Plasticity in dormancy behaviour of Calanoides acutus in Antarctic coastal waters

Pond, D. Plasticity in dormancy behaviour of Calanoides acutus in Antarctic coastal waters. 77: Copepods that enter dormancy, such as Calanoides acutus , are key primary consumers in Southern Ocean food webs where they convert a portion of the seasonal phytoplankton biomass into a longer-term energetic and physiological resource as wax ester (WE) reserves. We stud-ied the seasonal abundance and lipid proﬁles of pre-adult and adult C. acutus in relation to phytoplankton dynamics on the Western Antarctic Peninsula. Initiation of dormancy occurred when WE unsaturation was relatively high, and chlorophyll a (Chl a ) concentrations, predominantly attributable to diatoms, were reducing. Declines in WE unsaturation during the winter may act as a dormancy timing mechanism with increased Chl a concentrations likely to promote sedimentation that results in a teleconnection between the surface and deep water inducing ascent. A late summer diatom bloom was linked to early dormancy termination of females and a second spawning event. The frequency and duration of high biomass phytoplankton blooms may have consequences for the lifespan of the iteroparous C. acutus females (either 1 or 2years) if limited by a total of two main spawning events. Late summer recruits, generated by a second spawning event, likely beneﬁtted from lower predation and high phytoplankton food availability. The ﬂexibility of copepods to modulate their life-cycle strategy in response to bottom-up and top-down conditions enables individuals to optimize their probability of reproductive success in the very variable environment prevalent in the Southern Ocean.


Introduction
Lipid-rich copepods are important conduits of carbon flow from the base of the marine food web to higher trophic levels and support fish, mammal and seabird communities (Pervushin, 1968;Hopkins and Torres, 1989;Voronina, 1998). At high latitudes, primary production is strongly influenced by the availability of light resulting in distinct phytoplankton bloom cycles and a relatively short productive season (Ma et al., 2014). As the photoperiod increases after the winter minimum, the annual onset of stratification initiates the highly productive phytoplankton "Spring" bloom (Sverdrup, 1953;Huisman et al., 1999). The life cycles of copepods are strongly affected by this distinct seasonality, and they have developed specific adaptations to take advantage of short-term food availability to survive long periods of food scarcity. Calanoid copepods are an ecologically important order of marine copepods and several species, particularly those in the Calanidae and Eucalanidae families (Baumgartner and Tarrant, 2017), undergo ontogenetic vertical migration from relatively shallow to deep water where they spend a large proportion of their life cycle in dormancy (Hagen et al. 1993;Sartoris et al., 2010), a resting stage with reduced metabolism and swimming activity (Maps et al., 2014). To outlast the winter months, these copepods accumulate large lipid stores within a membranebound oil sac that can occupy a large part of the body cavity (Lee et al., 2006). Not only does this lipid store have to provide fuel for metabolic processes during dormancy, but also it must provide the energy needed to re-ascend to the surface, continue development to adulthood, and fuel early egg production . As in other oceans, copepods dominate the mesozooplankton across most of the Southern Ocean, in terms of biomass, abundance, grazing activity, and secondary production (Atkinson et al., 2012b). Calanoides acutus is a major contributor to zooplankton biomass (Shreeve et al., 2005;Marrari et al., 2011) and an abundant species of herbivorous copepod that spends a large proportion of their life cycle (up to 7-9 months) in dormancy (Hagen et al., 1993;Tarling et al., 2004;Atkinson et al., 2012b). Mating occurs in deep water during late winter and the males perish shortly after while the females migrate to surface waters to feed on the phytoplankton bloom and begin spawning at the start of spring/summer. Nauplii and early copepodites develop in the euphotic zone until vertical migration to deeper water of pre-adult (CIV and CV) and adult (CVI) stages at the end of summer when feeding in surface waters is terminated (Atkinson et al., 1997;Tarling et al., 2004). Little is known about the life span of overwintered late stage copepodites with some studies suggesting a 1 year life cycle (Marin, 1988;Atkinson et al., 1997) whilst others suggest that individuals may re-enter dormancy and survive an extra year Hagen and Schnack-Schiel, 1996).
