A comparison of SNARF-1 and skeletal δ 11 B estimates of calcification media pH in tropical coral

Coral skeletal boron geochemistry offers opportunities to probe the pH of the calcification media (pH CM ) of modern and fossil specimens, to estimate past changes in seawater pH and to explore the biomineralisation response to future ocean acidification. In this research we grew 2 Stylophora pistillata coral microcolonies over glass coverslips to allow analysis of the pH sensitive dye SNARF-1, in the extracellular calcification medium at the growing edge of colonies where the first aragonite crystals are formed, under both light and dark conditions. We use secondary ion mass spectrometry (SIMS) to measure the boron isotopic composition ( δ 11 B) of the skeleton close to the growth edge after 2 to 3 days of additional calcification had enlarged the crystals until they joined, generating a continuous sheet of aragonite. Mean skeletal δ 11 B-pH CM estimates are higher than those by SNARF-1 by 0.35 – 0.44 pH units. These differences either reflect real variations in the pH of the calcification media associated with each measurement technique or indicate other changes in the biomineralisation process which influence skeletal δ 11 B. SNARF-1 measures directly the pH of the extracellular calcification medium while skeletal δ 11 B analyses aragonite potentially formed via both extracellular and intracellular biomineralisation pathways. Analysis of a third coral specimen, also growing on a glass slide but with a 5 cm long branch, indicated good agreement between the δ 11 B value of the apex of the branch and the skeletal growth edge. The tissues overlying both these regions were transparent indicating they had low symbiont densities. This suggests that the biomineralisation process is broadly comparable between these sites and that studies growing corals over glass slides/coverslips provide representative data for the colony apex.


Introduction
The δ 11 B of marine carbonates has potential as a proxy of seawater and/or mineralising fluid pH (Branson 2018, Rollion-Bard et al., 2003) and the δ 11 B of tropical coral aragonite is positively correlated with external seawater pH (Hӧnisch et al., 2004;Reynaud et al., 2004).δ 11 B analysis of coral skeletons has been used to suggest that seawaters are acidifying much more rapidly in the Great Barrier Reef compared to NW Pacific reef sites (Shinjo et al. 2013) and that the equatorial/subequatorial Pacific was a more significant source of atmospheric CO 2 during the last glacial/interglacial transition than in the present day (Kubota et al., 2014).In addition, skeletal δ 11 B is used to probe the pH of coral calcification media and to explore the biomineralisation process (Allison et al., 2014) and its response to ocean acidification (Krief et al., 2010, Allison et al. 2018, 2021).
The speciation of the two main boron species in seawater is strongly pH dependent with borate, B(OH) 4 − , predominant at high pH and boric acid, B(OH) 3 , more abundant at low pH (Sanyal et al., 1995).There is a large B isotope fractionation between the 2 species and the δ 11 B of each species is therefore pH dependent (Klochko et al., 2006, Nir et al., 2015).B(OH) 4 − is predominantly incorporated into aragonite (Balan et al, 2016) and the δ 11 B of synthetic aragonite, precipitated from relatively low ionic strength solutions (0.1-0.2 M, Noireaux et al., 2015) approximately follows the δ 11 B -pH relationship predicted from the fractionation factor measured between seawater and inorganic aragonite (Klochko et al., 2006).However, the coral aragonite δ 11 B-seawater pH relationship is offset to higher δ 11 B (i.e.inferring higher pH) than predicted from this fractionation factor (see summary in McCulloch et al, 2018).The growth of coral skeletal crystals is observed in an extracellular space on the apical side of the calicoblastic epithelium (Venn et al., 2011;Tambutté et al., 2012).Recent research suggests that amorphous calcium carbonate (ACC) particles initially form intracellularly, in vesicles in the coral tissue, and are then exocytosed into the extracellular calcifying medium where the coral skeleton forms.Skeletal formation occurs by both ACC particle attachment and ion-by-ion growth (Mass et al., 2017, Sun et al., 2020).Corals increase the pH of the extracellular calcifying medium above that of seawater (Al-Horani et al., 2003;Venn et al., 2011) to promote aragonite formation but the chemistry of the medium in the intracellular vesicles is unknown.Extracellular calcifying medium pH can be estimated by direct observation of the pH sensitive fluorescent dye SNARF at the edge of coral colonies growing on glass coverslips using confocal microscopy (Venn et al., 2011).To date comparisons of SNARF and δ 11 B calcification media pH estimates have been limited.Holcomb et al. (2014) found that SNARF estimates were offset to lower pH by ~0.2 to 0.4 pH units than suggested from δ 11 B analysis of the same colonies.However, in their study SNARF measurements were made at the very outermost growth edge of the skeleton (which had very low symbiont densities) while δ 11 B analyses were performed on much thicker, more mature parts of the skeleton which visibly contained symbiotic dinoflagellates.The authors concluded that real changes in calcification medium pH occurred between these regions of the skeleton (Holcomb et al. 2014).
In this research we compare δ 11 B and SNARF-1 estimates of calcification media pH from comparable sections of coral skeletons, with low densities of algal symbionts.We grew microcolonies of Stylophora pistillata on glass coverslips and used calcein staining to identify the position of the skeletal growth front at the start of the experiment.We used SNARF to estimate the pH in pockets of extracellular calcifying medium (ECM) at the edge of the growing skeleton where the first aragonite crystals are formed.We then used secondary ion mass spectrometry (SIMS) to analyse the δ 11 B of the skeleton deposited in these regions after further calcification had occurred and the skeleton had grown and thickened into a continuous aragonite sheet.We also grew a larger coral colony with a branch ~5 cm high onto a glass slide and compared the δ 11 B of the aragonite deposited at the growth edge (on the glass) with that deposited at the top of the colony to determine if δ 11 B varied between the growth edge and colony apex.Throughout the manuscript we use the abbreviation pH CM to denote pH estimates of the coral calcification media.These may be of the ECM (i.e.pH estimates made by SNARF-1) or potentially of both the extracellular and intracellular calcification media (defined CM, i.e. estimates made by δ 11 B).

