A mixture of innate cryoprotectants is key for freeze tolerance and cryopreservation of a drosophilid fly larva

Various organisms from bacteria to animals respond to environmental stressors such as heat, cold, freezing, drought, hypersalinity and high hydrostatic pressure by accumulating a stereotypic set of cytoprotective compounds including amino acids, sugars, polyols and methylamines (Somero, 1986; Yancey, 2005). This is also true for insects that naturally tolerate extracellular freezing during overwintering in temperate and polar habitats (Lee, 2010; Storey and Storey, 1988). In such insects, the accumulation of low molecular weight carbohydrate and free amino acid cryoprotectants (CPs) is considered a fundamental tenet of their freeze tolerance (Storey, 1997; Storey and Storey, 1988; Storey and Storey, 1991; Toxopeus and Sinclair, 2018). Extensive knowledge has accumulated over many decades about which CPs are present in many different species (Asahina, 1970; Salt, 1961; Sømme, 1982; Storey and Storey, 1991), but our understanding of how they protect the insect tissues and cells is far from complete (Toxopeus and Sinclair, 2018).

Here, we focus on an important but previously neglected aspect of natural freeze tolerance: the components of complex mixtures of multiple CPs may contribute to freeze tolerance via additive or even synergistic mechanisms (Storey and Storey, 1986; Storey and Storey, 1988; Toxopeus et al., 2019). It is well known from practice in clinical medicine and biotechnology that combinations of CPs protect the viability of cryopreserved cells or materials better than single substances (Elliott et al., 2017). The mixtures may even have emergent properties, i.e. those beyond the summation of properties of the individual components. Relatively recently, the existence of specific mixtures, so-called NADES (natural deep eutectic systems) in various organisms was proposed (Dai et al., 2013; Choi et al., 2011). These mixtures consist of stereotypic sets of natural primary metabolites such as sugars, sugar alcohols, organic acids, amino acids and amines, and are characterized by a much lower melting point than that of the individual components, mainly because of the formation of intermolecular hydrogen bonds (Dai et al., 2015). These mixtures have been proposed to serve various physiological functions, including survival of organisms in extreme drought and cold (Castro et al., 2018; Gertrudes et al., 2017; Liu et al., 2018), and have also been proposed as new agents for cell cryopreservation (Hornberger et al., 2021). We believe that there may be major practical consequences of exploring the complex composition of insect CP mixtures, as they could aid in the development of applicable cryogenic techniques. Freeze-tolerant insects also offer the potential for the discovery of novel components, for the study of interactions between individual components, and for the analysis of the principles underlying survival after freezing in a complex organism composed of different cell types and tissues; a complexity that still poses a challenge for medical cryopreservation (Fahy et al., 2006; Pegg, 2001).

It is reasonable to first analyze the suite of CPs accumulated by the most freeze-tolerant model species, and larvae of the malt fly, Chymomyza costata (Diptera: Drosophilidae), are among the most cold-hardy animals known (Des Marteaux et al., 2019). Overwintering cold-acclimated larvae in diapause can survive freezing of all osmotically active water (68% of total body water), down to −75°C, with survival rates similar to those in unfrozen controls, and they can even survive long-term (18 month) cryopreservation in liquid N₂ (Moon et al., 1996; Rozsypal et al., 2018; Shimada and Riihimaa, 1988). We have previously shown that high concentrations of the free amino acid proline acquired by diapausing C. costata larvae during cold acclimation are essential for the survival of freezing and cryopreservation stress (Koštál et al., 2011b; Rozsypal et al., 2018).

Here, we extend our previous results to include a whole complex of CPs accumulated by these freeze-tolerant larvae. Using a combination of four chromatographic mass spectrometric (MS) analytical platforms, we established metabolite profiles of four different C. costata tissues in two contrasting larval phenotypes that differ greatly in freeze tolerance and cryopreservability in liquid N₂. By combining the results of the analytical chromatographic MS analysis with matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI), we show that freeze-tolerant larvae accumulate putative CPs in all tissues, but especially in hemolymph. Using the same technique, we observed changes in CP localization during slow extracellular freezing associated with partial dehydration of hemolymph and tissues. We then used differential scanning calorimetry (DSC) to determine the thermal phase transition properties of artificial aqueous CP mixtures mimicking the concentrations of the five most abundant metabolites in the hemolymph. Finally, we show in a bioassay that mixtures of CPs administered through the diet work in a synergistic rather than an additive way to induce strong freeze tolerance and cryopreservability in otherwise freeze-sensitive larvae.