Prior to dormancy, overwintering stages concentrate phytoplankton lipids and accumulate large stores of total lipid (TL), mainly as wax esters (WEs). High concentrations of TL can be accumulated per individual (>500 mg) representing an energy-rich food source for higher trophic levels (such as fish) and a large reservoir of essential fatty acids (FAs) such as 20:5(n-3) (eicosapentaenoic acid, EPA) and 22:6(n-3) (docosahexaenoic acid). This accumulation of lipids represents the long-term storage of shortterm phytoplankton production. The presence or absence of copepod species that enter dormancy plays a fundamental role on a global scale in determining whether or not a region supports a lipid-rich food web (Record et al., 2018). The timing and duration of dormancy has implications for carbon flow as ontogenetic vertical migration events (VMEs) dictate time periods when stored lipids (carbon) are either available in surface waters or sequestered to the deep ocean due to respiration during dormancy (Jó nasdó ttir et al., 2015). Potential triggers influencing vertical distribution patterns have been identified, such as lipid accumulation above a threshold level (Rey-Rassat et al., 2002;Maps et al., 2012) and utilization during dormancy (Johnson et al., 2008); however, cues remain poorly understood and seasonal datasets are scarce. Following the development of a population over an entire life cycle could provide a better understanding of how these mechanisms influence behaviour. Pond et al. (2012) showed that a critical minimum threshold of $50% WE unsaturation within the copepod is important for dormancy initiation. The composition of these lipids, and not the bulk amount, then provides the ability of the WE store to shift from a liquid to a solid phase allowing dormant copepods to become neutrally buoyant in cold deep water and conserve energy (Visser and Jó nasdó ttir, 1999;Lee et al., 2006;Pond and Tarling, 2011). Although copepod species that enter a period of dormancy have evolved in the open ocean, where such a strategy could provide a key evolutive advantage, genetic programming of individuals that inhabit relatively shallow coastal and shelf sea environments is likely to result in the accumulation of large WE reserves with high levels of unsaturation (Falk-Petersen et al., 2009;Clark et al., 2012). As the FA component of WE is mainly derived from the diet, a suitable food source is critical. High proportions of FA 16:1(n-7), 20:5(n-3), 18:4(n-3), and 22:6(n-3) in storage lipids indicate the importance of diatoms and dinoflagellates in the zooplankton diet Falk-Petersen et al., 2000;Budge et al., 2006). Diatoms often dominate the biomass of phytoplankton blooms in the Southern Ocean (Ducklow et al., 2012) and are an important source of primary nutrition as they produce long-chain polyunsaturated FAs (PUFAs) [such as 20:5(n-3)] in abundance (Kattner and Hagen, 2009). The timing and duration of blooms varies spatially and temporally, as a result of physicochemical factors, and could help explain why the timing of descent in dormancy inducing copepods can be so variable between regions and years (Heath et al., 2004;Johnson et al., 2008;Pepin and Head, 2009). More research is required to better understand the trophic transfer of lipids, particularly in a time of global climate change shown to affect the composition, timing, and magnitude of phytoplankton blooms (Smetacek and Nicol, 2005;Sommer and Lengfellner, 2008;Rozema et al., 2017). These changes are expected to have major implications for the capacity of copepods to undertake their seasonal life cycles successfully (Pond et al., 2014). Understanding population dynamics is fundamental to predict how copepods might respond to future climate change and how Antarctic ecosystems may be influenced by bottom-up forcing.
In this study, in Ryder Bay on the Western Antarctic Peninsula, the lipid content and composition of C. acutus individuals (the pre-adult CV and adult CVI) were analysed over a "summer (S1)-winter-summer (S2)" time series and the FA composition was determined to investigate links between the accumulation and composition of WE and diet. Research in the Southern Ocean often focuses on copepods inhabiting the deep-water open ocean (>500-1000 m depth); however, many populations successfully overwinter in relatively shallow coastal and shelf sea environments (<500 m depth, Clark et al., 2012). Ryder Bay is a marginal habitat that is relatively advection free due to localized gyre-like circulation features (Beardsley et al., 2004;Moffat et al., 2008;personal observation), which promotes the retention of phytoplankton and zooplankton populations. This study, which repeatedly sampled the same population in the same local area, is one of very few that presents complete seasonal sampling, including winter and the transition periods between winter and the productive summer, and provides valuable information to better understand the behaviour of copepods in a coastal environment.
Seawater samples for chlorophyll a (Chl a) concentration and taxonomic composition analysis were collected from the standard monitoring depth of 15 m by using a 12-l Niskin bottle deployed from a small boat. One to eight litres were filtered over GF/F glass fibre filters (47 mm, Whatman, Eindhoven, The Netherlands), after which the filters were carefully folded and wrapped in aluminium foil, snap-frozen in liquid nitrogen, and stored at À80 C until analysis in the home laboratory using high-performance liquid chromatography (HPLC) and chemical taxonomy analysis [CHEMTAX, see Biggs et al. (2019) for methodological details].
Calanoides acutus samples were collected from two depth profiles, i.e. 500-200 m (deep) and 200-0 m (shallow), to separate the overwintering "winter" population from the active "summer" population (Huntley and Escritor, 1991;Schnack-Schiel et al., 1991). A 200-mm mesh ring net (0.26 m 2 opening) was used to obtain population abundance counts (copepodite stages CI-CVI) in shallow water, and a 500-mm mesh ring net (0.26 m 2 opening equipped with a double release mechanism) was used to obtain CV (pre-adult) and CVI (adult males and females) individuals from both shallow and deep water. Sampling of copepods occurred weekly (weather dependent) during summer 1 (S1, 23 November 2012-18 April 2013) and summer 2 (S2, 14 November 2013-21 February 2014). To include community dynamics year round, sampling was performed on three occasions during "winter". Zooplankton were kept in a plastic portable cooler and transported back to the laboratory within 2 h of capture. Calanoides acutus in good condition were quickly sorted for lipid analysis using a binocular microscope and the remainder (of the net haul) preserved in 200 ml of formaldehyde (5% final concentration), and stored at 4 C. Formaldehyde-preserved zooplankton samples were split (1/3) using a plankton splitter, and abundances were determined under a binocular microscope.

