Effect of temperature rise and ocean acidification on growth of calcifying tubeworm shells (Spirorbis spirorbis): an in situ benthocosm approach

The calcareous tubeworm Spirorbis spirorbis is a widespread serpulid species in the Baltic Sea, where it commonly grows as an epibiont on brown macroalgae (genus Fucus). It lives within a Mg-calcite shell and could be affected by ocean acidification and temperature rise induced by the predicted future atmospheric CO2 increase. However, Spirorbis tubes grow in a chemically modified boundary layer around the algae, which may mitigate acidification. In order to investigate how increasing temperature and rising pCO2 may influence S. spirorbis shell growth we carried out four seasonal experiments in the Kiel Outdoor Benthocosms at elevated pCO2 and temperature conditions. Compared to laboratory batch culture experiments the benthocosm approach provides a better representation of natural conditions for physical and biological ecosystem parameters, including seasonal variations. We find that growth rates of S. spirorbis are significantly controlled by ontogenetic and seasonal effects. The length of the newly grown tube is inversely related to the initial diameter of the shell. Our study showed no significant difference of the growth rates between ambient atmospheric and elevated (1100 ppm) pCO2 conditions. No influence of daily average CaCO3 saturation state on the growth rates of S. spirorbis was observed. We found, however, net growth of the shells even in temporarily undersaturated bulk solutions, under conditions that concurrently favoured selective shell surface dissolution. The results suggest an overall resistance of S. spirorbis growth to acidification levels predicted for the year 2100 in the Baltic Sea. In contrast, S. spirorbis did not survive at mean seasonal temperatures exceeding 24 C during the summer experiments. In the autumn experiments at ambient pCO2, the growth rates of juvenile S. spirorbis were higher under elevated temperature conditions. The results reveal that S. spirorbis may prefer moderately warmer conditions during their early life stages but will suffer from an excessive temperature increase and from increasing shell corrosion as a consequence of progressing ocean acidification.


Introduction
Atmospheric carbon dioxide (CO 2 ) is a primary substrate for life on Earth, but is also a major driver of global scale environmental change, causing ocean acidification (Greene et al., 2012), controlling climate variability (Retallack, 2002;Galeotti et al., 2016), and initiating mass extinctions (Jaraula et al., 2013;Veron et al., 2009).The recent rapid CO 2 rise from anthropogenic emissions is a source of ocean acidification including pH reductions and alterations in fundamental chemical balances (Doney et al., 2007).Since the beginning of the industrial era, atmospheric pCO 2 rose from about 280 to 405 µatm (NOAA-ESRL, 20171 ) due to human activities such as fossil fuel combustion, cement production and deforestation.At the same time surface seawater pH decreased by 0.1 units, corresponding to 30% increase in the hydrogen ion concentration (Raven et al., 2005;Cao and Caldeira, 2008).It is predicted to further decrease by 0.3 to 0.4 pH units until the year 2100 when atmospheric pCO 2 levels may reach 950 µatm (IPCC, 2013).By the end of this century, the average surface ocean pH could be lower than it has been for more than 50 million years (Caldeira and Wickett, 2003) with severe consequences for marine calcifying organisms (Orr et al., 2005;Andersson et al., 2008;Erez et al., 2011).
The CO 2 rise also caused an increase of sea surface temperatures (SST) of about 1 • C on a global scale (European Environment Agency, 2015).However, mid and high latitude SSTs are more variable and increase more rapidly than the global average.For instance, the Baltic Sea annual mean SST warmed by up to 1 • C per decade between 1990 and 2008 (Elken et al., 2015).Warming of up to 6 • C and prolonged summer heat waves are expected until the end of 21st century (HELCOM, 2007;Gräwe et al., 2013).Rising temperatures and summer heat waves may increasingly affect mid/high-latitude marine ecosystems in the future, e.g., through micro-/macroalgae ecological functions, impacts on food-web structures or reduced reproduction (Knight-Jones et al., 1972;Graiff et al., 2015a;Werner et al., 2016).Stress from elevated temperatures can cause a depletion of organism's energy supplies resulting in energy deficiencies and increased mortality (Ivanina et al., 2013).
Coastal water pCO 2 and pH can be much more variable than that of the open ocean due to the effects of run-off, upwelling, eutrophication, atmospheric deposition and remineralisation (Doney et al., 2007).The Baltic Sea is an intra-continental nontidal brackish water environment with highly variable seasonal dynamics of pCO 2 and pH.Annual pH ranges vary from 8. 1-8.4 in the Kattegat area to 7.4-8.4 in the less saline eastern Baltic (Havenhand, 2012).Kiel Fjord and Eckernförde Bay are narrow coastal embayments in the western Baltic Sea.Surface water data from Kiel Fjord show a seasonal pH range from 7.3 to 8.5 (NBS scale) with pCO 2 varying from 385 to 2500 µatm (Thomsen et al., 2010(Thomsen et al., , 2013;;Wahl et al., 2015).Significant variations in pH and pCO 2 were observed along the coast line of the Kiel bight (Winde et al., 2017).In the Fucus meadows of Eckernförde Bay diurnal pH variations from 7.3 to 7.8 were found during an upwelling episode, while during normal summer conditions pH varied between 8.0 and 8.4 (Saderne et al., 2013).These observed ranges of pCO 2 and pH by far exceed the predicted levels at the end of the 21 st century.Therefore, the question arose: Are calcifying organisms living under such dynamic conditions better adapted for future ocean acidification?Consequences of ocean warming and acidification for marine organisms have been investigated in many studies (e.g., Reynaud et al., 2003;Marshall and Clode, 2004;Veron et al., 2009;Saderne and Wahl, 2013;Wisshak et al., 2013;Cornwall et al., 2016;Wahl et al., 2016).However, only few studies investigated combined effects of simultaneously increased temperature and CO 2 on whole ecosystems (Wahl et al., 2015).To study the combined impact of temperature rise and elevated CO 2 on typical marine calcifiers from the Baltic Sea, we carried out experiments in the Kiel Outdoor Benthocosms ("KOB", Wahl et al., 2015) to investigate calcification of the serpulid tubeworm Spirorbis spirorbis under near-natural habitat conditions as sessile epibionts on the thalli of Fucus seaweeds.
The brown algae Fucus vesiculosus and Fucus serratus are among the most widespread brown seaweed found on the coasts of the Baltic Sea.The pH in the seaweed ecosystems shows significant diurnal variations due to photosynthesis (high pH during the day) and respiration (low pH during the night) (Saderne et al., 2013).A diffusive boundary layer (DBL) of typically 50 µm to 2 mm thickness surrounds the algal thalli depending primarily on the flow conditions (Larkum et al., 2003;Spilling et al., 2010;Hurd and Pilditch, 2011;Wahl et al., 2016).Micro-or macroepibionts living in the DBL are affected by conditions with variable concentrations of chemical compounds (e.g.O 2 , DIC and pH) that are created by algal bioprocesses (Larkum et al., 2003).In the DBL of F. vesiculosus pH was found to increase by up to 1.5 units from dark conditions to bright daylight (Spilling et al., 2010;Wahl et al., 2016).Consequently, this surface boundary layer of the algae can potentially provide a shelter from ocean acidification during daylight time (Hendriks et al., 2014;Pettit et al., 2015).
Water temperature significantly influences growth, photosynthesis and metabolism of algae.Optimal temperature for growth of Baltic F. vesiculosus is in the range of 15 to 20 • C, but growth decreases rapidly when the water temperature exceeds 27 • C for several days (Graiff et al., 2015a).High temperatures may therefore have indirect adverse effects on epibionts, like S. spirorbis, because the ecological functions of their host algae may be reduced or damaged.
It is a common species of the Baltic Sea, where it lives in coastal macrophyte meadows characterised by large pH variations (>1 pH unit) and frequent aragonite under-saturation (Ω arag >0.6, Saderne et al., 2013).
S. spirorbis shells are purely calcitic.No or only questionable indications of aragonite have been reported (Ippolitov and Rzhavsky, 2015).The tubes consist of Mg-calcite with about 10 mol% MgCO 3 (Bornhold and Milliman, 1973) (Ni et al., in prep.), which has a similar solubility as aragonite (Plummer and Mackenzie, 1974;Walter and Morse, 1984;Morse and Mackenzie, 1990).Obviously, S. spirorbis is able to prosper in temporarily CaCO 3 under-saturated water.Other serpulid worms have even been reported to calcify in abyssal waters below the calcium carbonate compensation depth (Kupriyanova et al., 2014).
Previous work on Baltic S. spirorbis in laboratory experiments (Saderne and Wahl, 2013) found significantly reduced growth only at pH values lower than 7.7 (Ω arag <0.8).The study confirmed that the tubeworms were able to calcify in aragonite undersaturated water (Ω arag <1).This points to a high short-term tolerance for ocean acidification for at least some of the serpulid worm species.
Several recent ocean acidification experiments have included serpulid worms of a variety of species.Most studies focused on the tropical species Hydroides elegans (Lane et al., 2013;Chan et al., 2012Chan et al., , 2013;;Mukherjee et al., 2013;Li et al., 2014).The results indicated reduced growth, increased porosity and reduced mechanical strength of the worm tubes, as well as increased mortality of larvae at lowered pH (<7.9).
In the present study, we compare growth rates and corrosion features of Spirorbis spirorbis grown under ambient and elevated pCO 2 and temperature conditions in four seasonal experiments to test their sensitivity to ocean acidification and warming.Our results also provide new information about the life cycle and shell microstructure of S. spirorbis.The growth experiments were carried out in the Kiel Outdoor Benthocosms under near natural conditions, exposed to the weather and water conditions of the Kiel Fjord, by using a flow-through setup with water pumped directly from the fjord (Wahl et al., 2015).