Coral culturing
Microcolonies of the model scleractinian coral species Stylophora pistillata, were maintained in aquaria at the Centre Scientifique de Monaco.Coral fragments were attached to glass coverslips or glass slides with resin (Devcon™) and left to grow laterally across the glass surface (Muscatine et al.1997;Raz-Bahat et al., 2006;Venn et al. 2011).Aquaria were supplied with flowing seawater from the Mediterranean Sea (exchange rate 2% h − 1 ), pH ~ 8, salinity 38, with temperature maintained at 25 • C, under an irradiance of 200 µmol photons m − 2 s − 1 on a 12 h:12 h light:dark cycle.Aquarium seawater pH was monitored continuously by pH probes (Ponsel-Mesure, France) linked to a custom-made monitoring system (Enoleo, Monaco) and verified by m-cresol purple measurements (Dickson et al. 2007).Seawater pH in the aquaria varied by ≤0.02 units during the experiment.Total alkalinity was 2570 µmol kg − 1 , determined from a single measurement of seawater supplied to the aquaria during the experiment.Food was provided daily as frozen rotifers and twice weekly with live Artemia salina nauplii.Microcolonies were left to grow for 3 to 4 weeks under these conditions before the experiment.

Samples and timing of SNARF-1 and δ 11 B analysis
A total of three different microcolonies were analysed in this investigation.Corals 1 and 2 were grown on glass coverslips and consisted of flat, laterally growing fragments with no branch (Fig. 1a).Coral 3 was grown on a glass slide and consisted of both flat laterally grown material and a vertically orientated branch about 5 cm high (as in Fig. S1).To record the skeletal growth front at the start of the investigation, the microcolonies were transferred to glass dishes and incubated in a seawater solution of the fluorescent dye calcein (Sigma) (160 µM) for 30 min and then rinsed for a further 20 min in seawater before being returned to the aquaria (Tambutté et al. 2012;Venn et al. 2013).This calcein incubation produced a clear staining band that was visible in the skeleton using fluorescence microscopy.Skeletal growth beyond this band could therefore be identified as a zone of new growth and could be targeted for SNARF-1 and δ 11 B measurements.
SNARF-pH CM of corals 1 and 2 was analysed under light and dark conditions, one and two days after the calcein staining respectively.Coral 3 was grown at the same time and in the same conditions as corals 1 and 2 but was not analysed using SNARF-1.All corals were sacrificed 4 days after the calcein staining by submergence in 10% sodium hypochlorite and the skeleton was rinsed in distilled water, oven dried and shipped to the UK for skeletal δ 11 B analysis.

pH measurements by confocal microscopy and SNARF-1
In vivo pH CM measurements were made at the Centre Scientifique de Monaco by inverted confocal microscopy (Leica SP8) using the pH sensitive dye SNARF-1 (Thermofisher Scientific) according to methods reported previously (Venn et al., 2019(Venn et al., , 2022) ) in the area of new skeletal growth after the calcein stain.For the measurements, the microcolonies were transferred from the aquaria and fitted in cell-culture perfusion chambers (POC-R2, PeCon) mounted on a temperature-controlled stageinsert on the confocal microscope.Irradiance (for in the light measurements) was provided at 200 μmol photons m − 2 ⋅s − 1 and temperature maintained at 25 • C with perfusion rates at a 50% per minute renewal rate of the volume of liquid in the chamber.For both light and dark measurements, microcolonies were first perfused with seawater for 20 min, before being perfused with a seawater solution of 45 µM cellimpermeable SNARF-1 for a 5 min loading period.Perfusion continued for a further 10 min during which repeated pH measurements were taken by obtaining confocal Z stacks of optical sections that started at the level of the glass coverslip and moved upwards into the coral tissue or seawater.Light and dark measurements were carried out consecutively on the same samples before returning them to the aquarium.Magnification was at 40x, with excitation provided at 552 nm and fluorescence captured in two channels at emission wavelengths of 585 ± 10 nm and 640 ± 10 nm.For SNARF-pH CM measurements, Z stacks were recorded at the growing edge of each microcolony in the zone of new growth between isolated crystals (Fig. 2).For seawater pH measurements, Z stacks were captured in the seawater/SNARF-1 solution surrounding the corals within 100 µm (=boundary layer, BL) and at 5 mm (=seawater, SW) from the edge of the microcolony.
For each microcolony, Z stacks were obtained in 3 or 4 different positions of the growing edge.During image analysis, digital regions of interest (ROI) were drawn in several areas of the ECM in each position or Z stack (Figs. 1 and 2).The ratio of mean SNARF-1 fluorescence at 585 ± 10 nm and 640 ± 10 nm was calculated in each ROI.This ratio was converted to pH with a calibration curve according to the procedures described in (Venn et al. 2011).Repeat SNARF-1 measurements in a seawater (pH total = 8.05) have a standard deviation (1 s) = 0.02 pH units.After SNARF-1 measurements, microcolonies were removed from perfusion chambers rinsed in seawater for 20 min and returned to the aquaria.