A colony of C. costata, Sapporo strain (Riihimaa and Kimura, 1988), was reared on artificial diet in MIR 154 incubators (Sanyo Electric, Osaka, Japan) as described previously (Kostal et al., 1998; Lakovaara, 1969). Two phenotypic variants (LD and SDA) of the 3rd larval instar were generated according to our earlier acclimation protocols (Des Marteaux et al., 2019; Koštál et al., 2011b; Rozsypal et al., 2018) (Fig. S1a). The LD larvae (warm acclimated, active larvae reared at long-day conditions) have limited survival after freezing (35% survive to −5°C, 10% to −10°C, and none survive freezing to −20°C or below). In contrast, practically all SDA larvae (diapause larvae reared at short-day conditions and gradually acclimated to cold) survive deep freezing to −30°C or even −75°C, and 42.5% survive at least 18 months of cryopreservation in liquid N₂ (Rozsypal et al., 2018). In addition, SDA larvae were frozen (variant SDA-frozen) according to a previously developed optimal freezing protocol (Rozsypal et al., 2018) schematically presented in Fig. S1b. Survival after freezing and cryopreservation in liquid N₂ was assessed as the ability to pupate and metamorphose into the adult stage within 42 days of thawing (at a constant 18°C, LD).

To collect the hemolymph, 30 larvae were gently pierced and torn on a piece of Parafilm, creating a large droplet of pooled hemolymph. This droplet was then extracted using a calibrated glass capillary (Drummond Sci., Broomall, PA, USA). Collection of hemolymph in a single group of 30 larvae took less than 2 min. To obtain fat body, muscle and midgut tissues, the larvae were quickly dissected under binocular microscope in ice-cold phosphate-buffered saline. Dissected tissues were collected in ice-cold extraction buffer (see below) and stored at −80°C until analysis. Dissection of a single larva took less than 2 min, and we took care to sample a similar part of the respective tissue from each individual. This allowed us to express each metabolite concentration as a ‘pool per tissue’. Only in hemolymph samples where we could measure the exact sample volume, could we also calculate the concentration in mmol l⁻¹.

Analyses were performed for whole C. costata larvae (pools of five larvae taken in four replicates) or dissected tissues (pools of 30 tissues in four replicates). Whole larvae were weighed to obtain fresh mass, immersed in liquid N₂, and stored at −80°C until analysis. Frozen samples of whole larvae/tissues were melted on ice and homogenized in 400 μl of extraction buffer (methanol:acetonitrile:deionized water in a volume ratio of 2:2:1). The methanol and acetonitrile (Optima LC/MS) were purchased from Fisher Scientific (Pardubice, Czech Republic) and the deionized water was prepared using Direct Q 3UV (Merck, Prague, Czech Republic). Internal standards, p-fluoro-DL-phenylalanine, methyl α-D-glucopyranoside (both from Sigma-Aldrich, St Louis, MI, USA) were added to the extraction buffer, both at a final concentration of 200 nmol ml⁻¹. Samples were homogenized using a TissueLyser LT (Qiagen, Hilden, Germany) set to 50 Hz for 5 min (with a rotor pre-chilled to −20°C). Homogenization and centrifugation (at 20,000 g for 5 min at 4°C) were repeated twice and the two supernatants were combined.

We performed targeted analyses of 49 select metabolites (Table S1) using a combination of four different mass spectrometry-based (MS) analytical platforms that were described previously: ECF-LC/MS and ECF-GC/MS (Štětina et al., 2018), SILYL-GC/MS (Škodová-Sveráková et al., 2020) and HILIC-LC/MS (Škodová-Sveráková et al., 2021). More details on instrumentation and chromatographic columns are in Table S1. All 49 metabolites were identified against relevant standards (Sigma-Aldrich) and subjected to quantitative analysis using a standard calibration curve method. The analytical results were validated by simultaneously running blank samples (no larvae in the sample), standard biological quality control samples (the periodic analysis of a standardized larva/tissue sample – the pool of all samples) and quality control mixtures of amino acids (AAS18, Sigma-Aldrich).

The MALDI-MSI (Khalil et al., 2017; Tuthill et al., 2020; Yang et al., 2020) was conducted in LD and SDA larvae and also on SDA frozen larvae (Fig. S1); the LD frozen variant was not considered relevant as all LD larvae die when exposed to freezing to −30°C. Larvae were rinsed in water, incubated for 3 min in 2% gelatin, and then submerged into 5 ml of 10% gelatin in a plastic mold (Tissue-Tek Cryomold, Sakura, Mdesa, Czech Republic) placed on an ice-cold stage. The molds with solid gelatin were then immersed in isopentane pre-cooled to −140°C in liquid N₂ vapors for 3 min and transferred to −80°C for storage until further processing. Preparation of the SDA frozen larvae was modified as shown in Fig. S2.

