are fed a low P diet and have a reduced appetite due to elevated CO 2 levels. Taken together, these results concerning growth and mineralization further support the notion that reduced growth alone is not the causal factor for the increased mineralization observed in the 0.5P/20CO 2 group. This brings up the importance of the previously discussed increased Ca absorption via gill uptake, intestinal Ca, and P absorption, and enhanced renal phosphate reabsorption which argues in favor of the systemic animal response as the main candidate factor for a regular bone mineralization in 0.5P/20CO 2 animals.

The Negative Effect of High P and CO 2 on Growth of Animals
High P diet and 20 mg/L of CO 2 led to the reduction in feed intake and SGR. Reduced feed intake in animals fed a 3P diet is possibly related to the high inclusion of MAP and a high ash content in the diet. Similarly, Sugiura et al. [103] observed that high P diets with high ash content resulted in a reduced feed intake in rainbow trout. Animals reared in 20 mg/L CO 2 showed a reduced SGR in all diet groups. In studies on European sea bass (Dicentrarchus labrax), gilthead seabream (Sparus aurata), and rainbow trout, high CO 2 resulted in olfactory system impairment and reduced SGR [104][105][106][107]. Comparable to other teleost fishes, Atlantic salmon rely on smell (olfactory system) to detect feed [87,88] which could provide an explanation for the reduced feed intake under high CO 2 in the current study and in other studies [104,105]. In addition, acid-base regulation and stress related to increased plasma CO 2 are metabolically costly [108][109][110][111]. In accordance with the current findings, it was shown that plasma Cl − levels are significantly reduced in Atlantic salmon smolts exposed to 32 mg/L and 21.8 mg/L CO 2 , compared with animals reared in control conditions (1.7-7.0 mg/L CO 2 ) [24,108]. Such a decrease in plasma Cl − is likely associated with the increased bicarbonate levels under high water CO 2 as a consequence of the HCO 3 − /Cl − exchange mechanism [112]. The cumulative effect of impaired smell, the cost of acid-base balance under high CO 2 condition, stress, and the high ash content of the diet are possible factors for the reduced feed intake and SGR in animals fed the 3P diet and in all diet group animals reared in 20 mg/L CO 2 .