Boron isotopes 2.4.1. Preparation of samples for SIMS
Cleaned skeletons were photographed (Fig. 3a) and the calcein stain in the skeleton was recorded using a Leica SP8 laser confocal microscope at the University of St. Andrews (Fig. 3b).Corals 1 and 2 were prepared for SIMS by positioning a 2.5 cm diameter mould over the cleaned skeleton and filling the mould with epoxy resin (Struers Epofix).The glass coverslip at the bottom of the coral served as the base for the mould.After the epoxy hardened the mount was polished to partially remove the glass coverslip whilst leaving the underlying coral aragonite at the skeleton edge exposed for analysis (Fig. 3c).Samples were polished using silicon carbide papers (up to 2400 grade, lubricated with water) and diamond suspensions (3 and 0.25 µm).
Samples were prepared from the growth edge and branch apex of coral 3. The skeletal growth tip was removed using a hand held rotary saw, placed in a 2.5 cm diameter circular mould, embedded in epoxy resin and polished to produce a cross-section across the outermost tips of the coral skeleton.The skeletal growth edge (growing laterally over the glass slide) of this colony was too large to embed within a single 2.5 cm diameter mount.Epoxy resin was dabbed over the edge of the coral (giving support to this fragile surface) and, after drying, the glass slide was sawn into smaller pieces which were individually embedded in epoxy resin and then polished to remove the glass slide as before.All mounts were photographed under reflected light to produce maps to identify locations for SIMS analyses.

SIMS measurements
Skeletal δ 11 B was determined by SIMS using a Cameca 7f in the School of GeoSciences at the University of Edinburgh.Sections were gold coated and analysed with a primary O 2 -beam of ~9 nA, accelerated at 5 keV and focussed to an oval ~30 × 40 μm.Instrument conditions were energy offset = 0 eV, imaged field = 25 µm, entrance slits 150 μm and exit slits 535 μm (mass resolution was ~2400).Secondary ions were collected by a single electron multiplier cycling the magnetic field through the mass range.Singly charged cations were collected at masses 10 B (11 s per cycle) and 11 B (3 s) yielding typical count rates of ~2500 and 9000 cps respectively.Count rates were corrected for electron multiplier deadtime (24.8 ns).Doubly charged 40 Ca 2+ ions were collected at mass 20, to track the sputtering of Ca from the sample, and typical count rates were ~110000 cps.Background counts on the electron multiplier were <0.02 cps and are considered insignificant.Each analysis is the sum of 60 cycles.A pre-analysis sputter time of 60 s in spot mode was used to remove surface contamination.Internal repeatability (the precision at a single point) was calculated from the standard deviation of the measured population (1 s) of the 60 cycles in each analysis (s/(√60)) and was typically 1.1‰.The standard deviation of multiple δ 11 B analyses (n = 28-47) on each coral sample was typically 1.6-2.6‰.A Desmophyllum sp.cold water coral chip which exhibited limited heterogeneity in δ 11 B (Allison et al., 2021) was analysed as a reference material under the same conditions to confirm that there was no instrumental drift within and between days.Multiple analyses were completed on this reference material each day (n = 9-11) to yield a 95% confidence limit of the mean (calculated from the standard deviation, s, as t multiplied by s/√n, where t is the test statistic at n-1) which was typically ~1.0‰ and was always better than ±1.1‰.δ 11 B was measured on the edges of 2 microcolonies imaged with SNARF (Corals 1 and 2) and on the apex and edge of one larger colony (Coral 3).We identified several features in the coral section that could be correlated between the calcein maps and the polished mount micrographs (Fig. 4a and b).These features include outgrowths at the edge of the coral that have unique morphology, voids in the section and dark areas which are likely to be centres of calcification i.e. the start points of aragonite deposition.These features can be used to accurately correlate the calcein maps and micrographs and thereby identify locations of suitable areas for SIMS analysis.Multiple SIMS analyses were made across each mount.In corals 1 and 2 (both calcein stained) analyses were focused on sections of the skeleton which were either deposited in the outermost edge of the skeleton, beyond the calcein stain or were deposited after the calcein staining but were enclosed by skeleton which had been stained (Fig. 4).45 and 74 δ 11 B analyses were conducted on corals 1 and 2 respectively and these spanned a broader area than analysed by SNARF-1.In coral 3 analyses were made at the very edge of the skeleton which had grown over the glass slide (comparable to the analyses made in corals 1 and 2) and on the outermost tips of the apex skeleton.
Each SIMS analysis is the sum of 60 cycles and count rates of secondary ions typically exhibit a small decrease (15-20%) over this period (see the Desmophyllum reference material in Fig. 5).We observed these characteristic profiles in most analyses on the coral samples (Sample 1, Fig. 5) but some analyses on the edge samples showed large variations in secondary ion counts (Sample 2, Fig. 5).These discrepancies may indicate that the primary beam has passed through the depth of the coral sample to the underlying epoxy resin.The B and Ca counts of epoxy resin are ~3% and <0.01%respectively of the counts on coral aragonite analysed by SIMS (Allison et al., 2007) However sputtering through the last CaCO 3 can increase Ca counts as sputtering of the remaining CaCO 3 accelerates.Analyses which demonstrated these anomalies in secondary ion counts after 30 cycles were limited to the first 30 cycles of data.We calculated 10 B/ 11 B for each analysis on the Desmophyllum reference material, first using the full 60 cycle dataset and then limiting the data to the first 1-30 cycles.There was no significant difference in 10 B/ 11 B (1-30 cycles = 0.27009, 1-60 cycles = 0.27005, paired t test p = 0.80, n = 33).
Accurate calibration of SIMS analyses is hampered by the heterogeneity of carbonate standards used for SIMS.The Desmophyllum reference material used here is relatively homogenous (1 s of multiple δ 11 B by SIMS is ~1.4‰) but is too small to microdrill to allow accurate characterisation by bulk methods whilst still leaving material for SIMS analyses.To calibrate the δ 11 B of the SIMS samples we identified a Porites lutea coral skeleton which had been cultured at constant temperature, salinity and seawater pCO 2 and exhibited little δ 11 B heterogeneity during previous analyses (reference P. lutea 1 cultured at 750 µatm seawater pCO 2 , 1 s of multiple δ 11 B by SIMS = 0.8‰, Allison et al., 2018).We used this coral skeleton as a reference material.We completed multiple SIMS analyses over a short section (200 µm) of this coral skeleton in the same session as the S. pistillata microcolonies to yield mean 10 B/ 11 B = 0.27873 ± 0.00030, 1 s, n = 9.We then drilled a powder from an adjacent transect of the same coral skeleton and analysed this by multicollector inductively coupled plasma mass spectrometry (MC-ICP-MS, Section 2.4.3) in 2 separate sessions yielding δ 11 B = -17.43± 0.09‰ (2 s, n = 6).This negative value occurs as the P. lutea coral was cultured in artificial seawater made with boric acid with a significantly different δ 11 B to that of natural seawater (Allison et al., 2021).We observed good agreement between SIMS analyses of an additional coral standard, M93-TB-FC-1, calibrated using this method (δ 11 B = 25.1‰) and values determined by MC-ICP-MS and thermal ionisation mass spectrometry (δ 11 B = 24.6-25.0‰)(Kasemann et al., 2009).We did not have sufficient material from M93-TB-FC-1 or P. lutea 1 to permit SIMS analyses of these reference materials each day.
pH CM was estimated from skeletal δ 11 B as: .
using α B = 1.0266 (the mean of two empirical estimates which agree within error i.e 1.0272 ± 0.0006 Klochko et al., 2006 and 1.0260 ± 0.0010 Nir et al., 2015), pK B calculated from the temperature and salinity of the culture seawater (25 • C and 38 respectively) and assuming that the δ 11 B of the calcification media (δ 11 B CM ) is the same as the culture seawater (Allison et al., 2014).This calculation assumes that calcification media δ 11 B is the same as seawater and that B(OH) 4 -only is incorporated into the aragonite lattice.Thus skeletal δ 11 B reflects δ 11 B of B(OH) 4 -in the calcification media and can be used to derive media pH.