For cryo-sectioning, the frozen molds were placed into a cryostat chamber (Leica CM1950, Germany) pre-cooled to −20°C. Sections of 10 µm thickness were mounted on indium tin oxide coated glass slides (ITO, Bruker, Czech Republic), previously washed with hexane and then with 2-propanol. Sections of the LD and SDA larval variants were thaw-mounted on warm glass (standard procedure MALDI-TOF/MSI) (TOF, time of flight), while sections of the SDA frozen larval variant were mounted on a gold-coated ITO glass pre-cooled to −20°C in order to preserve the tissue organization as it formed during slow pre-freezing (extracellular ice formation and tissue freeze-dehydration). The sections were dried for 15 min in a desiccator at low air pressure (at room temperature for the LD and SDA variants, while at −20°C for the SDA frozen variant), then sealed in plastic foil inside a plastic mailer, evacuated, and stored at −80°C until further processing. The ITO glasses with sections were transferred to room temperature, left to temper for 15 min, and dried in a desiccator for 15 min. Samples were then sprayed with a matrix solution of 9-aminoacridine hydrochloride, 7 mg ml⁻¹ in 70% ethanol (v/v). A TM-Sprayer 3 (HTX-Technologies, USA) was used to cover the tissue with the matrix solution using the following settings: nozzle temperature, 50°C; 12 passes; flow, 0.035 ml min⁻¹; nozzle velocity, 1000 mm min⁻¹; track spacing, 2 mm, HH pattern; gas pressure, 10 psi; gas flow rate, 2 l min⁻¹; drying time, 0 s; nozzle height, 40 mm; and propelled with 50% methanol (v/v). After spraying, the samples were dried in a desiccator for 15 min.

Spectral images were acquired using a rapifleX MALDI-TOF/TOF spectrometer (Bruker, Germany) in reflector negative mode with a 355 nm smartbeam™ 3D laser with a spatial resolution of 10×10 µm (pixel) in an m/z range of 20–1000, at a constant laser fluence of 72% and laser frequency of 5 kHz. Two hundred images were accumulated from every position. The instruments were set up with: ion source 1, 19.973 kV; PIE, 2.664 kV; lens, 11.353 kV; reflector 1, 20.810 kV; reflector 2, 1.034 kV; reflector 3, 8.577 kV. The pulsed ion extraction time was set to 90 ns and detector gain was 2302 V. Data were acquired with digitizer speed, 2.5 GS s⁻¹. Samples were measured in random order. Calibration was done externally using red phosphorus, achieving precision up to 5 p.p.m. (Kolářová et al., 2016).

After MALDI-MSI, tissues on ITO glass were washed three times for 2 min with 70% ethanol and afterwards washed with water. The samples were then stained with Hematoxylin and Eosin (H&E, a general staining method, which allows to visualize the shape, and structure of cells in a tissue sample; Hematoxylin stains mainly heterochromatin and nucleoli, while Eosin stains cytoplasmic components including collagen and elastic fibers) and scanned with Axio Scan.Z1 (Zeiss, Germany) operated by Zen 2 software (blue edition; Zeiss, Germany). Single raw data files on the intensities of the MALDI-TOF signals of individual m/z peaks were processed using SCiLS Lab software (v.2021c, SCiLS, Bruker, Germany). The raw data were smoothed with a convolution algorithm (width 20) and all subsequent computations and images rendering were done with total ion count square normalization. We collected MALDI-TOF signal-intensity datasets for 5 transverse sections of LD and SDA larvae and for 6 sections of SDA frozen larvae, plus one longitudinal section of an SDA larva. For relative quantification of signal intensities and statistical analyses, the transversal sections were used. Galleries showing the complete results for all transversal sections are available in figshare (https://figshare.com/articles/figure/Appendice_MALDI_MSI_images_pdf/19590694).

Regions (or tissues) of interest in the sections were manually delimited using H&E-stained images in QuPath software (v.0.2.3, University of Edinburgh, UK) (Bankhead et al., 2017) and imported into SCiLS software. The summary spectrum was exported into mMass software (v.5.5.0) (Niedermeyer and Strohalm, 2012), internally re-calibrated, and annotated based on m/z values with 50 ppm tolerance using an in-house modified database based on the HMDB metabolite database considering [M-H]⁻ and [M+Cl]⁻ adducts (Wishart et al., 2018). Selected peaks were then analyzed using SCiLS software with mean spectral intensity scaled in arbitrary units (A.U.) for each region and treatment. These spectral intensity data were then statistically compared using ANOVA followed by Bonferroni's multiple comparison test (Prism, v.6.07, GraphPad Software, San Diego, USA). In addition, we verified the identities of m/z peaks for trehalose and proline in the experiment where LD larvae were fed diet augmented by ¹³C6 glucose (for details, see Figs S13 and S14). The ¹³C6 glucose was purchased from Sigma-Aldrich (product ID 389374).