Lipid analysis
A total of 331 samples of C. acutus were collected for lipid analysis consisting of 1635 individuals. Lipids were extracted from stage CV and CVI (male and female) following Pond and Tarling (2011). Individuals and bulk samples were initially transferred to 1.1-ml tapered vials (Chromacol) containing 500 ml of chloroform:methanol (2:1 v:v) and stored at À80 C. After transport to the home laboratory, the solvent containing each C. acutus sample was pipetted into 8-ml glass vials using glass pipettes. The 1.1ml vials (containing the exoskeleton) were briefly re-extracted and vortexed a further three times with 1 ml chloroform:methanol (2:1 v:v), and the final volume was adjusted to 4 ml. After the addition of 1 ml of potassium chloride (0.88% w:v), samples were vortexed and centrifuged for 2 min at 400 Â g to promote phase separation. The lower chloroform phase, containing the TL extract, was removed using Hamilton glass syringes containing a Teflon tipped plunger, into pre-weighed 4-ml glass vials before being evaporated under nitrogen and stored in a vacuum desiccator overnight prior to reweighing.

Lipid class analysis
Aliquots of TL (10 mg) were subjected to high-performance thinlayer chromatography (HPTLC) using a hexane:4 diethyl ether:acetic acid (90:10:1) solvent system (Pond and Tarling, 2011). The plates were sprayed with 8% (v:v) phosphoric acid containing 3% (w:v) copper acetate solution, followed by heating at 160 C for 13 min to char the lipid classes and create dark areas on the HPTLC plate. Lipid classes were then quantified by scanning densitometry (Shimadzu Dual-wavelength TLC Scanner, CS-930), the different lipid classes being identified by comparison with known standards (Pond et al., 1995). The degree of WE unsaturation was used to calculate an unsaturation index as described in Pond and Tarling (2011). In short, HPTLC separated the WEs into two bands: the upper band was rich in saturated fatty and monounsaturated FAs and the lower band was dominated by PUFAs. The unsaturation index was calculated by dividing the amount of PUFA WE by the total WE, providing an index of the degree of unsaturation ranging between 0 and 1.0 (Stevens et al., 2004) and presented as a percentage between 0 and 100%.

FA analysis of TL
Dried aliquots of 150 mg of TL were used for FA analysis. Samples were derivatized with 1 ml of 2% sulphuric acid-methanol (after the addition of standards: 5 mg 19:0 and 1 mg 12:0) and incubated for 4 h at 80 C. After cooling in water, 2 ml of Milli-Q water and 2 ml of hexane were added and vortexed and the upper layer was transferred to a 10-ml glass vial. Two millilitres of hexane was added once more and vortexed, and the upper layer was transferred to a 10-ml glass vial. The sample containing the FA methyl esters was transferred in 200 ml of hexane to a 2-ml vial and stored at À20 C until analysis on a ULTRA Trace gas chromatograph (GC). The GC was equipped with a BPX-70 column with hydrogen as the carrier gas.

Statistics
Comparisons of FA and WE unsaturation data were performed by linear regression in SigmaPlot V14.0 (Systat Software Inc., San Jose, CA, USA).

Results
Two main periods of phytoplankton accumulation occurred in S1 (2012-2013), 30 November-2 January and 11 February-15 April (Figure 2), as indicated by Chl a dynamics, which were dominated by diatoms (86 6 14% n ¼ 13 and 88 6 19% n ¼ 17, of total Chl a, respectively; Biggs et al., 2019). Female C. acutus are important for population growth (due to spawning) and contributed most across all time points to populations in both shallow (37 6 40%, n ¼ 32) and deep (61 6 33%, n ¼ 28) water. At the start of S1, females were the most abundant stage and only found in deep water (Figure 3a and b) until the abundance of females inhabiting shallow waters (shallow female) increased between 30 November and 12 December (Figure 3b), from 4 to 108 ind 100 m À3 , indicating the first ontogenetic VME (VME 1) of S1 and post-winter dormancy termination. The increase in shallow female abundance coincided with a decline in numbers of females inhabiting deep waters (deep females, from 44 to 4 ind 100 m À3 ) at the same time as rapidly increasing Chl a concentrations (from 0.5 to 3.5 mg l À1 , Figure 2). Following the decline in the initial phytoplankton bloom at the beginning of January in S1 ( Figure 2), shallow female numbers began to decline at the same time as a rise in the number of deep females (0-18 ind 100 m À3 , Figure 3a) and deep CVs (0-9 ind 100 m À3 , Figure 3c), indicating VME 2 (29 December to 17 January) and the descent of individuals to deep water (dormancy initiation). The number of deep females continued to increase until a steep decline between 20 February and 2 March (from 30 to 8 ind 100 m À3 , Figure 3a) alongside a brief increase in shallow female abundance (from 8 to 33 ind 100 m À3 , Figure 3b) and indicated a third VME (VME 3) and the second ascent of females in S1 (dormancy termination). Chl a concentrations had been increasing for 2 weeks prior to VME 3 (from 0.81 to 5.79 mg l À1 ) and a high biomass phytoplankton bloom (max ¼ 16 mg Chl a l À1 ) developed over a period of 2 months ( Figure 2).