MC-ICP-MS
The culture seawater was sampled at the start, in the middle and at the end of the experiment.Samples were filtered using 0.2 µm polyethersulfone filters, acidified to pH 2 and stored in acid washed high density polyethylene bottles.Boron isotope ratios of the culture seawater and the P. lutea standard were analyzed in the Isotope Geochemistry Laboratory at the MARUM (Center for Marine Environmental Sciences, University of Bremen).The coral powder was dissolved in 1 M HNO 3 .Boron was separated from seawater and coral solutions by microsublimation as described in Hüpers et al. (2016).Boron isotope ratios were measured using a ThermoScientific Neptune Plus MC-ICP-MS with the standard-sample-bracketing method in low resolution at concentrations of 100 ng/g B (detailed in Wilckens et al. 2018).Boron isotope values are reported in the conventional δ 11 B (‰) notation relative to NIST SRM 951, with the uncertainty of measurements reported as 2 s and based on a minimum of three isotope measurements.

Results
All SNARF-1, δ 11 B and MC-ICP-MS data are included in the supplementary data.

Confocal-SNARF-1 analysis
Confocal-SNARF-1 analyses were carried out at the growing edges of samples of S. pistillata (Figs. 1 and 2).The growing edge is characterised by lateral extension of the coral skeleton where the overlying tissue produces isolated, rapidly growing CaCO 3 crystals on the glass coverslip that eventually fuse to form a continuous sheet of skeleton (Fig. 1).The tissue was relatively transparent in this zone due to low densities of symbiotic dinoflagellates relative to the centre of the sample (Fig. 1).Confocal analysis combined with light microscopy allowed visualisation of the ECM and calicoblastic epithelium surrounding the isolated crystals (Fig. 2).The calcein stain that marked the beginning of the investigation was also visualized by confocal imaging (Fig. 2).pH was determined from analysis of SNARF-1 fluorescence in ROIs drawn in the ECM between isolated crystals that had grown after the calcein staining.This analysis was made at three to four different locations along the growing edge of each sample.Mean estimates of pH CM were calculated from all the ROIs measured at each location in each sample and are presented in Table 1.No significant differences were found between pH CM in light and darkness, or between corals (one way ANOVA, p = 0.18).