High Dietary P Does Not Improve Bone Health
Increased dietary P supplementation in Atlantic salmon diets has been suggested to enhance bone mineralization and prevent vertebral deformities [34,90,91]. Likewise, a diet deficient in dietary P and high water CO 2 are factors that were suggested to induce vertebral column deformities in Atlantic salmon [30,34,54]. Based on X-ray images, and detailed analyses of Alizarin Red S samples and serial histological sections, the occurrence of mild deformities was equally low in all diet groups. A previously hypothesized association between low P diet, high CO 2 , and the development of deformities is thus not supported by this study.
Based on various analytical methods (radiography, mineral content analysis, mechanical property analysis, Alizarin Red S staining, non-demineralized histological sections) this study concludes that the increased level of dietary P does not further enhance bone mineralization when compared to recommended levels of dietary P (1P). The analysis of the 3P diet revealed a higher content of Fe (148 mg/kg) relative to 0.5P (78 mg/kg) and 1P (102 mg/kg) diet, comparable to diets used in Drábiková et al. [39] containing on average 163 mg/kg Fe. The likely cause for this increase is a common presence of Fe impurities in MAP [113]. Concentrations of Fe above 60 mg/kg in feed for Atlantic salmon are suggested to reduce vitamin C absorption while dietary Fe oversupply in rodents and in zebrafish were shown to increase bone resorption and reduce bone mineral content [114][115][116]. No adverse effects of Fe content on bone mineralization or bone resorption were observed in the current study nor in the study by Drábiková et al. [39]. Furthermore, the analysis of the expression levels of genes associated with bone mineralization and bone formation (enpp1, entpd5, opn, col1a1a, and bgp) revealed no differences among the diet groups. The unchanged expression of the bone biomarker genes further confirmed that excessive dietary P had no positive effect on bone formation and bone mineralization. This is in agreement with studies on parr and early-seawater stages of Atlantic salmon where high dietary P intake neither increased bone mineralization nor bone formation [39][40][41].
Under low dietary P conditions in Atlantic salmon and in the zebrafish, formation of the bone matrix is continuous while the bone mineralization is halted [39][40][41][42]. Studies on mineralization of bone in Atlantic salmon with osteomalacia show that the supplementation of P leads to a prompt mineralization of the previously non-mineralized bone matrix [39,40]. This implies that regular bone matrix was formed even under low dietary P conditions and its mineralization was simply delayed by weeks instead of days as is commonly the case [96][97][98]. In line with the previous studies, this study shows a regular bone matrix formation under the 0.5P diet and decoupling of bone formation from bone mineralization. Comparable expressions of gene markers for osteoblasts (bgp) and collagen type I (col1a1a) in all groups suggest that the bone matrix formation continued regularly in all animals irrespective of P intake. Similarly, the expression of gene markers for inhibition of bone mineralization (opn and enpp1) and ENTPD5 where the function is to facilitate bone mineralization were not different indicating that proteins required for a proper bone mineralization are synthetized appropriately even under low P conditions. An upregulation of fgf23 observed in this study as well as in studies by Smedley et al. [54] and Fjelldal et al. [36] indicates increased renal phosphate excretion in animals fed high dietary P. Indeed, several studies provide evidence that a reduced content of dietary P that is within the limits of P requirements, benefits P retention and reduces P waste [117][118][119][120][121][122][123]. A fishmeal-free diet with a reduced content of non-phytin P, supplemented with defluorinated rock phosphate (phosphorus pentoxide), also shows a reduction in solid P waste in juvenile rainbow trout [120]. Rodehutscord et al. [121] determined that rainbow trout fecal P excretion increases gradually with increasing dietary P while non-fecal P excretion increased substantially only after the estimated P requirements are met. In rainbow trout, a high-lipid diet with a lower level of dietary P (315 g lipid, 5.6 g/kg available P) reduces P excretion by nearly 50% and increases P retention by 20% compared with a commercial diet (173 g lipid, 6.9 g/kg available P) [117]. These studies and the current findings demonstrate benefits of reduced dietary P inclusion concerning reduced P effluents from fish farms without compromising the animals' bone health. The current study is especially interesting in the context of emerging aquaculture needs for use of higher stocking densities and recirculating systems, both of which are inherently associated with increased CO 2 release into the water [1,2,6]. Other studies have already elaborated on increased bone mineral content in animals reared in a high CO 2 environment [1,4,9,44,45]. Here it is indicated that the requirement for dietary P in seawater Atlantic salmon is reduced when reared in water with 20 mg/L CO 2 . A study under regular commercial farming conditions could show if the reduction in dietary P in the Atlantic salmon diet is possible for fish reared in recirculating systems or high stocking densities.

Conclusions
This study analyzed the synergistic effect of high water CO 2 (20 mg/L CO 2 ) and low or high dietary P on bone formation and bone mineralization in post-smolt Atlantic salmon. High dietary P (3P) had no further effect on bone mineralization and bone matrix formation in post-smolt Atlantic salmon. High dietary P (3P) and 20 mg/L CO 2 had detrimental effects on SGR, arguably as a result of the decreased feed intake. Animals fed low dietary P (0.5P) and reared in 20 mg/L CO 2 either showed bone mineralization comparable to Atlantic salmon fed the recommended (1P) and high (3P) levels of dietary P. However, some animals from the 0.5P/20CO 2 developed a mild low-mineralized phenotype. This phenotype was, nevertheless, less severe than the phenotype in the Atlantic salmon 0.5P/5CO 2 group. The positive effect of high water CO 2 on bone mineralization coincided with the downregulation of fgf23 expression which suggests that increased renal phosphate reabsorption is responsible for maintaining normal levels of mineralization. The possible reduction in dietary P requirement offers an opportunity to lower the P content in feeds for salmon reared in systems in which CO 2 accumulates. These include land-based systems with high stocking densities and recirculating systems.

Institutional Review Board Statement:
The animal study protocol was approved by the Norwegian Food Safety Authority (Oslo, Norway) (protocol code 25657 and 15.12.2020).

Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the corresponding author.