Copepodite stage CV was, overall, the second most dominant life-cycle stage contributing on average 22 6 27% n ¼ 28 to total numbers. At the beginning of S1, CVs were absent from net hauls (Figure 3c and d) and numbers remained relatively low until a rapid increase in the abundance of CVs inhabiting shallow waters (shallow CVs) between 1 and 20 February (from 12 to 123 ind 100 m À3 , Figure 2d), 7-10 weeks after the peak in abundance of shallow females during VME 1. Numbers of shallow CVs remained high until 13 March ( Figure 3d) and declined at the same time as increasing numbers of CVs found in deep water (deep CVs, from 7 to 129 ind 100 m À3 , Figure 3c) indicating a fourth VME (VME 4, 13 March to 18 April) and the start of the CV overwintering period (dormancy initiation). Chl a concentrations during this late-season diatom bloom peaked on 12 March and declined thereafter, until low concentrations on 15 April ( Figure 2) indicated the end of the phytoplankton productive season.
At the beginning of the winter dormancy period (18 April), CVs contributed 88% to total population abundance ( Figure 4). A steady decline in the number of deep CVs was observed until 14 November (down to 0 ind 100 m À3 , Figure 3c), at the same time as a steady rise in the number of deep females (up to 39 ind 100 m À3 , Figure 3a), and indicates the maturation of copepodites from CV to CVI stages. The greatest change in the ratio of CV:CVI occurred between 16 May and 12 August (from 13 to 1) with relatively more males in deep nets on 12 August (30%) than females (18%, Figure 4). Male abundance remained high (and only observed in deep water) until 14 November and declined thereafter to zero by mid-December in S2 (Figure 3e). A 68% decline in population abundance over winter indicates that (in retrospect) each female at VME 1 needed to contribute a minimum of 3 (CV) individuals to the population at the start of the overwintering period (VME 4) to maintain population numbers (at VME 5 compared with VME 1) and offset losses during winter. This suggests that, generally, even low numbers of females can reconstitute the numbers required to ensure the population longterm viability.
At the beginning of S2 (2013-2014), phytoplankton biomass began to slowly increase during November and rapidly increased from 0.4 to 4.8 mg l À1 between 2 and 14 December (Figure 2). One extended period of increased Chl a concentrations followed (2 December-14 February) with the phytoplankton community continually dominated by diatoms (83 6 15%, of Chl a, n ¼ 32; Biggs et al., 2019]. Between 14 November and 14 December, a sharp decline in deep females (39-2 ind 100 m À3 , Figure 3a) and an increase in shallow females (0-38 ind 100 m À3 , Figure 3b) indicated a fifth VME (VME 5) and post-winter dormancy termination. Shallow female abundance peaked on 17 January (96 ind 100 m À3 , Figure 3b) and declined until the end of S2 (41 ind 100 m À3 ); however, the number of deep females remained low until an increase between 30 January and 14 February (from 1 to 18 ind 100 m À3 , Figure 3a) at the same time as declining surface Chl a concentrations (from 5.6 to 0.7 mg l À1 between 12 and 17 February, Figure 2). This indicates a sixth VME (VME 6, 30 January-14 February) and the descent of females at the end of S2 (dormancy initiation). CVs were absent during S2 until an increase in shallow CVs on 17 January (20 ind 100 m À3 ; Figure 3d) that continued until the end of the season (52 ind 100 m À3 by 21 February).
When examining the population dynamics of all C. acutus copepodite stages (CI-CVI), a period of increased abundance occurred in S1 between 17 January and 2 March 2013 (304 6 144 ind 100 m À3 , n ¼ 7, Figure 3f). At the initial increase on 17 January (333 ind 100 m À3 ), a rapid rise in the share of CIs was observed (59%, Figure 4) occurring 5 weeks after the peak in  shallow females at the beginning of S1 (Figure 3b). A second rapid increase in the share of CIs occurred on 13 March (40% of total, Figure 4) when total abundance was lower (97 ind 100 m À3 ) and 3 weeks after the second peak in numbers of shallow females (33 ind 100 m À3 , Figure 3) on 20 February. This indicates two rounds of spawning occurred during S1. During S2, population abundance remained low until a sharp increase on 14 February 2014 (1291 ind 100 m À3 , Figure 3f). Only one sharp rise in CI abundance was captured in the dataset during S2 (Figure 4) with 3106 CI-CIV 100 m À3 on 14 February in S2 (data not shown) appearing 1 month later than the initial CI increase in S1.