Sampling
Healthy F. vesiculosus plants bearing intermediate amounts of live S. spirorbis were collected for 4 seasonal experiments in less than 1.5 m water depth in Eckernförde Bay (54 • 27'N, 9 • 53'E, Western Baltic Sea, Germany) in March 2013, June 2013, October 2013 and January 2014.The location is described in detail by Saderne et al. (2013) and Winde et al. (2017).Individual Fucus plants were selected by visual inspection to contain approximately the same volume of blades and similar amounts of S.
spirorbis.The typical density of S. spirorbis tubes at the start of the experiments is shown in Fig. 1.The collected plants were transported in a cool-box to GEOMAR (Kiel, Germany) for subsequent treatments.

Culturing
The samples were stained outdoor at the quay at GEOMAR in a closed 10 L transparent plastic box for three days in Kiel Fjord seawater with ∼50 mg/L calcein.The box was continuously bubbled with ambient air.S. spirorbis were fed at the start of the staining with Rhodomonas algae.The staining box was placed in a flow through water trough with seawater pumped from the Kiel Fjord to keep the temperature close to ambient conditions in the Fjord.The absorption of the dye into newly grown tubes provides a well defined starting point for growth under the experimental conditions.After three days staining, twelve individual Fucus plants were transplanted into the twelve subunits of the Kiel Outdoor Benthocosms (Wahl et al., 2015), fixed on a plastic grid at the bottom of the basins under 0.4 m of water.The incubations started immediately after staining.Average S. spirorbis starting populations were on the order of 100-200 specimens per subunit.
The twelve benthocosm subunits were assigned to four treatments, each treatment had three replicates (Wahl et al., 2015): "control treatment" with ambient pCO 2 (380 -400 µatm) and water temperature, "+CO 2 treatment" with 1100 µatm pCO 2 in the headspace of the subunit, "+T treatment" with water temperature elevated by 5 • C over ambient conditions, and "+CO 2 +T treatment" as a combination of both elevated pCO 2 and temperature.These conditions are considered as representative for acidification and temperature changes at the end of 21 st century (Wahl et al., 2015).Each benthocosm subunit had a volume of 1500 L and was continuously flushed with ambient fjord water, pumped from 1 m below the surface at a flow rate of about 65 L/h.Water in the subunits was additionally mixed by artificial waves with a frequency of 30 waves per hour.Four seasonal experiments were carried out: "spring" (04 April -19 June 2013), "summer" (4 July -17 September 2013), "autumn" ( 10October -17 December 2013), "winter" (16 January -1 April 2014).In total, each subunit contained 21 Fucus plants, but only one with Spirorbis, and a fauna of mollusks, arthropods and echinoderms.Details of the KOB setup and experimental parameters are described in Wahl et al. (2015), Graiff et al. (2015b) and Werner et al. (2016).After 10 -11 weeks of incubation, the twelve algal plants with S. spirorbis were collected from the benthocosms for freeze drying and further analyses.

2.3
Measurements and statistics of S. spirorbis growth S. spirorbis specimens were peeled off from the algal surfaces and photographed under an epifluorescence microscope.The initial and final diameter (in millimeter) of S. spirorbis shells and the length of the newly grown tube segments (mm) were measured after observing the position of the staining line (Fig. 2).The absolute tube length increase (mm) was measured as the length of the newly formed external arc of the tube between the staining front and the terminal tube edge, following Saderne and Wahl (2013).
From the spring, summer and and autumn experiments S. spirorbis tubes were collected from some basins for chemical analysis (Ni et al., in prep.).From each basin the newly grown tube parts of up to 20 specimens were cut off at the stain line, pooled, bleached, washed, dried and weighed.Bleaching was carried out using sodium hypochlorite with 1% active chlorine.
The measured length increase and final diameters were normalized by the initial diameter.In our analysis we compared the resulting five growth parameters: (1) Initial diameter, D i , (2) final diameter, D f , (3) growth, Gr, (4) growth/initial diameter, Gr/D i , and (5) final diameter/initial diameter, D f /D i .In order to test the robustness of the different parameters we measured Gr and D f of specimens with similar initial diameters in the autumn, winter, and spring populations.The results showed that Gr measurements were more sensitive in detecting growth differences than the D f measurements.
Normalization to the initial diameter was applied because growth of S. spirorbis tubes is strongly size-dependent.However, as the dependence is not strictly linear (see Section 3.5) we based all growth rate comparisons on the condition that the initial diameters of the starting populations were in the same range.The clearly bi-modal populations in the autumn experiment were treated separately (autumn-big and autumn-small).The summer populations and autumn-small populations, which both were dominated by juveniles, differed significantly from the autumn-big, winter and spring populations, dominated by adults.
Therefore no comparisons were carried out between these two sets of populations, because there was very little overlap in the initial sizes (compare Results Section, Fig. 5).Initial diameters of the summer and autumn-small populations overlapped to a high degree, but the medians differed significantly (two-way ANOVA, p=0.02).In order to derive comparable populations with similar D i in the summer and autumn-small data we selected sub-populations that had similar D i ranges and similar median D i values.Tubes outside this D i range were not used in the statistical analysis.For the autumn-big, winter and spring populations the initial diameters were not significantly different, as verified by Tukey's HSD tests.
Three-, two-, one-way ANOVA and Tukey's HSD tests were used for testing statistical significance of differences between the median values from different treatments and seasons.Each treatment had three replicates but, with a total duration of one year, seasonal experiments were not replicated.Median values were calculated for each of the treatment replicates based on the measured values, resulting in 12 basin medians for every seasonal experiment.In the three-way ANOVA, the three factors were temperature, pCO 2 and season.The temperature and pCO 2 factors had two levels, elevated and ambient.It should be kept in mind that the season factor here is a multiple factor which includes a range of parameters/conditions such as fjord temperature, pH, saturation state, nutrients and ontogenetic effects of S. spirorbis.Only differences caused by the temperature and pCO 2 offsets between the treatments were tested for statistical significance.The "seasonal" factor had no independent (multi-annual) replicates.Differences between seasonal experiments may consequently arise from any of the above mentioned seasonal factors, as well as from other unknown factors.
Assumption of normality of the models' residuals and homogeneity of residual variances were tested with Shapiro-Wilk's tests and box plots, respectively.Statistical analyses were conducted with R (Version 3.2, cran.r-project.org),PAST (Version 3.13, Hammer et al., 2001) and Microsoft Excel (Data Analysis Tool).A probability value of <0.05 was considered significant.