δ 11 B analyses
δ 11 B analyses were sited on areas which were deposited at the outermost edge of the skeleton, beyond the calcein stain or in spaces within the calcein stain zone.All skeletal δ 11 B analyses are illustrated in Fig. 6.To test if the distance from the growth edge influences δ 11 B we divided the dataset for Coral 2 into SIMS points located in the outermost 1 mm of skeleton and analyses sited further back in the skeleton.We observed no significant differences in δ 11 B between these regions (Table 2).As the coral extends over the glass coverslip the skeleton becomes thickened to produce coenosteal spines, up to several mm in length, which radiate from the central part of the colony to the outer growth edge (Fig. 1b).The spines were apparent in the SIMS mounts of the growth edges of the microcolonies while the thinner skeleton deposited between spines was often lost during polishing (Fig. 4d).To explore if δ 11 B varied between these features we divided the analyses of corals 1 and 2 into those sited on spines and those sited in between spines.The data for analyses on spines in each of corals 1 and 2 and for analyses in between spines in coral 2 were normally distributed (Shapiro Wilk test, p ≥ 0.05).The dataset for in between spine analyses in coral 1 was too small (n = 7) to test for normality but we assume this data is also normally distributed and use the t test to compare δ 11 B between features  N. Allison et al. within each coral (Table 2).In coral 1 δ 11 B is not significantly different between these 2 regions but in coral 2 δ 11 B is significantly reduced in the skeleton between spines.We observed no significant difference in the δ 11 B of the colony branch apex and growing edge in coral 3 (Table 2).
Seawater δ 11 B in the culture system ranged from 39.08‰ ± 0.06 (2 s, n = 3) at the start to δ 11 B = 39.71‰ ± 0.05 (2 s, n = 3) at the end.It is unclear when the analysed skeleton was deposited during the experiment and the mean seawater δ 11 B over all samples (start, middle and end = 39.3‰ ± 0.3, 1 s, n = 9) was used to estimate pH CM from skeletal δ 11 B. A seawater δ 11 B change of the observed magnitude (0.6‰) affects reconstructed pH CM by 0.041 which is small compared to the variation observed between SIMS analyses (Fig. 6).

Comparing SNARF-1 and δ 11 B pH CM
Seawater pH, coral tissue boundary layer pH and pH CM estimated from SNARF-1 and δ 11 B are summarised in Table 1 and illustrated in Fig. 7.We combined all light and dark SNARF pH estimates taken from different ROIs at the growing edge into a single dataset for each coral and used the t test to compare the means of pH estimates by SNARF and δ 11 B (both normally distributed, Shapiro Wilk test, p ≥ 0.05).pH estimates were significantly different between methods for both coral 1 (p = 8.2 × 10 − 16 ) and coral 2 (p = 9.7 × 10 − 14 ).

Comparing SNARF-1 and δ 11 B pH CM estimates
Mean pH CM estimates by δ 11 B are higher than by SNARF-1, in both the light and the dark (Fig. 7).The offset between the average SNARF-1 and δ 11 B pH CM estimates is substantial: 0.44 and 0.35 pH units in corals 1 and 2 respectively.Coral calcification is typically faster in the light than the dark (Gattuso et al., 1999) and if δ 11 B measurements are biased towards light pH CM conditions then the offset between SNARF-1 and δ 11 B pH CM estimates is larger than suggested by this comparison.There is excellent agreement between SNARF-1 estimates of seawater pH in the vicinity of the coral and measurements of culture seawater by pH electrode and spectrophotometric methods, suggesting that SNARF-1 yields accurate estimates of media pH.The typical uncertainty in SNARF measurements is ≤0.18 pH units.δ 11 B pH CM estimates included the uncertainty on α B (i.e.±0.0006 from 2 estimates by Klochko et al. (2006) and Nir et al. (2015), equivalent to ±0.02 pH units), the uncertainty in seawater δ 11 B (equivalent to ±0.02 pH units, see Section 3.2) and the uncertainty in multiple coral δ 11 B measurements (equivalent to ±≤0.06 pH units, Table 1).Combining these yields a compounded uncertainty of <0.07 pH units.We did not detect variations in temperature or salinity in the culture system but note that small changes have little effect on pK B , equivalent to ±0.005 pH units for ±1 salinity unit and ±0.006 pH units for ±0.5 • C. We conclude that there is a genuine offset between mean SNARF-1 and δ 11 B pH CM estimates.This could reflect real variations in the pH CM associated with SNARF-1 and δ 11 B measurements or changes in the biomineralisation process.

