The ocean is absorbing CO2 from anthropogenic emissions resulting in a drop in carbonate saturation and ocean pH (Bates et al., 2012). Cold waters take up and store more CO2 and thus the high latitudes will be amongst the first to experience undersaturated conditions (Orr et al., 2005). The response of marine calcifiers to ocean acidification has been shown to be taxon-specific (e.g. Ries et al., 2009; Pörtner et al., 2014); consequently, understanding the response of important key marine habitat builders is imperative to estimate potential impacts on their future ecosystem service. A large number of studies have concentrated on the physiological aspects of changes in carbonate chemistry (see Pörtner et al., 2014), much less is known about the impact this has on the skeleton grown by these organisms. While some species have been shown to continue to grow even under low pH conditions, a weakening of the ultra-structure can impair ecosystem functionality, i.e. its ability to withstand predators and wave action (Chan et al., 2012; Ragazzola et al., 2012; Melbourne et al., 2015).

Cold-water corals are important habitat builders that offer a range of microhabitats sustaining high biodiversity and provide nursery grounds for various species of fish (Fosså et al., 2002; Henry and Roberts, 2007; Roberts et al., 2008). The maintenance of their structural integrity is essential not just for the species itself but also for a wide range of species which depend on this habitat. Lophelia pertusa is the most common species of cold-water corals and has a cosmopolitan distribution with a wide temperature (4–12 ∘C) and salinity range (35–37 psu) suggesting a relatively high tolerance to environmental drivers. The species is typically found in regions with strong water currents and high productivity (Genin et al., 1986; Mienis et al., 2007). The modern distribution of cold-water corals appears to be constrained by the aragonite saturation horizon (the depth below which the waters become undersaturated with respect to aragonite), with 88.5 % of all cold-water coral records found above the aragonite saturation horizon (Davies and Guinotte, 2011; Guinotte and Fabry, 2008). Importantly for their future distribution, the aragonite saturation horizon has shoaled by 80–400 m in the North Atlantic since the industrial revolution (Feely et al., 2004) and model projections suggest a shoaling of up to 2000m by the end of this century resulting in vast areas of their current habitat being undersaturation with regards to aragonite (Orr et al., 2005).

Despite the strong link between the distribution of cold-water corals and the aragonite saturation horizon, Lophelia pertusa can calcify in undersaturated conditions (Form and Riebesell, 2012; Hennige et al., 2014; Maier et al., 2009, 2012), likely facilitated by its ability to increase the internal calcifying fluid pH at the site of calcification (pHcf), termed “up-regulation”. Most indications for up-regulation come from indirect determinations, e.g. measuring the boron isotopic composition (δ11B) of bulk skeleton samples of corals (Anagnostou et al., 2012; Holcomb et al., 2014; McCulloch et al., 2012; Trotter et al., 2011). Measurement of the pHcf at the site of calcifications in several corals confirmed an ability of the organism to influence the internal pHcf with a range of physiological processes (Al-Horani, 2003; Ries, 2011; Venn et al., 2013). The skeletal δ11B was observed to decrease with lower saturation state and pH of seawater (in total scale: pHT), suggesting a relative lowering of the internal pHcf in response to external pH decrease. At low seawater pHT, internal pHcf is still significantly higher than seawater pHT (up-regulation intensity, where ΔpH = pHcf–pHT, Anagnostou et al., 2012; McCulloch et al., 2012; Trotter et al., 2011), but does not reach internal pHcf levels observed under control conditions (Holcomb et al., 2014; Trotter et al., 2011).

This up-regulation ability has several implications: firstly, the potential to moderate the impact of projected future saturation state depends on the strength and efficiency of this mechanism (less efficient up-regulating species may be more adversely affected). Secondly, such differences in efficiencies will affect the reliability of δ11B as pH proxy when applied to paleo-climate reconstruction. Thirdly, the establishment of a pH gradient between external seawater and internal site of calcification requires energy reallocation (Al-Horani et al., 2003; Chalker and Taylor, 1975) and altered energetic demands may affect skeletal structure and strength.