Microstructures
For localization of the calcein stain line S. spirorbis specimens were photographed with an epifluorescence microscope (Ax-ioScope A1, Carl Zeiss, Germany).Polished longitudinal and cross sections were used for electron microscopy.Samples were wet polished with grinding paper followed by polishing solutions of 9 µm, 3 µm and 1 µm grain size until no more scratches were visible on the polished surface.Backscatter electron images (BEI) and element concentration maps of calcium were taken with a JEOL JXA 8200 "Superprobe" electron microprobe (EMP) at GEOMAR Kiel, Germany.High resolution (2-3 µm per pixel) maps of calcium were recorded with 50 nA beam intensity at 15 kV, eight accumulations and 100 ms dwell time.Internal structures of stained skeletons were imaged on polished cross sections with a Zeiss Axio Imager.M2 microscope using white field and differential interference contrast (DIC).The Cy3 filter set was applied for detection of calcein.Images were acquired with a resolution of 1360*1024 pixels.Excitation wavelength was 495 nm.Emission from calcein (517 nm) was recorded.

Seawater chemistry
Temperature and pH NBS in all benthocosm treatments and the fjord water inflow were logged at two-hour intervals by GHL temperature sensors (PT1000) and pH glass electrodes, respectively.Air pCO 2 in the head space of the +CO 2 treatment subunits was monitored using infrared spectroscopy and kept at a constant level as described by Wahl et al. (2015).
Additionally, pH NBS values were measured daily using a Seven Multi1InLab Expert Pro (pH, Mettler Toledo GmbH, Giessen, Germany).The pH electrode was calibrated with NBS pH-buffer solutions (4.001, 6.865) kept at in-situ temperature (Winde et al., 2014(Winde et al., , 2017)).The pH 9/10 buffer was avoided to prevent impact of possible CO 2 contamination under field conditions.
The stability of the electrodes' Nernst slope and the applicability of the 2-point calibration to higher pH was previously tested by the measurement of a pH 9 calibration solution.Independently calculated pH values (using CO2SYS) based on measured dissolved inorganic carbon (DIC) and total alkalinity (TA) values showed good agreement with the measured pH in the range between 8 to 9 (Wahl et al., 2015).Discrete water samples were taken as described by Wahl et al. (2015) and analysed for TA two times a week, as well as for DIC on a monthly base.Water samples for DIC analysis were filled bubble-free into 50 mL Winkler bottles, poisoned by the addition of one drop of saturated mercury chloride (HgCl 2 ) solution and measured via coulometric titration (Johnson et al., 1993).To remove microbes and particles TA samples were filtered through 0.45 µm Minisart syringe filters (Sartorius SFCA, Sartorius) and measured by potentiometric titration using 0.01 M HCl with a Schott titri plus and an IOline electrode A157.NaCl was added to avoid changes in ionic strength during the analysis.The titration cell was kept at 25 • C. Measurements were calibrated using certified seawater standards for DIC and TA (Dickson et al., 2003(Dickson et al., , 2007)).
The speciation in the dissolved carbonate system, including the carbonate ion concentration, was calculated from pH NBS , TA, temperature and salinity using the code of the CO2SYS software package for MATLAB, Version 1.1 (Lewis and Wallace, 1998;van Heuven et al., 2011) with constants recommended for best practice (Dickson et al., 2007;Orr et al., 2015), i.e.K 1 and K 2 from Lueker et al. (2000), K S and K B from Dickson (1990), K F from Dickson and Riley (1979), K W , K 1-3P and K Si from Millero (1995) and the total boron-salinity relationship from Uppström (1974).The K 1 and K 2 constants from Lueker et al. (2000) are defined for a salinity range from 19 to 43, while the brackish Kiel Fjord water ranged from about 10 to 20 psu during the experiments.An alternative set of equations for K 1 and K 2 is available from Millero (2010) for salinities as low as 1.However, as discussed in Orr et al. (2015) applications of the latter showed discrepancies on different pH scales.Therefore, we used the Lueker et al. (2000) constants.Using the latter to calculate carbonate ion concentrations at salinities as low as 10 psu resulted in offsets of less than 0.5 % compared to the Millero (2010) constants, which is negligible for our interpretations.
Salinity and concentrations of Ca, Si and P were measured in all benthocosm treatments two times a week.Dissolved Si, P and Ca were analysed by inductively-coupled plasma optical emission spectrometry (iCAP 6300 DUO, Thermo Fisher Scientific) after appropriate dilution.The accuracy and precision was routinely checked with the certified seawater standard CASS-5 as previously described (Kowlaski et al., 2012).PO 4 was also measured by spectrophotometry using a QuAAtro nutrient analyser (SEAL Analytical; (Winde et al., 2014).Accuracy and precision checked by replicate analyses of a solution from powdered phosphate salts were better than 8 % RSD.Si and P concentrations were usually too low to have a substantial impact on alkalinity (Winde et al., in prep.).Alkalinity and salinity behaved conservatively in our experiments and showed no significant systematic variability on diurnal timescales (Winde et al., in prep.;Wahl et al., 2015).Calcium concentrations ranged from about 3.5 to 6 mM and were closely coupled to salinity (R 2 >0.9).
The saturation state in the benthocosm treatments with respect to the calcium carbonate of S. spirorbis tubes was calculated considering the shell composition.It has been shown, that the thermodynamic stability of biogenic Mg-calcites differs from pure calcite (Plummer and Mackenzie, 1974;Busenberg and Plummer, 1989) and varies with the Mg content.The MgCO 3 content of S. spirorbis tubes is about 10±1 mole% (Ni et al., in prep., Bornhold and Milliman, 1973).Unfortunately, the solubility of S. spirorbis was not explicitly determined, so far.According to different experimental studies biogenic Mg-calcite with about 10 mol% MgCO 3 has a solubility which is in thermodynamic equivalence to aragonite (Walter and Morse, 1984;Morse and Mackenzie, 1990;Andersson et al., 2008).Therefore, the saturation state with respect to aragonite (Ω) was taken as an estimate for the Mg-calcite forming the S. spirorbis shell.It should, however, be kept in mind that the solubility of biogenic Mg-calcites may not only differ with shell composition, but may also depend on crystal ordering, trace element impurities and other mineralogical factors (Mackenzie et al., 1983).Most of these factors increase the solubility of Mg-calcite.
Saturation states in the benthocosms at the measured in-situ temperatures and salinities were calculated from carbonate ion concentrations, calcium ion concentrations and the apparent solubility constant (K * sp ) of aragonite (Mucci, 1983): Only pH and temperature were measured with two-hourly resolution (Wahl et al., 2015).All other parameters ([Ca 2+ ], TA, salinity, Si, P) were interpolated to calculate diurnal variations of Ω (Fig. S1).Linear interpolation is justified by the conservative behaviour of these properties.The resulting two-hourly resolved time series of Ω were used to estimate the mean saturation state and the percentage of time when treatments were undersaturated with respect to S. spirorbis tube Mg-calcite.We use the daily insolation sum measured at the GEOMAR meteorological observatory2 , situated close to the benthocosms, to compare with the diurnal pH variations.Average daily insolation ranged from 0.2 kWh/m 2 in December to 6.6 kWh/m 2 in July (Fig. 3).