Thermal analysis was performed using the Q2000 calorimeter (TA Instruments, New Castle, DE, USA) as previously described (Pokorna et al., 2020). The analyzed solutions (see Table S2 for a complete list) were loaded and hermetically sealed into aluminium pans. The amount of loaded solution (∼5 µl) was controlled by weighing with precision to 0.01 mg using a Sartorius CP 225 D balance (Sartorius AG, Goettingen, Germany). The solutions were frozen and thawed according to the following protocol: (i) start at 25°C, (ii) cool to −90°C at 10°C min⁻¹, (iii) hold for 5 min at −90°C, and (iv) warm to 25°C at 10°C min⁻¹. An empty pan was used as a reference, and samples were run in technical triplicates (same solution analyzed three times). DSC results were analyzed using TA Universal Analysis 2000 software (v.4.5A Build 4.5.0.5). The onset of exotherm on the cooling scan was taken as the supercooling point (temperature of ice crystallization). Two major thermal events were analyzed on heating scans above −60°C. (1) Glass transition: an inflection point of the second-order phase transition was read as the temperature of vitrification (T_g). The change in specific heat capacity (ΔC_p) was derived from the difference in heat flow between the onset and the end of the glass transition. (2) Melting of bulk water: an onset of melting endotherm (first-order phase transition) was read as the melting point (m.p.), while the enthalpy of melting (ΔH, calculated from the area under the endothermic peak) served to estimate the fraction of melted (i.e. crystallized) water using the standard heat of fusion for the ice/water transition of 334 J g⁻¹. The remaining fraction of unfrozen water was considered osmotically inactive (%OIW).

Seventeen-day-old freeze sensitive (LD) larvae (i.e. roughly mid-3rd instar, voracious feeding stage) were extracted from the standard diet and groups of 20 individuals were moved to one of a variety of specific diets: (i) fresh standard diet (control); (ii) ‘CP mix 508’ diet – a standard diet augmented with five select CPs (proline, product ID P0380; trehalose, T9449; glutamine, G3126; asparagine, A0884; and betaine, 61962; all from Sigma-Aldrich) in concentrations corresponding to those observed in SDA larval hemolymph (i.e. 508 mmol kg⁻¹ water in total, see Table S3); (iii) five different standard diets augmented with the select CPs individually; (iv) ‘CP mix minus Pro 313’ diet – a mixture of the four select CPs but without the 313 mmol kg⁻¹ proline [because proline was the only metabolite capable of supporting freeze tolerance in a previous experiment (iii)]. After exposure to these diets for 3 days, the larvae were removed and used for the liquid N₂ survival bioassay (Fig. S1b).

Results of targeted quantitative analysis of 49 metabolites are summarized in Table S1 and plotted in Fig. 1A (hemolymph) and Fig. S3 (other tissues). For most metabolites, the pool in the hemolymph was significantly larger than the pools in other tissues (Fig. S4). The volume of hemolymph in one larva is approximately 200 nl (see Table S1) representing only 10% of the total body volume or body mass (∼2 mg). Nevertheless, the hemolymph contained approximately 50% of the total metabolite pool in both LD and SDA larvae (Table S1). Entry into diapause and cold acclimation (associated with the acquisition of extreme freeze tolerance) induced accumulation of specific metabolites (putative CPs) in SDA larvae (Fig. S1). The sum of the molar concentrations of the 49 metabolites in LD hemolymph was 182 mmol l⁻¹, but this increased to 554 mmol l⁻¹ in SDA hemolymph (Table S1). Four compounds that contributed most to this increase were: proline (increase of 277 mmol l⁻¹), trehalose (increase of 59 mmol l⁻¹), glutamine (increase of 22 mmol l⁻¹) and asparagine (increase of 17 mmol l⁻¹). The same four compounds also ranked first in order of concentration in SDA hemolymph: proline (313 mmol l⁻¹), trehalose (108 mmol l⁻¹), glutamine (55 mmol l⁻¹) and asparagine (26 mmol l⁻¹). Glycine-betaine (hereafter ‘betaine’, 6.1 mmol l⁻¹) was fifth in concentration, although it increased only moderately (by 1.7 mmol l⁻¹) in SDA compared with LD larvae. In addition, three other compounds had statistically higher hemolymph concentrations in SDA than in LD larvae, although their concentrations in SDA hemolymph were relatively low in absolute terms: glycerophosphoethanolamine (GPE, 4.0 mmol l⁻¹), glycerophosphocholine (GPC, 2.9 mmol l⁻¹) and sarcosine (0.5 mmol l⁻¹). For the purposes of this article, we consider the above compounds as putative CPs.

Using MALDI-MSI, we localized select metabolites in larval sections. A longitudinal section through the SDA larva provides the best overview of tissue structure (Fig. 1B). The highest intensity of the MALDI-TOF signal for ATP is localized predominantly in muscle tissue (corresponding to its role as the most important intracellular phosphagen) (Fig. 1C). In contrast, the highest intensities of proline and trehalose signals are localized in the hemolymph (Fig. 1D,E), where they are also colocalized with the high-intensity signals of glutamine, asparagine and GPE (Fig. S5).