In order to understand the interaction of biomineralization response, we analysed L. pertusa skeletons grown under natural control (Sula Reef pCO2= 405 µatm) and elevated CO2 conditions (CRSIII pCO2= 982 µatm). We uniquely combined Raman spectroscopy, Secondary Ionisation Mass Spectrometry (SIMS) and Synchrotron X-Ray Tomographic Microscopy (SXRTM) to examine whether ocean acidification causes any change in skeletal morphology of L. pertusa, such as thickness and growth patterns, or in the biomineralization processes. SIMS δ11B transects are compared between the high pCO2 (CRSIII) treatment and the natural conditions (Sula Reef). The δ11B are converted to pHcf to examine potential physiological adjustments in coral biomineralization under anticipated future ocean conditions of lower pHT.

The main growing edge outlined by the Alizarin staining lines marks the border to the experimentally precipitated distal skeleton (Fig. 2). In microscopic images, the staining line is only visible at the main growing edge (Fig. 2b), whereas in Raman fluorescence maps the outer skeletal surface before the start of the experiment can be traced over the entire colony (Figs. 2, 3). The Raman maps clearly display the orientation of the skeletal fibres and the location of the early mineralization zone (EMZ: Cuif and Dauphin, 2005, or rapid accretion front, RAF: Stolarski, 2003) and were used to compare skeletal formation before and during experimental conditions (Fig. 2c, d). At the macromorphological level, i.e. the arrangement of the main skeleton entities (EMZ and fibres), no notable difference between the natural and high pCO2 sample can be detected (Figs. 3c, d, 5, 6b).

Skeletal tomography reveals a large degree of morphological variability within the L. pertusa skeleton. Both the thickness of the outer wall and septa vary strongly as do the shape and length of the septa (Fig. 7). In addition, the vertical extension of newly grown material (after staining) was not even (Fig. 7). To enable a direct comparison between the natural material and that grown at high CO2, sections were taken directly above and below the Alizarin stain (Fig. 7b). These sections show that there is no change in structure for three different pCO2 treatments (Fig. 7c–h) which was confirmed by measurements on longitudinal polyp sections (Fig. S3). Overall thickness is slightly higher below the staining line than above (thickness ratio below/above of 1.10 ± 0.07) and range from 0.82 (±0.04), 1.14 (±0.09) to 1.32 (±0.16) for the CRSII, CRSI and CRSII, respectively (see Fig. S3).

All the samples and transects analysed for boron isotopes are summarized in Table 2. Repeated cross sections of the same corallites are reproducible (same colony and same polyp δ11B mean ± SE: 26.41 ‰ ± 0.83 vs. 26.08 ‰ ± 0.61 and 27.62 ‰ ± 1.09 vs. 27.55 ‰ ± 0.57) as were transects comparing two coral polyps from the same coral colony (27.96 ‰ ± 0.48 vs. 27.62 ‰ ± 1.09). Hence, we observed consistent values within the population grown in their natural environment within error.

All transects show heterogeneity in δ11B varying from ∼ 19.8–32.2 ‰, in particular when old branches with secondary thickening were analysed (Figs. 4, 6, S4). The variability of δ11B in L. pertusa spans approx. 14 ‰ and reveals lower values within the primary skeleton around the EMZ 22.48 ± 1.58 ‰ (mean ± SD, see Supplement Table S1) and an increase towards the outer skeletal rims (Figs. 4, 6, S4, S5). The secondary thickening is characterized by a higher δ11B of 26.97 ± 4.73 ‰ compared to the primary material and slightly reduced values at opaque nucleation sites (Figs. 6, S4).