Seawater carbonate chemistry and saturation state
Variations of physical and chemical parameters (TA, pH, temperature, salinity, etc.) in the control treatments of the four seasonal experiments are shown in Fig. S2.The calcium carbonate saturation state of the seawater (Ω) in all basins was dominantly controlled by the pH.Average diurnal cycles showed a minimum in pH and Ω around sunrise followed by a late afternoon maximum (Fig. S1).The pH values showed strong diurnal fluctuations in all treatments.Average day/night pH differences were smallest (<0.05) in December 2013 and largest (up to 0.6) in June, July and August 2013, as well as in February/March 2014 (Fig. 3).The pH amplitudes clearly follow insolation showing saturation behaviour at high insolation values, most pronounced in the ambient CO 2 treatments (Fig. 4).Generally, pH values declined from the spring to the autumn experiment and reached a minimum in November and December (Fig. 3).
Saturation states closely followed the pH dynamics.Average saturation was highest during the spring experiment when all treatments were generally oversaturated with respect to aragonite and Mg-calcite (Ω>1).Basin waters were undersaturated (Ω<1) only during 6 to 51% of the time during the spring experiment (Table 1).The lowest saturation states occurred during the autumn experiment with Ω<1 during 81 to 100% of the experiment duration.Average autumn saturation ranged from 0.6 to 0.8 (Table 1).It was only slightly elevated during daytime (Ω max of 0.6 to 1.1, Fig. 3).Average day-night differences in Ω were largely tracking the diurnal pH amplitudes with smallest differences during the autumn experiment (<0.1 in December 2013) and large fluctuations in February/March, June and July (up to 2.2, Fig. 3).

Spirorbis spirorbis tube size and ontogenetic cycle
The sizes that the S. spirorbis shells reached before the experiments in their natural environment are indicated by the initial diameters.They reflect the size distributions under natural conditions.In contrast, the final diameters of our specimens reflect changes from the initial sizes under experimental conditions.Note that only stained specimens were included in the analysis.
Therefore, juveniles that settled during the experiments and specimens that did not calcify during the staining were not included.
The final and initial diameters of 2782 stained and photographed S. spirorbis tubes from all four seasonal experiments were in a range of 0.2 mm to 4.0 mm (Fig. 5).The tube with the biggest final diameter (∼4.0 mm) was found in the winter experiment.
The smallest measured shell diameters (0.2 mm) occurred in summer and autumn (Fig. 5 B,C).The size distributions of the shells indicate distinct populations that, in summer and autumn, were separated by a minimum in shell counts at a diameter of about 1.3 mm (Fig. 5 B,C).Accordingly we classified S. spirorbis specimens into two general populations: "small" (diameter <1.3 mm) and "large" (diameter>1.3mm).Seed et al. (1981) observed reproduction of S. spirorbis at shell diameters >1.9 mm.Therefore, our "small" populations consist of juveniles, while the "large" populations are mostly adults but may include immature specimens.
"Large" specimens were observed in the starting populations of all seasons (Fig. 5 B-E), including spring and summer, which is in accordance with the maximum life span of S. spirorbis of about 1.5 years (Seed et al., 1981).Most "small" specimens grew to "large" sizes during the ∼10 weeks duration of the experiments (Fig. 5 F-I).The majority of S. spirorbis in the maximum size range (>3 mm) occurred by the end of the winter and spring experiments (March, June, Fig. 5 A,H,I).The initial shell diameters of the S. spirorbis autumn population showed a clear bi-modal distribution (Fig. 5 C).A juvenile population with a modal diameter of 0.6 mm ("autumn-small") was clearly separated from an "adolescent/adult" population with a modal diameter of 1.8 mm ("autumn-big").Initial diameters in the intermediate range of 1.4 mm -1.5 mm were scarce.A similar size distribution was found in the summer experiment.However, the "large" population had very few specimens in summer (Fig. 5 B).
Juveniles occurred in all four seasons but were rarely observed in winter and spring.The proportion of juveniles in the initial populations decreased from July (Fig. 5 B) to April (Fig. 5 E).Accordingly, the majority of the S. spirorbis specimens at the start of the summer and autumn experiments were in the juvenile stage (<1.3 mm), while the winter and spring experiments were dominated by "large" specimens (Fig. 5).The modal initial diameter increased systematically with the sequence of the seasons from July (∼0.7 mm, Fig. 5 A,B) until April (∼2.4 mm, Fig. 5 A,E).The spring, winter and autumn-big populations started with similar initial diameters (modes of 1.8 to 2.5 mm, Fig. 5 C-E) and all grew into a typical final diameter range (modes of 2.5 to 2.8 mm, Fig. 5 G-I) representing the most common size of adult S. spirorbis.
As shown above the diameter increase of S. spirorbis tubes during the experiments strongly depended on the season and the initial size distribution of the populations.Diameter increases ranged from 4 µm/day for the adult-dominated population in spring to 20 µm/day for the juvenile population in autumn.Modal diameter increases of the summer, autumn-big and winter populations were similar (Fig. 5) with values of about 10 µm/day.This ontogenetic influence has to be taken into account when interpreting growth rates in terms of temperature and saturation state.

Tube microstructure
SEM pictures of S. spirorbis sections (Fig. 6) show a relatively rough and irregular outer tube wall surface whereas the inner surface is smooth.The internal wall structures consist of convex-forward lamellae or chevrons (Fig. 6b).New lamellae were laid down by the worm on the anterior tube surface, forming curved convex-forward layers, wrapping the end of the tube wall to completely cover the end of the anterior tube wall with a new layer.Thin crescent pores exist in the wall interior between the chevrons (Fig. 6b).These pores taper towards the inner and outer rims of the tube wall where the chevron lamellae fuse into a dense, calcium-rich wall (Fig. 6).The high calcium concentrations indicate that not organic but strongly calcified dense layers armour the inner and outer tube wall surfaces.
A comparison of a cross section (Fig. 6a) and a longitudinal section (Fig. 6d) through S. spirorbis shells reveals the complex shape of the growth lamellae.The convex-forward layers are additionally curved upward, forming convex-upward lamellae in longitudinal sections.The convex-upward lamellae were built upward successively from the bottom on both sides of the tube and then converge at the tube top.The growth direction is indicated by the convex layering.
In addition, the inner and outer sides of each convex-forward layer of the tube walls are asymmetric (Fig. 7).The fluorescent, stained skeleton outlines the pattern of lamellae which were accreted during the 3 days staining period.The newly grown lamellae cover a large area along the inner tube wall surface, while little new material is attached to the outer tube wall surface.
The bottom of the tube, which was attached to the substrate, is relatively thin and characterized by parallel planar lamellae.
An idealized sketch of the S. spirorbis tube structures is shown in Fig. 8.Where the wall of a new whorl attached to an older whorl it formed a thickened wedge-like structure partly filling the gap between the old and new whorl (Fig. 6d, 8).These wedges are usually calcium-rich, densely calcified (Fig. 6e), increasing the stability of the shell.The tube diameter of the whorls and the tube wall thickness generally increased as the S. spirorbis shell grew (Fig. 6d).The wall thickness ranges from about 30 to 180 µm.It is thicker in the fully developed shell parts and tapers towards the tube opening (Fig. 6a, b).