As it was difficult to obtain enough high-quality longitudinal sections, we used transverse sections through the middle region of the larvae for the statistical analysis of the relative intensities of the MALDI-TOF signals. Quantitative analysis of these signals is technically difficult for small metabolite molecules freely dissolved in biological aqueous solutions (Fig. S6). Nevertheless, our analysis suggests that there are no differences in the overall localization patterns of the five putative CPs (proline, trehalose, glutamine, asparagine and GPE) between LD and SDA larvae. However, the signal intensities for all five CPs were much stronger in SDA than in LD larvae (compare the y-axes of Fig. S7 with Fig. S8). The MALDI-MSI thus confirms the results of the quantitative MS analysis (Fig. 1A) that the putative CPs are accumulated in all tissues, but especially in the hemolymph, of SDA larvae (for details, see MALDI-MSI images in figshare (https://figshare.com/articles/figure/Appendice_MALDI_MSI_images_pdf/19590694).

During slow inoculative freezing to −30°C (rate of 0.1°C min⁻¹_; see Fig. S1b for the freezing protocol) large masses of extracellular ice crystals developed in between partially dehydrated pools of larval hemolymph and tissues (Fig. 2A–D). We observed a striking difference in the ‘behavior’ of two major components of the CP mixture during slow extracellular freezing: whereas trehalose was concentrated in the pool of dehydrated hemolymph of SDA larvae, a large quantity of proline apparently ‘migrated’ from the hemolymph toward the ice region (Fig. 2E,F) and formed thin ‘layers’ of concentrated proline at the boundary between the extracellular ice and the dehydrated hemolymph and tissues (Fig. 2G–I). In these layers, proline colocalized with high-intensity signals from glutamine and asparagine, while trehalose colocalized with GPE in the dehydrated hemolymph (Fig. S9; for details, see figshare (https://figshare.com/articles/figure/Appendice_MALDI_MSI_images_pdf/19590694).

To assess the influence of different CP mixture components on water binding and thermal phase transitions, we compared heating curves from DSC scans of aqueous solutions containing single components or different CP mixtures with real SDA hemolymph. We focused on melting of bulk water (Fig. 3A) and glass transition (Fig. 3B). The full DSC analysis results can be found in Table S2.

We designed the CP mixtures according to the concentrations of the five most abundant (and commercially available/affordable) compounds in the SDA larval hemolymph. CP mix 508 mimics the composition of the major CPs in the hemolymph of SDA larvae prior to freezing, while CP mix 1588 takes into account a 3.125-fold increase in the concentration of all components resulting from extracellular freezing of osmotically active water (and, this way, it mimics the composition in the SDA-frozen variant). In addition, we tested all components individually at concentrations equivalent to those estimated in SDA hemolymph after extracellular freezing, and we also designed two reduced-composition mixes: Pro-Tre 1316 (proline 978+trehalose 338) and Pro-Gln-Asn 1231 (proline 978+glutamine 172+asparagine 81) (see Table S2 for details).

We found that depression of the bulk water melting point, as well as the fraction of ‘bound’ water (unfrozen, osmotically inactive water, %OIW), correlated linearly with the molal concentration of a solution, regardless of whether it was a single component solution or a mixture. Even the values for real hemolymph collected from SDA larvae fitted well to a linear relationship (Fig. 3C).

A clear glass transition was observed whenever trehalose was present in a mixture (Fig. S10). The temperatures of glass transition (T_g) in different trehalose-containing solutions varied from −29.6°C (trehalose 338) to −49.7°C (Pro-Tre 1316) (Fig. 3D). The change of thermal capacity (ΔC_p) during de-glassing was slightly above 0.3 J (g °C)⁻¹ in the solutions containing a relatively high concentration of trehalose (338 mmol kg⁻¹) but lower, ∼0.1 J (g °C)⁻¹, in the solution containing a relatively low concentration of trehalose (CP mix 508 contains 108 mmol kg⁻¹ of trehalose) and in real SDA hemolymph (which also contains 108 mmol kg⁻¹ of trehalose) (Fig. 3E). No glass transition was observed in the concentrated proline 978 solution (Fig. S11). The presence of proline in a mixture with glutamine and asparagine (Pro-Gln-Asn 1231) resulted in the elimination of small glass transitions observed in single-component mixtures of glutamine or asparagine (Fig. 3D,E). We also observed minor exotherms in heating curves of glutamine and asparagine at −22.8°C and −14.6°C, respectively (Fig. S12). The presence of such exotherms was also eliminated by the addition of proline in the Pro-Gln-Asn 1231 mix (Table S2).