The TL content of females at the beginning of S1 (23 November) was 112 mg ind À1 (Figure 6a) and 28% unsaturation of WE (Figure 7a). TL content and WE unsaturation separated by depth and stage are available in Supplementary Online Resources 2 and 3. During dormancy termination (VME 1, 30 November-12 December), female TL increased slightly to 136.8 mg ind À1 and remained relatively constant throughout (131.5-141.4 mg ind À1 , Figure 6a); however, WE unsaturation declined from 31 to 22% (Figure 7a). During VME 2 (dormancy initiation), female TL increased from 124 to 350 mg ind À1 (Figure 6a) and WE unsaturation increased from 33 to 49% (Figure 7a). At the beginning of VME 3 (second diapause termination in S1), female TL peaked at 419 mg ind À1 on 22 February and declined to 284 mg ind À1 at the end (2 March, Figure 6a), and female WE unsaturation declined from 66 to 32% (20 February-3 March, Figure 7a). This suggests lipid utilization by females during VME 3. Although abundances were low, the highest TL content of females during S1 occurred at the start of VME 4 (13 March, 426 mg ind À1 , Figure 6a); however, unsaturation continued to increase until the end (57% on 18 April, Figure 7a) at the same time as peak unsaturation of CV WE (71%, Figure 7b).
Similar to females, CV TL content ( Figure 6b) and WE unsaturation ( Figure 7b) were low (12 mg ind À1 and 11%, respectively) at the beginning of S1 (similar to abundances) and increased to 254 mg ind À1 on 17 January (Figure 6b) and 61% on 11 January (Figure 7b). A decline in CV TL followed between 17 January and 2 March (down to 78 mg ind À1 ) and coincided with the     appearance of greater numbers of CIII and CIV stages; however, WE unsaturation remained relatively high (>51%) during this time (Figure 7b). WE unsaturation of both CVs and females peaked on 18 April (71 and 57%, respectively, Figure 7b) at the end of VME 4; however, the TL content of CVs continued to increase until 16 May (to 270 mg ind À1 , Figure 6b), at the same time as a quantitatively similar rise in female TL (Figure 6a). Prior to overwintering, female unsaturation increased to 57% on 18 April and declined to 32% on 14 November, a 25% decrease during dormancy (Figure 7a). Although female TL (in S1) peaked on 13 March at the start of VME 4 (426 mg ind À1 , Figure 6a), TL was lower at the end of VME 4 on 18 April (347 mg ind À1 ), increased to 385 mg ind À1 on 16 May, and declined to 212 mg ind À1 on 14 November (start of VME 5). This corresponds to a decrease of 174 mg TL ind À1 (16 May to 14 November) and 130 mg ind À1 of PUWE. CVs were absent in net hauls on 14 November; therefore, winter CV lipid utilization estimates cannot be calculated. At the end of the main overwintering period, male numbers were still relatively high (9 ind 100 m À3 on 14 November, Figure 3e); however, their TL content (27 mg ind À1 , Supplementary Online Resource 2e) and WE unsaturation (12%, Supplementary Online Resource 3e) were low and remained low (TL 20-30 mg ind À1 , unsaturation 21-25%) until male abundance declined to zero after 7 December (Figure 3e). Females dominated the population at the beginning of S2 and, similar to S1, TL remained relatively constant during dormancy termination (VME 5, 173-229 mg ind À1 , Figure 6a), unlike WE unsaturation that declined from 32 to 21% (Figure 7a). The TL content of females at the start of S2 (VME 5) was higher than at the start of S1 (VME 1, 137-132 mg, ind À1 ); however, declines in WE unsaturation were highly similar between seasons (from 31 to 22% in VME 1). At the start of VME 6 (female dormancy initiation in S2), the TL content of females had rapidly increased to 427 mg ind À1 with 64% unsaturation (Figures 6a and 7a, respectively). Although female WE unsaturation of 49% at VME 2 ( Figure 7a) suggests this level was sufficient for dormancy, the second high biomass diatom bloom of S1 (Figure 2) allowed females to increase their TL content and unsaturation at the end of the productive season

Chl a and VMEs
Although patchiness in time and space may theoretically have influenced the results, duplicate net hauls were comparable (slope of linear regression ¼ 0.92, r 2 ¼ 0.84, p < 0.0001, Supplementary Online Resource 4) and deep mixed layer depths (MLD) due to storm events were rare (maximum MLD of 112 m during winter 2013, data not shown) indicating that mixing events did not affect the vertical distribution of individuals (Clarke et al., 2008;Dong et al., 2008). In this study, the timing of ontogenetic VMEs suggests that dormancy was initiated by females as soon as sufficient stores of TL and PUWE were accumulated. This could have consequences for predators of copepods in surface waters as lipid-rich females were only available for $2 weeks during the first half of January in S1 and 2-4 weeks in S2 (from mid-January to mid-February), after which females migrated to below the base of the thermocline ($200 m) and a portion of carbon stored as lipids was sequestered to the deep ocean as a consequence of respiration during dormancy (Jó nasdó ttir et al., 2015).