Shell corrosion
Shell corrosion (Fig. 9) occurred in all treatments during all seasons, but was most commonly observed in the high pCO 2 treatments of the autumn and winter experiments (Fig. 10, Table 2).In the basins of these treatments up to 75 % of the specimens showed corroded shells.On average, the proportion of corroded samples (P corr ) was highest in the autumn +CO 2 treatment and in the winter +CO 2 +T treatments, with treatment averages of 58% and 62%, respectively.In contrast, corrosion was nearly absent in the control treatments, where in all four seasons P corr values were lower than 1.5%.Additionally, corroded specimens were nearly absent in all spring treatments, except for the +CO 2 +T treatment.
The percentage of corroded samples was clearly related to the saturation state (Fig. 10).Except for one basin from the spring +CO 2 +T experiment P corr was below 10% when average saturation (Ω) was above 1.For average saturation Ω>2 corroded shells were completely absent.On the other hand, although P corr = 0% was observed in basins with an average saturation as low as 0.8 (basin D2, autumn control), corrosion frequencies generally increased in undersaturated basins.For Ω<1 we observed a significant inverse correlation between P corr and saturation state: Notably, shells grew significantly even in undersaturated waters.Thus corrosion selectively affected the previously grown parts of the shell (Fig. 9b).
P corr was independent of temperature in autumn, winter and spring (R 2 =0.03, n=42, p=0.28), but temperature may have fostered corrosion and bioerosion in the summer experiments.Ambient temperature treatments of the summer experiments showed very low P corr values (Fig. 10).However, the few recovered samples from the elevated temperature experiments were highly corroded and showed very little net growth.Unfortunately, because very few specimens were recovered from these treatments of the summer experiment, P corr values could not be determined.
Strong bioerosion by microborers was observed in a cross section of a summer control specimen.Numerous microborings of about 5 to 45 µm diameter affected the outer tube wall (Fig. 11).The microborings penetrated the whole tube wall.This is in contrast to the shell corrosion of the other seasons, which mostly affected the outermost layer of the tube wall (Fig. 9b).

Growth rate
The length of new tube segments that grew during an experiment (Fig. 2: "growth", Gr) varied considerably between populations and seasons, ranging from less than 0.1 mm up to 7.3 mm.This corresponds to a range in growth rates of 1 to 100 µm/day.
The longest newly grown tube in all experiments (7.3 mm) occurred in the autumn-small population.Growth was found to be inversely correlated with the initial diameter of the shells (D i ), i.e. smaller tubeworms generally grew faster than bigger ones (Fig. 12).The correlation is highly significant Gr(mm) = −1.1±0.05• D i (mm) + 5.17±0.09;R 2 = 0.41; n = 2783; p = 0; (3) for D i ranging from 0.2 to 3.5 mm.
Growth of the winter populations showed the highest variability of all treatments, ranging from 0.4 mm to 6.3 mm (Fig. 12).
Growth rates and initial sizes in winter were similar to those of the autumn-big populations.This indicates that the tubeworms from these two experiments were in the same developing stage, although they represented different generations of S. spirorbis populations (Section 3.2, Fig. 5).
In a subset of specimens from the spring and autumn (control, +CO 2 +T) and the summer (control,+CO 2 ) experiments average weights of newly grown tube segments, W t , were determined (Table S1).The results show similar weight increases in spring and summer of 0.1 -0.9 and 0.2 -0.6 mg/shell, respectively.In contrast, W t values in autumn were significantly larger, ranging from 1.2 to 2.1 mg/shell.As shown above Gr varied seasonally (Fig. 12).For the weighed specimens mean Gr ranged from 2.2 to 3.8 mm and 3.8 to 5.4 mm in summer and autumn, respectively.It was only 1.0 to 2.3 mm in spring.We accordingly normalized W t by Gr.This resulted in overlapping W t /Gr ranges for spring and autumn of 0.1 -0.4 and 0.3 -0.4 mg/mm of tube, respectively (Fig. 13).The summer shells increased their weights by only 0.1 -0.2 mg/mm of tube.
Generally, this is in agreement with smaller final shell sizes in summer (Fig. 5) and consequently smaller tube widths (Fig. 2).
Assuming a cylindrical tube geometry and a constant wall thickness of 0.1 mm the measured tube width values (Table 3) allow to estimate average shell densities of 1.1±0.3 and 1.8±0.3g/cm³ (±1sd) for the summer and autumn specimens, respectively.
This indicates that density and/or tube wall thickness of the summer tubes was 38±13% lower compared to the autumn tubes.

Treatment effects
Only very few broken and strongly damaged S. spirorbis specimens could be recovered from the elevated temperature treatments (+T, +CO 2 +T) of the summer experiment.Growth of broken and damaged tubes was not measured (indicated by "no data" in Fig. 14).In these experiments a temperature-driven collapse of the grazer community had caused epiphytic overgrowth of Fucus thalli and S. spirorbis tubes leading to an increased mortality (Werner et al., 2016).Except for these elevated temperature summer treatments there was no significant treatment influence (pCO 2 or T) on growth in spring, summer, winter or autumn (big population).Notably, elevated pCO 2 had no detectable influence on growth in any of the four seasonal experiments (Fig. 14).
In the autumn-small population, temperature caused a significant increase of growth, but only under ambient pCO 2 condition (Fig. 14; three-way ANOVA and Tukey's HSD tests, p=0.03 for Gr/D i , p=0.04 for D f /D i ).There were marginally significant interactions (p<0.1)among the factors temperature, pCO 2 and season.Each factor influenced the growth parameters in each experiment differently due to the effects of the other two factors.
No significant correlation between saturation (Ω) and growth parameters (Gr, Gr/D i , D f /D i ) was found (p>0.14 to 1.0, R 2 = 0.00 to 0.40, seasonal basin data in Tables 1 and S2).The extension rates of S. spirorbis tubes were not negatively impacted by the saturation state of seawater.In contrast, weight increase (W t /Gr, Table S1, Fig. 13) showed a significant positive correlation with saturation state in spring (R 2 =0.94, n=6, p=0.002) and autumn (R 2 =0.68, n=6, p=0.04).Weight increases of the autumn specimens were similar to spring, although the tubes formed in undersaturated water.In the summer experiment no significant correlation was observed (R 2 =0.48, n=6, p=0.13), but the data lie close to the spring trend line (R 2 =0.88, p<10 -5 for summer and spring combined, Fig. 13).