To evaluate the ability of a complex cryoprotective mixture to increase freeze tolerance compared with individual components of the CP mix, we fed LD larvae with CP-augmented diets and then performed freezing survival assays. We prepared the CP mix 508 diet by augmenting the standard larval diet with a CP mix 508 solution (see Table S3). When LD larvae were fed with the CP mix 508 diet from hatching, more than 90% died (often as 1st instars). We therefore exposed only 3rd instar LD larvae to this diet and for only 3 days (at ages 17–19 days), which reduced larval mortality to 20%. The surviving larvae were significantly smaller at day 19 and pupariation was delayed by 2 days (Table S4), indicating that even short-term exposure to CP mix 508 appears to have some toxic effects.

A 3 day exposure to the CP mix 508 diet significantly increased whole-body concentrations of most CP mix components: proline increased 6.8-fold, trehalose 1.7-fold, glutamine 5.4-fold and asparagine 19.9-fold (see Table S5 for more details). In the hemolymph, the concentrations of proline, glutamine and asparagine reached high levels of 289, 80 and 49 mmol l⁻¹, respectively, while trehalose and betaine remained relatively low at 29 and 7 mmol l⁻¹, respectively (Fig. 4A, see Table S5 for more details). The sum molarity of the five CPs in hemolymph reached 454 mmol l⁻¹ in LD larvae fed with the CP mix 508 diet, which is broadly similar to the sum of 508 mmol l⁻¹ seen in SDA larvae.

No LD larvae that fed exclusively on the standard diet survived cryopreservation in liquid N₂. In striking contrast, 58% of the LD larvae survived cryopreservation after only 3 days of feeding with the CP mix 508 diet; 24.5% went on to pupate and 18.3% successfully metamorphosed to the adult stage (Fig. 4B). The components of the CP mix 508 administered via the diet had much less (or no) effect on larval freeze tolerance when they acted in isolation: betaine and asparagine had absolutely no effect; glutamine and trehalose resulted in 1 and 2 larvae surviving (out of 100), respectively, but these larvae did not pupate; proline augmentation was most effective, resulting in 37.7%, 10.3% and 7.2% survival for larvae, pupae and adults, respectively. When proline was excluded from the CP mix, the remaining components allowed only 4 larvae (out of 100) to survive liquid N₂ exposure, demonstrating that proline is the most important component of the mixture, or even the requisite for survival in liquid N₂. Nevertheless, the complete CP mix 508 was approximately 2-fold more effective at bolstering liquid N₂ survival than proline alone. This result was consistent at all three levels of survival (see t-tests in Fig. 4B) and also across five different generations of assayed insects (Table S6). Simple addition of the individual effects of the three components (proline+trehalose+glutamine) was not sufficient to explain the effect of the complete CP mix 508. These results suggest a synergy between CP mix 508 components that should be investigated in the future.

We identified the composition of the innate putative CP mixture which bathes the tissues of extremely freeze-tolerant C. costata larvae: proline, trehalose, glutamine, asparagine, betaine, GPE, GPC and sarcosine, which occur in hemolymph in a stoichiometric ratio of 313:108:55:26:6:4:2.9:0.5 mmol l⁻¹, respectively. It may not be surprising that the mixture is broadly similar to the cocktails known from other organisms exposed to various environmental stressors (Gertrudes et al., 2017; Choi et al., 2011; Somero, 1986; Yancey, 2005). The occurrence of the same classes of organic cytoprotectants and compatible osmolytes in different organisms is likely based on convergent evolution which favors accumulation of polar molecules with high solubility in water (high accumulation capacity) countered by low tendency to interact with charged groups of macromolecules and membranes (low toxicity) (Hochachka and Somero, 2002).

We performed a bioassay showing that supplementation of the larval diet with major components of the CP mixture in their native stoichiometric ratio changes freeze-sensitive larvae into freeze-tolerant/cryopreservable larvae within just 3 days. This result convincingly demonstrates that the accumulation of CPs has a strong functional (adaptive) significance in C. costata. However, this adaptation has probably evolved to cope with conditions typical for larvae overwintering in thermally buffered microhabitats under the bark of fallen trees, often under a blanket of snow (Band and Band, 1982; Grimaldi, 1986). Therefore, there may be different functional explanations for the roles of CPs in ecologically relevant situation in the field and for deep freezing and cryopreservation in liquid N₂ in the laboratory. Here, we avoid discussion of the eco-physiological aspects of CP-based freeze tolerance in animals and insects in the field, as these have already been covered for animals (Sinclair, 1999; Storey and Storey, 1988; Toxopeus and Sinclair, 2018; Zachariassen, 1985) and also specifically for C. costata (Rozsypal et al., 2018) elsewhere. Instead, we limit further discussion to the potential relevance of our results for deep freezing and cryopreservation in the laboratory.