VMEs were linked to the seasonal dynamics of Chl a, i.e. the frequency and duration of phytoplankton blooms. Dormancy termination was timed to coincide with high biomass blooms that consisted both of micro-sized (VME 1) and nano-sized (VME 3 and 5) diatoms (Biggs et al., 2019). Calanoides acutus is primarily herbivorous with a short reproductive period that was timed to coincide with high phytoplankton food availability (Atkinson, 1998;Pasternak and Schnack-Schiel, 2001a). Furthermore, the second ascent of females in S1 (VME 3) indicated that the high biomass diatom bloom provided a strong cue for ascent (Dezutter et al., 2019), even though female levels of TL (364-419 mg l À1 ) and unsaturation (66-48%) were still high, and only 34-64 days (VME 2-3) after the initial descent (VME 2). VME 3 represented the mass movement of carbon from deep to shallow water and may have allowed surface-dwelling predators access to an additional peak in the abundance of lipid-rich copepods. Prior to dormancy initiation (VME 2, 4 and 6), it is likely that individuals can directly sense reducing/low phytoplankton food concentrations (Bautista and Harris, 1992;Perissinotto, 1992;Atkinson, 1994;Pasternak and Schnack-Schiel, 2001a;Garrido et al., 2013); however, at dormancy depths (500-200 m), signals of phytoplankton biomass (at the surface) must be transmitted through the water column and are likely related to sedimentation of organic matter (Annett et al., 2010;Ducklow et al., 2012;Turner, 2015). During winter, when Chl a concentrations are low, increased proportions of smaller flagellated phytoplankton are observed (Rozema et al., 2017) and the shift to a high biomass, diatom-dominated period (at VME 1 and 5) most likely promoted a more herbivorous diet with reduced coprophagy/coprorhexy (Pasternak and Schnack-Schiel, 2001a, b;Turner, 2002Turner, , 2015 and (at the same time as increased stratification during VME 1 and 5; Biggs et al., 2019) increased the load and sinking speed of faecal pellets (FPs) in the water column. Similarly, a relatively low biomass, flagellate-dominated period (Biggs et al., 2019) was observed prior to the second high biomass diatom bloom in S1, and the second ascent of females (VME 3). FPs produced from flagellated phytoplankton diets have been observed to sink almost ten times slower than diatom fed pellets (Ploug et al., 2008), likely related to mineral ballasting by biogenic silica (Voss, 1991). Diatoms also produce transparent exopolymer particles (TEPs) in abundance and are a major component of diatom aggregates (Passow, 2002). TEPs production combined with organic material released from melting sea ice (and increased stratification) may produce further deposition resulting in a rain of organic material at depth. Rapidly sinking FPs (>300 m d À1 ) (Ploug et al., 2008;Atkinson et al., 2012a) would enable organic material to reach the overwintering population of C. acutus (500-200 m) within 1-2 days and pigment degradation products, characteristic of the phytoplankton community, can be detected in FPs (Nelson, 1989). Microbial degradation of sedimenting organic matter may leak solutes into the surrounding water (Turner, 2015), which, through chemosensory mechanisms and probably hormonally mediated, may trigger the ascent (Irigoien, 2004).

Losses
Declines in abundance over the winter period (a 68% reduction between April and November 2013, VME 4-5, Supplementary Online Resource 5) suggest the 64 ind 100 m À3 at the start of S1 (total population abundance at VME 1) each needed to contribute $3 offspring (total of 192 ind 100 m À3 required on 18 April 2013) to balance losses during winter. At dormancy initiation (VME 4), population numbers were 48% less (100 ind 100 m À3 captured on 18 April 2013) representing (in retrospect) a minimum total requirement of $6 ind female À1 , over S1 and winter combined, to maintain the population over an annual cycle. This indicates during S1 either loss rates were twice as high (to maintain the population) or egg production twice as low. Low egg production may be related to relatively lower TL stores at the start of S1, compared with S2 (VME 1, 137 6 5 mg ind À1 , n ¼ 3; VME 5, 200 6 25 mg ind À1 , n ¼ 4) as well as reduced CI-CIV recruitment (1266 ind 100 m À3 at initial CI peaks in S1 and 3106 ind 100 m À3 in S2). Whilst the large amounts of lipid and high unsaturation alone could have triggered entrance into dormancy at VME 2 and 6, lower TL and unsaturation were observed at VME 2 compared with VME 6 (Figures 6 and 7). Eggs, nauplii, copepodites, and potentially even adults of C. acutus are within the potential prey range for many of the large zooplankton species that were often observed during both seasons, such as Metridia gerlachei, Rhincalanus gigas, (Para)euchaeta antarctica, euphausids, polychaetes and arrow worms. The 1 month earlier increase in largersized zooplankton abundance (>200 mm) in S1 than in S2 (Supplementary Online Resource 6) matched the 1 month earlier descent of female C. acutus, which may imply that enhanced predation pressure affects the initial decision of females to descend. The combination of increased numbers of larger-sized zooplankton and reducing/low Chl a concentrations (January S1) may have resulted in even higher predation pressure on smaller zooplankton (including nauplii and copepodites) by larger omnivorous individuals (Pasternak and Schnack-Schiel, 2001a) and contribute to high losses over S1.