Differences between seasonal experiments
The term "seasonal" in this study collectively describes differences between the four seasonal experiments which are not influenced by treatment effects.They may comprise truly seasonal variations but also variability on different time scales.
As described in Section 3.2 we observed a significant seasonal variation in the proportion of juvenile specimens (Fig. 5), indicating limited reproductive activity during the cold seasons.In spring and winter less than 10% of the stained specimens were juveniles (D i <1.3 mm) while there were more than 84% juveniles in summer and autumn.As a consequence D i values were seasonally biased, which can explain at least some of the seasonal variations of Gr (Fig. 12).
In order to detect additional seasonal impacts on S. spirorbis tube growth we compared the juvenile populations in the control treatments of the summer and autumn experiments.Populations with similar mean D i were selected (Table 3).No significant seasonal impact on growth (Gr) was found (Fig. 15).However, the final diameters of the autumn-small population were significantly larger than those of the summer experiment (Tukey's HSD test, p<0.01,Fig. 5F-G).Additionally, the width of the newly grown tubes (Fig. 2) differed significantly between the two seasons (Tukey's HSD test, p<0.001).The tubes that formed in autumn were wider than the summer tubes.Two-way ANOVA of tube width values from the control and +CO 2 treatments of the two seasons (Table 3) indicated no treatment effect (p=0.21) but a significant seasonal impact (p<0.0001).
There was no significant difference in growth of the "large" populations between the winter and autumn experiments (Fig. 12).There was no influence of temperature, pCO 2 or season on Gr/D i of these populations (three-way ANOVA, p>0.67).
All populations that had "large" sizes at the start of the experiments (spring, autumn-big, winter) grew to a similar final size distribution at the end of the experiments (Fig. 5 G-I).Consequently, because the initial diameters of the winter and autumn-big populations were generally smaller compared to spring (Fig. 5 C-E), average growth was higher in autumn and winter than in spring (Fig. 12).

Discussion
As shown in Figs. 3 and S2 there was strong intra-and inter-experimental variability in several environmental parameters, most prominently temperature, insolation, pH and saturation state, but also salinity and nutrient availability.Further, food supply and faunal/floral composition varied during the experiments as discussed below (Section 4.5) and shown in Werner et al. (2016).
This natural variability is an intentional part of the benthic mesocosm set-up as it allows to consider the dynamics of benthic communities reacting to environmental changes under near-natural boundary conditions (Wahl et al., 2015(Wahl et al., , 2016)).On the other hand, the lack of control on several environmental parameters also has drawbacks for the interpretation, comparability and reproducibility of results from different seasonal experiments.As described in Section 2.3, we use the term "seasonal factors" to collectively describe variations of experimental conditions between the four experiments, including environmental parameters and the ontogenetic development of S. spirorbis.While some of these factors are clearly dominated by seasonal change (e.g.light, temperature), others may vary on different time scales.Without multi-annual replicates we can not prove the seasonal nature of the observed changes in S. spirorbis growth between the four experiments.We therefore use the term "seasonal" as a simplifying descriptor of inter-experimental changes, although their seasonal nature needs to be verified in future multi-annual experiments.

Water chemistry
The aim of the study was to detect influences of elevated pCO 2 and temperature on growth and destruction of calcareous tubeworm shells under near-natural conditions in different seasons.The temperature manipulations produced consistent offsets of 4-5 • C between the respective treatments (Fig. 3).However, the basin water acidification (pH, saturation state Ω) induced by elevated pCO 2 was more complex.The average pH and Ω values were highest in the control treatments and lowest in the +CO 2 +T treatments.Intermediate values occurred in the +CO 2 and +T treatments.At the same pCO 2 level pH was lower in the elevated temperature treatments (Table 1; Wahl et al., 2015).This was probably caused by biological activity or nutrient cycling.
It cannot be explained by the carbonate chemistry, which would result in higher pH at elevated temperatures under otherwise constant conditions (Lewis and Wallace, 1998).The mean pH difference between the +CO 2 +T and the control treatments was 0.2 units in summer and autumn and 0.4 units in spring and winter (Tab.1).This simulated pH change is in good agreement with the predicted pH decrease at the end of this century (Omstedt et al., 2012;IPCC, 2014).
The seasonal fluctuations of pH (0.4 to 0.6) and Ω (0.9 to 1.9) exceeded the respective differences between treatments (pH: 0.2 to 0.4, Ω: 0.2 to 1.2).This has to be considered when comparing data from different seasons (e.g.Fig. 13).In addition, the strong diurnal cycles of pH (≤0.6) and saturation (≤2.2) complicate interpretations of carbonate chemistry impacts on tube growth and corrosion (e.g.Fig. 10).
Such interpretations are further hampered by potential impacts from the diffusive boundary layer (DBL) forming at the surface of Fucus, the substrate of S. spirorbis tubes (Spilling et al., 2010;Wahl et al., 2016).Photosynthetic activity during the day can elevate pH and saturation state in the algal DBL compared to the bulk fluid.Average saturation of the autumn experiment was as low as Ω=0.6 in the bulk fluids of some treatments.To elevate saturation from this value to slight oversaturation (Ω=1.1)pH has to be increased by log(1.1/0.6), i.e by about 0.3 pH units.A pH elevation of this magnitude was reported by Wahl et al. (2016) at a DBL thickness corresponding to the height of S. spirorbis tubes.However, these observations were made in stagnant water while conditions in the benthocosms were quite turbulent due to artificial waves generated every 2 minutes (Wahl et al., 2015).Additionally, considering that insolation was reduced during the autumn experiment (Fig. 3), it appears unlikely that photosynthesis-driven daytime pH elevation (Fig. 4) was sufficient to overcome under-saturated water conditions in the DBL.This means that S. spirorbis was able to build tubes with above-average rates (Gr of ∼5 mm, Tab S2; Fig. 12) in spite of constant under-saturation (t Ω<1 =100%) in the autumn treatments A1, B1 and E2 (Tab.1).S. spirorbis tube growth in under-saturated water was previously observed by Saderne and Wahl (2013).

Reproduction and life cycle
S. spirorbis reproduces and releases larvae predominantly during the warm seasons (Knight-Jones and Knight-Jones, 1977;Seed et al., 1981).Larvae settle in episodic pulses that may be coupled to fortnightly lunar or tidal cycles (De Silva, 1967;Daly, 1978).The episodic larval settlement provides an explanation for the presence of distinct populations in our experiments (Fig. 5 B-E).In line with previous studies, we found living (actively calcifying) juveniles at the beginning of all four seasonal experiments, i.e. in January, April, July, and October.This indicates that the Eckernförde Bay S. spirorbis population reproduces throughout the year, although, in January and April juveniles were very rare.At the end of the summer and autumn experiments (September and December, respectively) we found numerous unstained living S. spirorbis specimens on the Fucus thalli which had shell diameters <1.3 mm.These juveniles obviously had settled during the experiments, indicating continuous reproduction in the benthocosms from July to December.
In addition to temperature, fecundity of S. spirorbis is affected by salinity and food supply and increases with individual size and age (Daly, 1978;Kupriyanova et al., 2001).Salinity fluctuated strongly during the experiments (from ∼10 psu in June to ∼20 psu in November-January, Fig. S2), but our data do not allow to draw conclusions about salinity impacts on reproduction.
Food supply for the filter-feeding Baltic tubeworms is generally lower in winter and increases when increased light availability promotes phytoplankton growth in spring, summer and autumn.Juveniles were rare in the initial populations in April, when the water temperature was still <10 • C and in January when temperatures had decreased to <5 • C (Fig. 3).In April, however, phytoplankton biomass is already high in the Kiel Bay area (Rheinheimer, 1996).Therefore, temperature probably dominates over food availability in controlling S. spirorbis reproduction in Eckernförde Bay.
S. spirorbis tube walls are purely calcitic and two-layered with an irregularly oriented prismatic (IOP) ultrastructure in the chevrons of the wall's core and a spherulitic prismatic ultrastructure (SPHP) of the thin outer wall region (Ippolitov and Rzhavsky, 2015;Vinn et al., 2008).The IOP chevrons and the SPHP structure are also common in a range of other serpulid genera (Crucigera, Floriprotis, Pyrgopolon, Spiraserpula; Gee and Knight-Jones, 1962;Vinn, 2011).Weedon (1994) pointed out that this complex internal tube architecture is difficult to explain with simple pasting models for serpulid calcification, i.e.
secretion of calcium carbonate granules or of a mucus paste with small calcite crystals that are molded into the calcitic tube.
It is likely that extracellular organic matrices and scaffolds play a role in tubeworm biocalcification (Tanur et al., 2010).Thin layers of organic matrix could be secreted onto the surface of the growing shell, as indicated by the chevron-like pores between growth lamellae (Fig. 6b).
Chevron-like accretion of new tube lamellae is indicated by the shape of the shell's stain line (Fig. 7).The figure additionally shows that synchronously with the accretion of new chevron lamellae new material was added in a thin layer to the inner tube wall.This wall thickening is in agreement with the observed tapering of the tube walls near the tube mouth (Fig. 6a, b).The asymmetric chevron lamellae structure of the S. spirorbis shells reported here (Fig. 7) has not been recorded previously in serpulid tubes.It shows that the inner and outer tube wall linings are differently constructed.S. spirorbis prefer to consolidate the inner surface of the tube while constructing new chevron layers.
In many S. spirorbis specimens the chevron lamellae of the central tube wall became visible as ring structures when the outer tube wall layer broke off or dissolved (Fig. 9).The outer tube wall layer appears to be susceptible to corrosion in spite of its massive densely calcified nature (Fig. 6).