Table 1
Mean pH estimates (total scale) by different methods in each coral colony.SNARF-1 was used to estimate the pH of the seawater (SW, at 5 mm from the tissue edge), the fluid boundary layer of the coral tissue (BL, within 100 µm of the tissue edge) and pH CM in both the light and dark.δ 11 B analyses likely combine skeleton deposited in both the light and dark.na = not analysed.Means and standard deviations are included for each technique, Numbers of analyses (n) show total ROI analysed by SNARF and total SIMS points.The boron proxy assumes precipitation occurs from a media with the δ 11 B of seawater and proceeds via incorporation of borate into the aragonite lattice.The cell impermeable dye calcein is rapidly transported to the coral extracellular calcification site when added to seawater suggesting that certain ions and molecules in seawater reach the extracellular calcification site by paracellular transport (Tambutté et al., 2012).The Ca 2+ , pH and dissolved inorganic carbon are modified by active and passive transcellular transport mechanisms (Allemand et al., 2011, Zoccola et al., 2015).The transport of a specific boron species into the ECM would offset media δ 11 B. Preferential transport of B (OH) 3 has been observed across frog cell membranes (Dordas and Brown, 2001) and has been hypothesised in coral (Fietzke and Wall, 2022).B(OH) 3 is enriched in 11 B compared to B(OH) 4 − , and preferential transport into the calcification site would elevate the δ 11 B of aragonite and offset δ 11 B pH CM to higher values.Incorporation of B(OH) 3 into the skeleton could also drive δ 11 B pH CM to higher values although we note that theoretical calculations (Balan et al., 2016) and studies of B structural state (Sen et al., 1994;Klochko et al., 2009;Noireaux et al. 2015) indicate that B(OH) 4 -is the ion predominantly incorporated into the aragonite structure.
Recent research suggests that the coral biomineralisation proceeds via two crystallisation pathways (see Fig. 6 in Sun et al., 2020 for a schematic).The process begins via intracellular formation of amorphous calcium carbonate (ACC) precursors (Mass et al 2017, Sun et al., 2020) in vesicles in the calicoblastic epithelium.The ACC vesicles likely form from endocytosis of the ECM which is manipulated intracellularly to form ACC and then released from the apical side of the calicoblastic epithelium back into the ECM.Formation of the aragonite skeleton occurs in the ECM through attachment of ACC nanoparticles and space filling by ion-by-ion growth of aqueous CO 3 2-and Ca 2+ (Sun et al 2020).
Offsets between SNARF-1 and δ 11 B pH CM estimates may reflect the contributions of ACC particle attachment and ion-by-ion growth to each analysis type.SNARF-1 measurements record the pH of the extracellular calcification media, while SIMS analyses the δ 11 B of the final skeleton which could be a composite of both the intracellular formation of ACC and extracellular growth of the skeleton.The SNARF-1 and δ 11 B pH CM offset could reflect real pH changes between the biomineralisation pathways or the influence of a B transport process as discussed above.It's also unclear how aragonite formation via an ACC precursor influences aragonite δ 11 B (Branson 2018).The B/Ca of ACCs precipitated in vitro are positively related to seawater pH suggesting that B(OH) 4 − , which becomes more abundant at high pH, is preferentially incorporated into ACC over B(OH) 3 (Evans et al., 2020).However, B/Ca of ACCs are considerably higher (Evans et al., 2020) than for synthetic aragonites produced at comparable pH (Holcomb et al., 2016) indicating differences in B incorporation between ACCs and aragonites.Finally, it is unknown how conversion of ACC to aragonite will influence B incorporation and δ 11 B. ACC dehydration and conversion in corals occurs during transit of the ACC through the extracellular calcification media or after attachment to the existing skeleton (Sun et al., 2020).Conversion of ACC to CaCO 3 may occur by solid phase transformation (Nielsen et al., 2014) or via dissolution and reprecipitation (Walker et al., 2017).This latter process allows for the exchange of ions between the ACC and surrounding media and this could alter the δ 11 B signature of the formed aragonite from that of the precursor ACC.If conversion occurs in a confined space then the opportunity for this exchange may be limited (Branson 2018).However δ 11 B pH CM estimates in this study are ~0.6 pH units higher than of the culture seawater (Table 1) and we note that δ 11 B pH estimates from aragonites precipitated extracellularly in S. pistillata cell cultures maintained at pH ~8 are also ~0.6 pH units higher than that of the culture media (Drake et al., 2018).Confinement did not occur in the extracellular aragonite precipitation in the cell cultures providing preliminary evidence that this factor does not influence aragonite δ 11 B.
Rather the pH increase in the Drake et al., (2018) study was interpreted to reflect an influence of the coral cells on the media immediately adjacent to the cell.The standard deviation of the δ 11 B analyses within corals 1 and 2 is up to 2.5‰, equivalent to a pH change of 0.17.This is in good agreement with the observed variation in SNARF-1 measurements (up to 0.18 pH units).Although the corals were maintained at constant seawater pH and temperature, our SIMS analyses detect significant variations in δ 11 B (1.8‰) between skeletal regions in coral 2 (Table 2), equivalent to a pH decrease of 0.12 units between the coenosteal spines and the skeleton deposited in between spines (Table 2).We also observe large variations in skeletal δ 11 B (outwith the precision of single analyses, Fig. 6) within each coral analysed.The lowest pH CM values generated from δ 11 B analyses in analyses on the growth edges of the microcolonies are 8.30 and 8.26 (corals 1 and 2 respectively) and are in good agreement with mean light/dark SNARF-1 pH CM estimates within uncertainty (0.072 and 0.15 pH units, 1 s, for δ 11 B and SNARF-1 analyses respectively).The large variations in δ 11 B pH CM in the dataset may reflect variable pH CM during deposition of subtly different skeletal regions and/or contributions of different crystallisation pathways to the skeleton.
We cannot rule out the possibility that subtle changes in environment occur between the microscope chambers (where SNARF-1 is measured) and the aquaria (where the majority of skeleton is deposited).However, temperature and light are maintained at the same levels for chambers and aquaria and we consider that any minor differences are unlikely to generate an offset between SNARF-1 and δ 11 B pH CM of the magnitude observed here.For example, SNARF-1 pH CM varies by 0.10 to 0.12 units between light and dark measurements (Table 1) while doubling light intensity reduced skeletal δ 11 B pH CM by only 0.04 in a Acropora sp. in another study (Dissard et al., 2012).Similarly temperature changes of up to 6 • C had little impact on the δ 11 B pH (≤0.03 units) in the Acropora sp.coral (Dissard et al., 2012).
We also note that SNARF-1 and δ 11 B measurements relate to subtly different areas of skeleton.SNARF-1 measurements were made at the very edge of the coral colony in the ECM in areas between freshly nucleated small crystals (up to ~40 µm across) which were isolated from the fully matured skeleton (Figs. 1 and 2).SIMS δ 11 B measurements analysed areas where these crystals had grown and joined together to produce a continuous sheet of skeleton (e.g.see the voids in the calcein stain in Fig. 4c which subsequently fill with unstained aragonite).pH CM may vary between the calcification media associated with the deposition of these areas e.g. if pH CM is lower at the very growth edge of the skeleton where symbiont density is very low (Venn et al. 2011;Jokiel 2011).However we observed no significant change in δ 11 B between the outermost 1 mm of skeleton and material deposited further back and it is likely that both of these regions have low symbiont density (Fig. 1a) suggesting that this does not explain the SNARF-1 and δ 11 B pH CM offset.