The nano-sized cells during the second high biomass diatom bloom in S1 (Biggs et al., 2019) likely represented a suitable prey size for copepodites (Perissinotto, 1992) as they mature and attempt to increase lipid stores at the end of the productive season; however, relatively low TL content of CVs was observed (270 mg ind À1 ) at diapause initiation (VME 4). The biomass dominant phytoplankton population during this bloom (11 mm Ø diatom) was subjected to high rates of viral lysis (Biggs et al., unpublished), which may have reduced food availability during a crucial time of lipid accumulation prior to diapause. Increased substrate availability (due to the viral shunt) could explain the S1 peak in bacteria abundance mid-March (Biggs et al., unpublished), which likely benefitted micro-zooplankton grazers (Azam et al., 1991). Although considered primarily herbivorous, low phytoplankton abundance could stimulate carnivorous grazing (Pasternak and Schnack-Schiel, 2001a) by late-season recruits and explain why CV TL increased between 18 April and 16 May (from 233 to 270 mg ind À1 ) whilst WE unsaturation declined (from 71 to 61%). Reduced lipid stores combined with a relatively shallow diapause depth (<500 m) could be related to high winter mortality as deeper depths may allow females to better take advantage of the lipid phase transition effects of a highly unsaturated ($50%) lipid store (Pond and Tarling, 2011). If neutral buoyancy cannot be achieved, then high rates of lipid utilization and mortality may be due to additional energetic costs associated with swimming to maintain an optimum position in the water column. At the same time, predation pressures may be higher in the top 500 m than below (Yamaguchi et al., 2004;Harper and Peck, 2016) and individuals close to the bottom depth, where phase transition of WEs may occur, would be more concentrated and exposed to the benthic community. The relatively high unsaturation of CVs (compared with females) at the time of overwintering (71-61%) and low TL (233-270 mg ind À1 representing 132-140 mg ind À1 of PUWE and $32-34 mg ind À1 EPA) suggests that increased unsaturation was prioritized when TL was increased. The data of Pond and Tarling (2011) also indicate that, at depths <500 m, WE phase transition occurred at higher temperatures and was more pronounced at higher levels of unsaturation (70%). The greater increase in the unsaturation of CV WE prior to diapause may have been a mechanism to compensate for lower TL stores and take advantage of energetic savings at higher temperatures as lipid stores become increasingly dense with depth.

Diapause mechanisms and life cycle
The level of unsaturation of the WE lipid store, rather than the total amount (TL), is potentially a key mechanism influencing the dormancy behaviour of calanoid copepods (Pond and Tarling, 2011;Pond et al., 2012). In our study, unsaturation levels peaked at the start of the main dormancy period (VME 4) and declined over the winter. Dynamics were determined by the retention of FA 20:1(n-9) and 22:1(n-11) (during winter) and the selective accumulation (during summer) and utilization (during winter) of 20:5(n-3) (EPA). Changes in unsaturation were often not mirrored by changes in TL indicating a more functional role of FA composition. EPA is a precursor of eicosanoids that are locally acting "tissue hormones" and may influence functions such as reproduction, ion, and water transport (Persson and Vrede, 2006). The WE of many herbivorous copepods is also characterized by considerable amounts of long-chain monounsaturated fatty alcohols [20:1(n-9) and 22:1(n-11)], which are not present in significant amounts in their phytoplankton diet (Albers et al., 1996;Lee et al., 2006;Pond et al., 2012). This biosynthesis of fatty alcohols and esterification with FA to WE resulted in rapid lipid accumulation since there is both de novo synthesis and incorporation of dietary lipids (Graeve et al., 2005). During increases in unsaturation (summer), stores of 20:1(n-9) and 22:1(n-11) (retained during winter) are likely reduced to fatty alcohols Pond et al., 2012) and esterified to 20:5(n-3) acquired from dietary sources. Unsaturation level of the entire lipid store was therefore determined by synergistic changes in "pools" of 20:1(n-9) þ 22:1(n-11) and 20:5(n-3), rather than unsaturation dynamics determined solely by 20:5(n-3). This action has a double impact on unsaturation levels (þ1 PUFA, À1 MUFA), as FAs are converted to fatty alcohols (and vice versa), likely providing a stable signal regarding changes in unsaturation state. The physiological impact of selective EPA utilization during dormancy may further act as a biological timer (Häfker et al., 2017(Häfker et al., , 2018 and, combined with increased buoyancy (due to reducing unsaturation and phase transition), could stimulate ascent. In relatively deep waters of the Scotia sea, late-stage C. acutus individuals that typically diapause between 500 and 1000 m or deeper were observed with a tri-modal vertical distribution at one site (<100, 400-600, and 800-1000 m) and concentrated in the 300-500 m depth range at another . One explanation for this is that individuals had ascended from deeper waters to an intermediate depth to await the initiation of the "Spring" bloom . This initial "intermediate" migration is potentially triggered by reduced levels of unsaturation; however, in coastal and relatively shallow shelf sea areas such as Ryder Bay, bottom depth (520 m) is similar to this intermediate depth; therefore, signals of reduced unsaturation and increased phytoplankton standing stock may combine to initiate ascent.