Shell corrosion
In a recent study Saderne and Wahl (2013) incubated S. spirorbis in a laboratory experiment for 30 days at three different pCO 2 levels (450 µatm, Ω=1.8; 1200 µatm, Ω=0.8; 3150 µatm, Ω=0.4).They used specimens from the same site as in the current study, i.e. from Eckernförde Bay.The tubes exhibited substantial dissolution at the highest pCO 2 condition (3150 µatm, Ω=0.4), but not in experiments with lower pCO 2 , even though waters were slightly undersaturated with respect to S. spirorbis calcite (1200 µatm, Ω=0.8).In contrast, in the current study corrosion of S. spirorbis shell surfaces was common (>10% of the shells) when average seawater saturation state was below ∼0.9, indicating starting corrosion even at mildly undersaturated conditions.Shell corrosion increased with decreasing saturation when the seawater was undersaturated (Ω<1; Fig. 10), but occurred in only a few experiments for Ω>1.It was completely absent at Ω>2.
The S. spirorbis tubes in our experiments may have been more susceptible to shell corrosion compared to those of Saderne and Wahl (2013) for several reasons.First, the duration of the benthocosm experiments was much longer (>70 days) than the laboratory experiments (30 days).Second, during our experiments saturation state in the benthocosms fluctuated strongly between day and night (Fig. 3).Even with a mean saturation of 1 the shells may have been exposed to strongly undersaturated water during night time.Saderne and Wahl (2013) did not record pH or Ω on diurnal timescales, but the low biomass (1 g per 0.6 L flask) and constant vigorous gas bubbling most likely prohibited strong diurnal fluctuations of saturation states in their experiments.Third, with the more natural conditions in the benthocosms (unfiltered seawater, presence of natural fauna and flora) bioerosion by microbes may have fostered corrosion of the shells.Therefore, our experiments indicate that under natural conditions S. spirorbis can be significantly affected by shell corrosion at acidification levels expected for the end of the century.
Corrosion in the mostly under-saturated waters of all autumn and the high-CO 2 winter experiments (t Ω<1 >70%; Tab. 1, Fig. 10) was likely induced by mineral dissolution.In contrast, during the summer experiment when waters were mostly oversaturated (t Ω<1 <50%) the tubes were affected by bioerosion.Boring organisms play an important role in the ecology of many marine habitats (Warme, 1977).Microborings were observed in a S. spirorbis shell from the summer control experiment.
They probably affected the stability of the worm tube (Fig. 11).The few tubes recovered from residual Fucus thalli in the summer experiments with elevated temperatures (+T and +CO 2 +T) were mostly broken and strongly corroded (Fig. 16).These observations hint at a detrimental influence of elevated summer temperatures on S. spirorbis shells, either directly by affecting the worm's metabolism or indirectly through the reduction of grazing organisms (Werner et al., 2016) and increased antifouling activities of the Fucus host plants (Raddatz et al., 2017).Additionally, irreversible damage of Fucus algae at high summer temperatures (>27 • C, Graiff et al., 2015a) leads to substratum loss for S. spirorbis, which preferentially settle on Fucus (De Silva, 1962;O'Connor and Lamont, 1978).
S. spirorbis tubes that grow during the warm season might be especially susceptible to mechanical stress and bioerosion.As shown in Section 3.5, summer tubes were significantly lighter than expected for their size.This indicates thinner and/or less dense tube walls compared to autumn and spring specimens.With the increasing frequency and duration of summer heat waves in Central Europe predicted for the 21 st century (Beniston et al., 2007;Gräwe et al., 2013), increased bioerosion and loss of substratum can severely affect future S. spirorbis populations in the Baltic Sea.

Growth rate
Tube growth rates of S. spirorbis in our experiments were strongly controlled by the ontogenetic development.Growth rates were highest for juveniles and decreased when the worms got older and the tubes reached the maximum diameter range (Fig. 12).Similar growth-age relationships were found previously for S. spirorbis and other serpulid worms (O'Connor and Lamont, 1978;Kupriyanova et al., 2001;Riedi, 2012).
However, if only "large" specimens of similar initial sizes are considered, S. spirorbis tubes grew more rapidly in autumn and winter (October -March) than in spring (April -June), with the slowest growth occurring in the "large" summer population (July -September, Fig. 12).We know of no comparable published data.Reports of more rapid tube growth in summer compared to winter for several temperate serpulid species (Riedi, 2012;Kupriyanova et al., 2001) usually refer to the ontogenetic effect described above, i.e. enhanced growth of juvenile serpulids.The enhanced cold season growth of "large" S. spirorbis in our experiments was quite unexpected.Water was frequently under-saturated with respect to Mg-calcite during autumn and winter (48 to 98 % of experimental time, compared to 9 to 43 % in spring and summer, Table 1, Fig. 3).Other factors may play a role, like food availability and salinity.Food supply is generally lower in winter than during the warm seasons when increased light availability promotes phytoplankton growth.It consequently provides no explanation for the observed enhanced cold season growth of "large" S. spirorbis.Salinity was high during November-January and lowest in June.Spring salinities between 10 and 15 psu contrasted with autumn and winter values between 16 and 21 psu (Fig. S2).Enhanced calcification at higher salinities was previously observed in Baltic bivalves (Hiebenthal et al., 2012) and may potentially provide an explanation for enhanced growth of "large" S. spirorbis during the autumn and winter experiments.No significant treatment effects on growth were detected (see below).Consequently, the cold season growth enhancement of "large" S. spirorbis is not an artefact of increased temperatures and pCO 2 levels in the benthocosms.We suggest that, in addition to possible salinity effects, the reduced growth of adult specimens during the warm seasons reflects enhanced reproductive activity during this time (Knight-Jones and Knight-Jones, 1977;Seed et al., 1981), re-allocating energy resources from calcification to reproduction and thus reducing tube growth capacities.
The influence of increased pCO 2 on the growth of S. spirorbis worm tubes was previously studied in the experiments of Saderne and Wahl (2013).A significant growth rate reduction was only observed for "large" specimens at the highest pCO 2 (3150 µatm, Ω=0.4).No significant growth rate reduction was found at the intermediate pCO 2 level of 1200 µatm.In agreement with these results we found no significant change in tube growth parameters (Gr, Gr/D i , D f /D i ) when elevating pCO 2 from ambient levels to 1100 µatm (Fig. 14), corresponding to average saturation values as low as Ω=0.6 (Tab.1).We detected no significant impact of saturation state on growth (tube length or diameter) in any season.However, the average weights of newly grown tubes correlated with saturation states in the spring and autumn experiments (Fig. 13).
Apparently, the Baltic S. spirorbis worms are able to build their tubes with little changes in extension rates at pCO 2 levels as high as 1100-1200 µatm.Notably, these pCO 2 values are in the range of their natural habitats (385 to 2500 µatm; Thomsen et al., 2010Thomsen et al., , 2013;;Wahl et al., 2015).However, the tubes that are formed at lower CaCO 3 saturation may be more fragile.This is in line with results from cultured juvenile worm tubes of the tropical serpulid species Hydroides elegans, which showed reductions in shell hardness and wall thickness at lowered pH and CaCO 3 saturation states (Chan et al., 2012(Chan et al., , 2013;;Li et al., 2014).
As discussed in Section 4.4 high temperatures in the +T and +CO 2 +T treatments of the summer experiment (average T of 24 • C, Tab. 1) led to high mortality and strongly reduced growth of S. spirorbis tubes (Fig. 16).The only other significant temperature influence on growth was found in the juvenile populations of the autumn experiment.The higher temperature in the +T treatment induced higher growth rates of the juvenile populations compared to the control treatment (Fig. 14).There was, however, no significant temperature influence on the growth parameters at elevated pCO 2 (+CO 2 +T treatment), possibly indicating interactions between the effects of temperature and pCO 2 on growth.