Comparing colony apex and edge δ 11 B
We compared the δ 11 B of aragonite deposited at the tips of the skeleton at the colony apex and at the growing edge of the same colony as it extended over the glass coverslip.Both the growth tips at the colony apex and the growth edges of the colonies in the present study appeared transparent as the tissues in these areas contain low symbiotic dinoflagellate density (Figs. 1, S1).We observed no significant difference in δ 11 B between these sites (Table 2, Fig. 7).Holcomb et al. (2014) measured the δ 11 B of the apices and whole lateral areas (4 mm of mature lateral skeleton together with the growing edge) of microcolonies cultured over a range of seawater pH.t test analyses of the dataset in Holcomb et al. (2014) indicates no significant difference in mean δ 11 B between the apices and lateral areas of colonies grown at a similar pH (7.94) to that used here (8.00), but a significant reduction in δ 11 B (and thereby estimated pH CM ) in lateral areas compared to apices in corals cultured at lower seawater pH.The mean δ 11 B value reported by Holcomb et al. (2014) for corals cultured at pH 7.94 is 23.6‰ and is considerably lower than the skeletal values observed here (26.1-26.8‰),equivalent to a pH offset of 0.19 units.The skeletal samples analysed by Holcomb et al. (2014) were overlain by coral tissue visibly containing symbionts both at the apices and the growth edge (see Fig. 1 in Holcomb et al. 2014).Further work is required to understand if and how the presence of photosynthetic dinoflagellates decreases the coral calcification media pH or alters the 10 B: 11 B pool for CaCO 3 formation.
The good agreement between the δ 11 B in the colony apex and growth edge in this study suggests that the biomineralisation process is broadly comparable at these 2 sites at ambient seawater pH.The growth of coral over glass slides has transformed our understanding of the biomineralisation process.It allows visualisation of coral skeletal formation (Venn et al., 2011), confocal dye measurements of the site media (Venn et al., 2011(Venn et al., , 2013)), accurate positioning of microsensors for pH and [CO 3 2-] measurement (Sevilgen et al., 2019) and identification of material for geochemistry analysis (Reynaud-Vaganay et al., 2001;Ram and Erez, 2021).However the skeleton deposited at the edge of glass slides has a unique morphology.At the very edge of the colony aragonite crystals can be deposited on the glass slide in isolation from the main skeleton.These crystals grow until they meet each other and further extension cannot occur, forming the patchwork visible under the glass coverslip in Fig. 4b.In contrast, aragonite precipitation in natural coral colonies occurs via precipitation onto the existing skeleton/substrate and generates the organised features (early mineralisation zone and thickening deposits) which are recognised in scleractinian corals (Drake et al., 2020).Thickening deposits are composed of acicular aragonite crystals which can extend over greater distances than observed during skeletal growth on the glass slides.In spite of these differences our finding that the apex and coral growth edge deposit skeleton with similar δ 11 B suggests that formation conditions (e.g.crystallisation pathway and environment) of both regions is comparable.

Implications for the δ 11 B pH proxy
We observe a large offset in mean pH CM as estimated by SNARF-1 and skeletal δ 11 B. The offset, 0.35-0.44pH units, is substantial given that seawater pH has probably varied by only ~0.3-0.4 pH units over the last 16ky (Kubota et al., 2014) and is predicted to decrease by up to 0.3 units over the course of the 21st century (Pörtner et al., 2019).SNARF-1 measures the pH of the extracellular calcification media while skeletal δ 11 B measures aragonite potentially formed from both the extracellular calcification medium and from the intracellular medium (Sun et al., 2020).Given these differences it is not possible to determine if the offset reflects real changes in pH CM between these pathways or if skeletal δ 11 B is affected by other factors besides pH CM .If the SNARF-1 and δ 11 B pH offset reflects a real pH change then SNARF-1 and skeletal δ 11 B have different applications in tracking the pH CM of the extracellular calcification media and of both the extracellular and intracellular calcification media respectively.Further research is required to determine how environmental change influences the proportions of skeleton derived from the ACC pathway and from ion-by-ion attachment in the extracellular medium.If the latter is correct, then further works needs to be undertaken to explore other influences on the δ 11 B proxy such as transmembrane B(OH) 3 transport and how this is affected by changing environment.At present it is clear that both SNARF-1 and skeletal δ 11 B pH CM estimates are affected by seawater pH (Venn et al., 2013;Holcomb et al., 2014;Hӧnisch et al., 2004;Reynaud et al., 2004;Allison et al., 2021) and can be used to provide insight into changes in seawater pH in the past (δ 11 B) and the response of biomineralisation to ocean acidification into the future.