The late-season high biomass diatom bloom of S1 coincided with the second ascent of females (VME 3) at the same time as lipids appear to be utilized (TL decreased from 419 to 284 mg ind À1 and WE unsaturation decreased from 66 to 32%). The second S1 spawning event coincided with a decline in total female abundance over VME 3 (from 31 to 9 ind 100 m À3 , Supplementary Online Resource 5a) and suggests that the ascending females that spawned expired after a 1-year life cycle. In contrast during S2, a singular but prolonged high biomass bloom (Figure 2) resulted in a singular spawning event and a 1-month delay in dormancy initiation of (most likely "new") females. Chl a concentrations were relatively low between 21 February and 15 April post-S2 (1.5 6 0.7 mg l À1 , n ¼ 8), and these females likely remained in dormancy to complete a 2-year life cycle. Such a mechanism could explain why some studies suggest the overwintering stages of C. acutus complete their life cycle in 1 year (Marin, 1988;Quetin et al., 1996;Atkinson et al., 1997) and others that they re-enter diapause and survive an extra year (Hagen and Schnack-Schiel, 1996;Tarling et al., 2004). The lateseason phytoplankton bloom in S1 was associated with higher temperature (Biggs et al., 2019), similar to novel autumn blooms observed in the Arctic (Ardyna et al., 2014), and suggests that global warming could influence the lifespan of vertically migrating copepods and promote a 1-year life cycle, rather than a 2-year life cycle.
Although two spawning events were observed in S1, the abundance of CI-IV in shallow water on 2 March 2013 (initial peak of 531 ind 100 m À3 ) was similar to 17 January 2013 (735 ind 100 m À3 ). Maturing copepodites from the late-season spawning event would likely benefit from reduced predation pressure due to lower numbers of larger-sized zooplankton and an abundance of phytoplankton food (Pasternak and Schnack-Schiel, 2001a) and could explain the similarity in CI-CIV numbers, even though peak shallow female abundance was threefold higher during VME 1 (108 ind 100 m À3 ) than VME 3 (33 ind 100 m À3 ). Conversely, copepodites from the early-season spawning event would have matured during times of low Chl a (food availability) with increased proportions of cryptophytes (preference of diatoms as food type; Verity and Smayda, 1989;Head and Harris, 1994) and high numbers of larger-sized zooplankton (high predation pressure, Pasternak and Schnack-Schiel, 2001a). An early-season mismatch and late-season match, between peaks of Chl a, new C. acutus recruits and higher trophic predators (Durant et al., 2013), likely resulted in a relatively greater contribution of individuals from the late-season spawning event to annual reproductive success. This flexible dormancy strategy would enable females to maximize phytoplankton food availability and reproductive capability whilst reducing predation risk.
It is likely that multiple internal and external factors combine, such as a lipid-modulated endogenous clock (Johnson et al., 2008;Häfker et al., 2018), TL content (to support reproductive maturation and winter survival; Hagen and Schnack-Schiel, 1996;Rey-Rassat et al., 2002), WE unsaturation (buoyancy regulation; Pond et al., 2012), and food availability (Friedland et al., 2016) to serve as a timing mechanism to determine ontogenetic vertical migration behaviour.

Conclusions
The dormancy behaviour of C. acutus CV and CVI stages appears coupled to the frequency and duration of phytoplankton blooms. Dormancy termination post-winter coincided with reduced unsaturation ($30%) and increased phytoplankton standing stock. Once sufficient lipid stores were accumulated after the initial phytoplankton bloom period (350 mg TL ind À1 with $49% unsaturation), dormancy was initiated when Chl a concentrations declined. An early S1 increase in larger-sized zooplankton numbers (compared with S2) likely contributed to increased losses and early dormancy initiation. A late summer diatom bloom, driven by higher temperatures, stimulated ascent and spawning of dormant females for the second time in S1 and the resulting late-season copepodites would benefit from reduced predation (low numbers of larger-sized zooplankton) and high phytoplankton food availability, i.e. a match between peaks of Chl a, new C. acutus recruits, and higher trophic predators. The loss of females during dormancy termination (VME 3) suggests that individuals expire after two main "spawning events". Furthermore, the timing of these events is regulated by the frequency and duration of phytoplankton blooms, which have consequences for the lifespan of C. acutus females, either 1 or 2 years. This flexible strategy would enable females to maximize phytoplankton food availability and reduce the likelihood of predation, thereby increasing reproductive capability whilst promoting the survival of produced offspring via reduced mortality rates. As such, dormancy behaviour and copepod lifespan, and thus zooplankton's role in the biological carbon pump, are intimately linked to the structure and dynamics of the Southern Ocean food web.

Supplementary data
Supplementary material is available at the ICESJMS online version of the manuscript.