Conclusions
The results of our benthic mesocosm experiments clearly demonstrate that the growth of S. spirorbis tubes is predominantly controlled by ontogenesis.Elevated pCO 2 levels, lowered pH and calcium carbonate saturation states expected for the end of the 21 st century had no significant impact on tube extension rates.Rather, S. spirorbis is capable of calcifying in water under-saturated with respect to the Mg-calcite of its shell.New tube parts were observed to be formed in under-saturated water when at the same time parts of the older tube were being corroded (Fig. 17).This is clear evidence for a strict biological biomineralization control of S. spirorbis.
Opposed to the batch culture experiments of Saderne and Wahl (2013) significant shell corrosion occurred in our experiments at a pCO 2 of 1100 µatm.While acidification had no impact on shell extension, shell corrosion increased with progressing acidification and under-saturation.Additionally, increased bioerosion, reduced growth, and loss of substratum occurred at high summer temperatures.Most S. spirorbis were not able to survive at a mean temperature of 24 • C in the benthocosms.On the other hand, among the juvenile populations of the autumn experiment, elevated temperatures (15 • C) increased tube growth rates, but only under ambient pCO 2 conditions.
We conclude that under continued warming and ocean acidification, with conditions expected for the end of the 21 st century, S. spirorbis in the Baltic Sea could be seriously affected by high summer temperatures and by enhanced dissolution and     +T and control treatments.The effect is only significant in the "small" sub-population, while the "big" sub-population showed no significant temperature effect.However, the "small" sub-population dominates the autumn population.Thus the total population shows a significant temperature effect.In summer, no tubes were recovered from the elevated temperature treatments.Results from three-way ANOVA and Tukey's HSD tests; * significant difference (p <0.05).
Average diurnal amplitudes of saturation state, pH and temperature were calculated as follows: The pH and temperature time-series fromWahl et al. (2015) and the resulting saturation values have an interpolated resolution of 10 minutes.We averaged all values of the period of interest into 24 one-hour bins, resulting in a mean value for each hour of the day.The minimum and maximum values of the resulting mean diurnal cycle define the mean diurnal amplitude.The resulting values were averaged for each of the four different treatments.Each experimental period was subdivided into four sub-periods with durations of 17-19 days and mean diurnal amplitudes of each sub-period were calculated as explained above.

Figure 1 .Figure 2 .
Figure 1. A. Two subunits of the Kiel Outdoor Benthocosm with open hood.Subunits with closed hoods are visible in the background on the right.B. Spirorbis spirorbis specimens attached to living brown alga Fucus vesiculosus.Juvenile (white dots, yellow arrows) and adult (white spires, red arrows) specimens of S. spirorbis are visible.

Figure 3 .Figure 4 .Figure 5 Figure 6 .Figure 7 .Figure 8 .Figure 9 .Figure 10 .
Figure 3. Average water temperature, daily insolation, pH and saturation state with respect to aragonite (as proxy for S. spirorbis Mg-calcite) in the four different treatments.Each of the four seasonal experiments is divided into four sub-periods lasting 17-19 days (start and end dates indicated at x-axis).Error bars indicate minimum and maximum values of the mean diurnal cycle during the sub-periods, except for insolation where they indicate day-to-day variability (standard deviation).Insolation was measured at the GEOMAR meteorological observatory (www.geomar.de/service/wetter),about 100 m from the benthocosms.

Figure 11 .Figure 12 .Figure 13 .Figure 14 .
Figure 11.SEM image (BEI) of polished cross section of S. spirorbis shell from summer control experiment.Dark spots are microborings mostly affecting the outer tube wall.

Figure 15 .
Figure 15.Size and growth of juvenile populations in control treatments of summer and autumn experiments.Two-way ANOVA and Tukey's HSD tests; * significant difference (p <0.05).

Figure 17 .
Figure 17.Synchronous calcification and shell corrosion in the under-saturated waters (mean Ω<0.8) of the autumn experiment.Length of arrows indicates relative magnitudes of treatment influences on shell calcification (left, solid arrows) and shell corrosion (right, dashed arrows), respectively.Shell growth showed little variability between treatments, except for increased growth in the +T treatment.Corrosion was strongly increased in the high-CO2 treatments and slightly enhanced in the +T treatment.DIC: dissolved inorganic carbon.
provided invaluable help for collecting, handling and studying Spirorbis.Stefan Krause, Mario Thöner and Manuela Goos helped with preparation and imaging of polished samples.Mark Lenz and Stephanie Schurigt kindly helped with the intricacies of statistical analyses.Iris Schmiedinger was of invaluable help in the laboratory during water chemistry analyses.This study was funded by the collaborative project BIOACID Phase II of the German Federal Ministry of Education and Research (BMBF; FKZ 03F0655F) and by Leibniz IOW.Many thanks to the reviewers V.B.S.C. Chan and J. Bijma for their critical and constructive comments.

Table 1 .
Average water data of the four treatments in the four seasonal experiments.
Mean values for temperature, pH, saturation (Ω) and percent of experimental time when basins were undersaturated with respect to aragonite and Mg-calcite (t Ω<1 ).Columns show mean values for single basins (A1, A2, B1, B2, etc.) and averages for each treatment with ±1sd ranges.a: pH data only for last 2 weeks of experiment; b: no pH data recorded; c: no data recorded; d: data only for last 4 weeks of experiment.Data excluded from mean.
No specimens were recovered from elevated temperature treatments in summer

Table 3 .
Mean size and growth parameters of juvenile S. spirorbis populations in summer and autumn.Selected sub-populations with homogenous initial diameter range (similar median Di) from summer and autumn control and +CO2 treatments.Values are averages and standard deviations.Di: initial diameter, Df: final diameter, Gr: growth, TbWd: tube width.a: insufficient data to select sub-population with Di similar to other treatments.