Conclusions
We find that coral skeletal δ 11 B-pH CM estimates are higher than those of by SNARF-1 by 0.35-0.44pH units.This offset could reflect real variations in the pH of the calcification media associated with each measurement technique.This could occur if media pH is lower at the very edge of the skeleton where extracellular SNARF-1 measurements are made and aragonite is deposited as small isolated islands compared to the thickened areas of skeleton analysed by SIMS.SNARF-1 analyses the pH of the extracellular calcification media while δ 11 B could record both extracellular and intracellular media pH.Alternatively, the offset could indicate other changes in the biomineralisation process e.g. the transport of B(OH) 3 to the calcification media to offset δ 11 B CM to higher values than observed in seawater or the incorporation of B(OH) 3 (in contrast to B(OH) 4 − ) in the skeleton.Both SNARF-pH CM and δ 11 B-pH CM are influenced by changes in seawater pH and may both have applications in tracking seawater pH.

Collaborative group
EIMF consists of John Craven, Cristina Talavera and Cees de Hoog.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Coral 1, (a) Overhead view of microcolony growing on the glass coverslip.The centre of the microcolony is golden brown due to the presence of symbiotic dinoflagellates at high density, while the growing edge is transparent as symbionts are at low density.(b-d) SEM images taken in the yellow square shown in A at successively higher magnification.S indicates mature coenosteal spines.IS indicates immature coenosteal spines in the process of forming.Cr indicates groups of crystals isolated from the main skeleton.

Fig. 2 .
Fig. 2. In vivo images of the growing edge of a microcolony on a glass coverslip where confocal-SNARF-1 measurements of the ECM are made.(a) Transmitted light image showing isolated crystals (Cr), the margin of the coral tissue (M) and the external seawater (SW).(b) Merged confocal and transmitted light image showing the extracellular calcification media (ECM) stained with SNARF-1 (red) and the coral tissue including the calicoblastic epithelium (CE) that excludes the dye (grey).White circles indicate example regions of interest (ROI) where SNARF-1 fluorescence is analysed to obtain pH CM .Cal indicates calcein stain of skeleton and crystals made at the growing edge 24 h prior to the image being taken.

Fig. 3 .
Fig. 3. Coral 2 skeleton, (a) overhead view after cleaning.The white arrow indicates the position used for SNARF measurements.(b) The bottom half of the microcolony in a) showing the morphology of the skeleton edge (grey scale) and the position of the calcein stain (green).The images are taken from below using an inverted microscope so are a mirror reflection of the microcolony in (a).(c) A transmitted light micrograph of the same area of the microcolony as (b) after preparation for SIMS.The yellow arrows highlight the edge of the glass coverslip.Below this edge the surface of the mount is covered by the glass coverslip.Above this edge the glass coverslip and some of the skeleton has been polished away leaving patches of the aragonite skeleton at the mount surface.

Fig. 4 .
Fig. 4. Photomicrographs of part of the edge of coral 2, (a) before polishing showing the growth edge and the position of the calcein stain.The blue arrow denotes the edge of the skeleton used for SNARF measurements.(b) The mount after polishing photographed in transmitted light.The white arrows indicate the edge of the glass coverslip.Below this edge the surface of the mount is covered by the glass coverslip.Above this edge the glass coverslip has been removed leaving patches of the aragonite skeleton at the mount surface.Yellow arrows indicate markers in the skeleton (voids, dark spots or unique features) used to correlate images (a) and (b).(c) A close up of the outer edge in (a) and (d) a reflected light image of the polished section after SIMS analysis.Analyses appear as bright spots which puncture the blue colour of the gold coating.Black rectangles on (a) and (b) denote the positions of (c) and (d) respectively.The grey rectangle in (b) denotes the position of (c).

Fig. 5 .
Fig. 5. Profiles of (a) 11 B + and (b) 40 Ca 2+ counts over the duration of the 60 cycle analyses in representative sample and standard analyses.The Desmophyllum sp.cold water coral standard has a lower B/Ca than hermatypic corals(Allison et al., 2018) and is shown on a different axis in (a).

Fig. 6 .
Fig.6.Skeletal δ 11 B determinations (and equivalent reconstructed pH CM ) in all analysed corals.Analyses are divided into those conducted on coenosteal spines and in between spines in coral 1 and 2 (the microcolonies) and into analyses conducted on the apex and edge of coral 3. The typical internal precision (1 s) of analyses is shown by the error bar.

Fig. 7 .
Fig. 7. Mean seawater pH, boundary layer pH and pH CM estimates by each technique.Measurements are by SNARF-1 unless denoted as δ 11 B. Error bars show 1 s.

Table 2
Mean δ 11 B ± 1 s (‰) of different features/regions of coral samples and p values (t test for equal means) comparing features/regions within each coral.Significant differences between features are highlighted in bold.