We applied the MALDI-MSI technique to localize CPs in larval tissues and, for the first time in cryobiology literature, to determine how this localization changes during slow extracellular freezing, i.e. under conditions that allow larvae to survive deep freezing and subsequent cryopreservation in liquid N₂. So far, concentrations of CPs in insects have mostly been reported in whole body samples for two general reasons: first, insect tissues are small, making it technically difficult to estimate precise concentrations; and second, native CPs – being standard products of metabolism – have been considered to move freely across membranes (Somero, 1986). In cases where CP analyses were performed on separate tissues (Koštál et al., 2011a; Rozsypal et al., 2013; Toxopeus et al., 2019), the CPs were generally found in all examined tissues. Here, we have confirmed that SDA larvae of C. costata accumulate putative CPs in all examined tissues but that the highest concentrations occur in the hemolymph.

One of the most salient findings of this study is the observation that the two major components of the CP mixture, proline and trehalose, ‘behave’ differently during slow extracellular freezing. The MALDI-MSI analysis suggests that the bulk of trehalose molecules remain in their original location (before freezing) and concentrate in partially freeze-dehydrated hemolymph and tissues. In contrast, some of proline molecules appear to move out of the hemolymph and concentrate in thin layers separating the extracellular ice crystals from freeze-dehydrated tissues and hemolymph. At this stage of research, we can only offer hypothetical/speculative explanations for the physico-chemical mechanisms responsible for delocalization of proline and trehalose and for its possible functional significance in a cryopreserved insect. We suggest that the mechanism behind the de-localization could be explained by differences between trehalose and proline in water solubility, propensity to induce glass transition, and mobility in highly viscous freeze-concentrated solutions.

We have previously shown that in gradually frozen SDA larvae a glass transition of the residual solution occurs between −20°C and −30°C. Moreover, only larvae pre-frozen to temperatures below −30°C (i.e. below the temperature of glass transition) were able to survive abrupt submersion and cryopreservation in liquid N₂ and successfully develop into adults (Rozsypal et al., 2018). This result clearly demonstrated the importance of the glass transition for survival in liquid N₂. In our earlier work (Rozsypal et al., 2018), we attributed glass formation to high concentrations of proline in accordance with results of the DSC analysis performed by Rudolph and Crowe (1986). However, this observation was challenged in later DSC studies (Liu et al., 2020; Rasmussen et al., 1997) and our own analysis confirms that proline at a concentration of 978 mmol kg⁻¹ does not induce glass transition, at least not at temperatures above −60°C. Instead, trehalose is known in physical chemistry as one of the strongest glass transition inducers in aqueous systems (Cesaro et al., 2008; Green and Angell, 1989; Chen et al., 2000). Nicolajsen and Hvidt (1994) reported that a water:trehalose mixture (%weight, 89.7:10.3) de-glasses at −30.5°C with a ΔC_p of 0.38 J (g °C)⁻¹. These values agree well with our DSC analysis results showing de-glassing at −29.6°C with a ΔC_p of 0.33 J (g °C)⁻¹ for the trehalose 338 solution (which is 88.7:11.3%weight of water:trehalose). Accurate interpretation of the thermal behavior and glass transition parameters is a challenge for physical chemistry of simple binary mixtures such as trehalose:water (Cesaro et al., 2008; Chen et al., 2000; Nicolajsen and Hvidt, 1994; Olgenblum et al., 2020; Weng and Elliott, 2014). This limitation becomes almost insurmountable for more complex mixtures and, especially, for biological solutions. Nevertheless, our DSC analysis suggests that trehalose rather than proline may be the component mainly responsible for the glass transition in larval body fluids: (i) all artificial CP mixtures containing trehalose showed clear glass transitions with broadly similar T_g and ΔC_p parameters; (ii) the parameters were similar for real hemolymph collected from SDA larvae (containing 108 mmol l⁻¹ trehalose) and for the artificial CP mix 508 (simulating composition of major CPs in SDA hemolymph and also containing 108 mmol kg⁻¹ trehalose); (iii) no glass transition was observed in the concentrated proline 978 solution.

During glass transition, the viscosity of the supercooled liquid increases rapidly as the frequency, strength and lifetime of hydrogen bonds in a system increases. In dilute systems, most hydrogen bonds are formed between water molecules, but as the concentration of solute molecules increases the interactions between water–trehalose and trehalose–trehalose become more frequent until trehalose molecules form large hydrogen-bonded clusters that exclude water (Olgenblum et al., 2020; Weng and Elliott, 2014). The increase in viscosity slows the molecular dynamics and macromolecules become trapped in an amorphous hydrogen-bonded matrix, which probably prevents undesirable transitions such as protein unfolding. In addition, trehalose can directly form hydrogen bonds with macromolecules, which probably further increases their stability in the amorphous glass (Crowe et al., 1998; Olgenblum et al., 2020). Trehalose molecules involved in glass formation are relative immobile. Moreover, large proportions of water molecules are ‘bound’, i.e. osmotically inactive in SDA larval hemolymph (35.5%) or in whole-body C. costata (42%; Rozsypal et al., 2018) and therefore do not contribute much to trehalose solvation. As a result, trehalose can even locally reach the solubility limit (∼1.7 mol kg⁻¹ according to ALOGPS, http://www.vcclab.org/lab/alogps/), partly crystallize and form a eutectic system with ice crystals. Both crystallization and glass formation would make trehalose molecules immobile, i.e. gradually concentrated in dehydrated tissues and hemolymph, which we observed.