Material that was deposited along the same skeletal region has the same δ11B (shown by repeated transects on several polyps; Figs. 4, 5). Transects on different polyps with similar diameter of the same colony show the same δ11B (transect 3,4, Table 2). The secondary thickening has the same δ11B independent of whether it was precipitated during natural or the high pCO2 condition (26.97 ± 4.73 and 27.8 ± 1.94 ‰, respectively; Fig. 3c) corresponding to a pHcf of 8.94 ± 0.15 or 8.95 ± 0.13, respectively. The sample grown only under high pCO2 has slightly higher average δ11B (24.52 ± 1.21 ‰) within the primary skeleton (Fig. 6) compared to primary skeleton precipitated under natural conditions. During the formation of the primary skeleton the calculated internal calcifying fluid pHcf is lower than during secondary thickening (8.58 ± 0.11 vs. 9.01 ± 0.15). The primary skeleton formed during high pCO2 conditions reveals a stronger internal pH up-regulation (8.74 ± 0.08) but still lower values than what was measured within the secondary thickening during high CO2 conditions (8.95 ± 0.13).

Lophelia pertusa has been shown to grow in undersaturated conditions. The amount of aragonite deposited under higher CO2 was at least equivalent to that deposited under natural conditions in agreement with findings in other studies showing sustained calcification using different analytical approaches (Form and Riebesell, 2012; Hennige et al., 2014; Maier et al., 2012). As L. pertusa grows by both vertical extension and by thickening, measurements by buoyant weight though do not provide information on whether the morphology is affected, i.e. does the skeleton thicken or thin during low saturation or remain unchanged?

Tomographic analyses showed that the morphology of Lophelia skeletons is highly variable and does not change under high CO2 even in undersaturated waters. We observed no change in the internal structure, in contrast other calcifying organisms, which show wall deformation in tube worms or coralline algae (Chan et al., 2012; Ragazzola et al., 2012). Arrangement and size of the primary skeleton, the template of size and shape of the corallite, do not change between treatments. The succession of growth bands is maintained and layers are formed even at undersaturated conditions. This finding corroborates a strong biological control on coral biomineralization. The only exception was less distinct organic layers, which might represent an impact on biomineralization in response to ocean acidification. The wider implications of a changed skeletal organic matrix need to be investigated further to understand its full implications.

A strong biological control on the biomineralization should also be expressed in its chemical composition, especially the boron isotope compositions (McCulloch et al., 2012; Hönisch et al., 2004). As deep-water corals grown in relatively stable environmental conditions, the high-resolution spatial isotopic and elemental heterogeneities suggest a biotic control that changes during growth. The isotopic heterogeneity is associated with specific skeletal regions (Blamart et al., 2007) and is also observed in other isotopes, e.g. δ18O (e.g. Rollion-Bard et al., 2010) and elemental ratios, e.g. Mg / Ca (e.g. Cohen et al., 2006; Krief et al., 2010). The Early Mineralization Zone (EMZ) is characterized with relatively low δ11B compared to adjacent fibrous aragonite with a higher δ11B and increases towards the outer wall. The EMZ is also known to have systematically lighter C and O isotopic composition by ∼ 4–5 and ∼ 8–10 ‰, respectively, compared to the fibrous aragonite part (Juillet-Leclerc et al., 2009; Rollion-Bard et al., 2010). Differences in C and O isotopes are suggested to be related to a faster growth of the EMZ suggesting that different skeletal regions are grown under different control or potentially even precipitation mechanisms. While the degree of heterogeneity with respect to boron isotopes in our samples is roughly equivalent to that of Blamart et al. (2007), the absolute values are offset by ∼ 10–14 ‰. In their study the δ11B translates to a maximum pHcf of approximately 10.2, while our data suggest values around 8.8 to 8.9 and agree with the bulk measurements of L. pertusa (McCulloch et al., 2012) and direct measurements of pH of calcifying fluids in symbiotic Scleractinian corals (Al-Horani et al., 2003) which determined values of 9.28 at light (additionally elevated by symbiont activity) and 8.13 at dark with seawater values of 8.2.

The use of δ11B as pHT proxy is based on exclusive borate incorporation. Rollion-Bard et al. (2011), by re-analysing samples from Blamart et al. (2007), suggested that both borate and boric acid are incorporated in the skeleton of Lophelia pertusa. In their study they observed NMR differences in skeletal boron coordination, which was used as indicator for boric acid incorporation. Considering the fraction of boric acid incorporation, they obtained pHcf values similar to values obtained in other (McCulloch et al., 2012) and this study. In addition, they suggested that the EMZ incorporated a higher proportion of boric acid. In our data, we did not test whether boric acid incorporation plays a role. Our δ11B to pH calculations though do not need any changes in incorporation to yield values which are comparable to bulk measurements (McCulloch et al., 2012). Applying their model to our data using δ11B signature of the EMZ, the same individual would incorporate very different proportions of boric acid which is not likely given the broad range of literature on boron in corals in general. Therefore, we question this variable boric acid incorporation hypothesis.

Our interpretations are based on the internal pHcf regulation calcification model (McCulloch et al., 2012) which assumes that the pHcf is offset from seawater pHT. McCulloch et al. (2012) reported a decrease in δ11B with decreasing pHT in several cold-water coral species including L. pertusa. They suggested that biological pHcf up-regulation determines the calcification response as described also for tropical corals (Holcomb et al., 2014). In contrast, we find similar growth rates between pCO2 treatments (Form and Riebesell, 2012) and similar δ11B values (within a 0.3 pH unit error) which questions a decreasing internal pHcf (Anagnostou et al., 2012; McCulloch et al., 2012).

If our interpretation of maintenance of internal pHcf within the secondary skeleton is correct it suggests that pH regulation can be decoupled from external seawater pHT and is a mechanism to explain the cold-water coral resilience. There are a number of parameters underreported in most acidification studies which might explain the difference in our findings. For example it has been suggested for a range of organisms that given a sufficient food supply, calcification can be maintained despite low saturation state (Schoepf et al., 2013; Thomsen et al., 2013). One possible explanation why we did not see a difference between treatments and the natural environment could be that food was provided in higher amounts than other studies on Lophelia and hence additional energy might have been available to support calcification. If this is the case than limited food availability might have a strong impact on the ability to regulate the internal pH and Lophelia growth (Rodolfo-Metalpa et al., 2015). Respiration rates of Lophelia were observed to decline (Hennige et al., 2014; Form and Riebesell, 2012) or unchanged (Maier et al., 2013) under elevated pCO2 suggesting that other energy sources, e.g. lipids, were used to maintain growth assuming other metabolic processes remained constant. Unfortunately, our growth experiments did not monitor whether reduced respiration changes the availability of energy reserves or tissue biomass between the treatments and presents an important scope for a new study of the physiological consequences of pHcf up-regulation and associated energy requirements.

We have found indications of changes in the organic matrix (OM). These less distinct OM bands could be an important step in biomineralization compromised under ocean acidification. In a tropical coral, OM production has been found to be affected by both up and down regulation of certain OM protein encoding genes (Moya et al., 2012), which might result in changes in the quality of the OM. The layered growth of biogenic organisms is a prominent feature (Cuif and Dauphin, 2005) and suggests a strong biological control of growth. Thus, a less clear banding could indicate that OM formation is compromised. However, further studies are necessary to better characterize the role, function and importance of skeletal organic layers.

In conclusion, the lack of sensitivity of L. pertusa to changes in pCO2 in growth, mineralogy and boron isotopes corroborates their strong biological control over biomineralization that is not easily disturbed under elevated pCO2 conditions. Our results raise a number of questions: (1) can energy be reallocated to up-regulate the internal pHcf to a suitable level which would complicate the applicability of Lophelia skeletons δ11B record as a paleo-pH proxy given the small ranges of pH difference studies often aim to resolve and (2) the role of OM production and quality need to be considered to improve our understanding of cold-water coral biomineralization and their response to acidification.