In contrast to trehalose, proline does not induce a glass transition (Liu et al., 2020; Rasmussen et al., 1997) and its solubility in water is enormous (it may reach over 15 mol kg⁻¹; Held et al., 2014; Qiu et al., 2019). Nevertheless, physical chemistry shows that proline in high concentration also extensively hydrogen bonds with water and together they form a viscoelastic liquid (de Molina et al., 2017; McLain et al., 2007; Troitzsch et al., 2008). The formation of dense, viscoelastic liquids is a guiding principle underlying the theory of NADES systems (Choi et al., 2011). The components in a complex mixture forming NADES are connected to one another through a well-organized tridimensional system with optimum interactions via inter- and intramolecular hydrogen bonding (Dai et al., 2013; Dai et al., 2015). Through these interactions, NADES systems exhibit emergent properties: (i) the melting point is deeply suppressed (usually below −30°C); (ii) ice crystallization is completely (or at least strongly) inhibited; whereas (iii) the tendency to transition to amorphous glasses is enhanced (Castro et al., 2018; Liu et al., 2018; Martins et al., 2019). However, work with NADES has also shown that the interactions between the NADES components are weakened with water dilution and even broken with water content exceeding 50% weight, when each component recovers its own specific properties (Castro et al., 2018; Dai et al., 2015; Durand et al., 2016). Our artificial CP mixtures have relatively high water content (between 79 and 99%weight), as do the SDA larval hemolymph (74.4%weight) and whole larvae (69.3%weight). Using DSC analysis, we show that these solutions exhibit significant suppression of the melting point (e.g. to −9.2°C in CP mix 1588, and to −4.5°C in SDA hemolymph) and significant water-binding capacity (41.1%OIW in CP mix 1588, and 35.5%OIW in SDA hemolymph). Both parameters (the decrease in melting point and the increase in OIW fraction) were linearly dependent on the osmolality of a solution (from 0 for water to 1588 mmol kg⁻¹ for CP mix 1588). These results suggest that neither the CP mixes 508 and 1588 nor the SDA hemolymph are ‘deep eutectic systems’ (DES, in the strict sense of Martins et al., 2019), but regular aqueous solutions with relatively low transition temperature from liquid to solid phase [LTTM, in the sense of Durand et al. (2016)].

We suggest that a proline-dominated viscoelastic solution with NADES-like properties may form on a microscale level, locally, in slowly freeze-dehydrating larvae. As water molecules are osmotically driven from the gradually dehydrating hemolymph and tissues and migrate toward the extracellular ice during extracellular freezing, the highly soluble and relatively mobile proline molecules could move together with water until they reach a barrier of ice where they form a thin layer of concentrated proline, which we observed in MALDI-MSI. Such a viscoelastic liquid could contribute to the stabilization of deeply frozen C. costata larvae by forming a rubber-like zone between extracellular ice crystals and freeze-dehydrated tissues, thus reducing the thermo-mechanical stresses associated with temperature fluctuations during immersion in liquid N₂ and rewarming (Rubinsky et al., 1980).

Overall, our study reveals the composition of a specific mixture of putative cryoprotectants accumulated by the freeze-tolerant larvae of C. costata. We demonstrate in a bioassay that as little as 3 days of exposure to a diet enriched with an artificial mixture of CPs allows an otherwise freeze-sensitive larva to survive not only deep freezing but also cryopreservation in liquid N₂. The components of CP mixture act synergistically rather than additively to improve the freeze tolerance. These findings open avenues for functional experiments to better understand the mechanisms underlying cryoprotection and cryopreservation of complex multicellular organisms. Toward this goal, we demonstrate that the accumulated CPs suppress the melting point of body water, increase the fraction of osmotically inactive water and significantly reduce the ice fraction. During slow extracellular freezing of whole larvae, trehalose becomes concentrated in the partially dehydrated hemolymph and tissues and stimulates glass transition, whereas proline moves to the boundary between extracellular ice and dehydrated tissues where it probably creates a layer of dense viscoelastic liquid. We suggest that these mechanisms contribute to larval extreme freeze tolerance and enable their cryopreservation by protecting their cells against thermomechanical shocks associated with freezing and transfers into and out of liquid N₂.

We thank Irena Vacková, Jana Járová, and Karolina Kowalska for maintenence of C. costata colonies, processing samples for metabolomics, and assistance during preparation of MALDI-MSI samples, respectively.