T. miqueliana fruits were harvested in November 2020 at the Huang Zangyu National Forest Park (117°03′–117°06′E, 34°–34°06′N), Anhui Province, China. The fruits were placed in net bags and stored at room temperature.

Fruits with full seeds were selected for testing based on X-ray image analysis. Fruits were placed on No. 3 enlarging paper and radiographed using an HY-35 X-ray machine (Xiang Xi Instrument Factory, Hunan, China) at a voltage of 80 kV, 5 mA electric current, 120 s exposure time and a focus-film-distance of 25 cm.

The fruits were treated with liquid N (Group 1) or dehulled manually (Group 2) as follows:

Group 1: fruits were placed in a stainless-steel box, and liquid N was poured into the box (volume ratio of fruits: liquid N = 1:1.5). After 20, 40, 60, and 80 s in contact with liquid N (-196°C), the fruits were placed on the laboratory bench and the pericarp was tapped with a 2.6-cm diameter hammer.

Group 2: The fruits at room temperature were squeezed in a vice to break the pericarp.

For Group 1 and 2 treatments, four replicates of 100 fruits were used. The processing time for 100 fruits was recorded and used to estimate the average processing time per fruit. Finally, undamaged seeds were removed with tweezers while damaged seeds were discarded and the percentage of damaged seeds during hulling determined.

Seeds collected in 2017, 2018, and 2019 were treated with liquid N for 40 s and tapped with a hammer to remove the pericarp. After seed removal, undamaged seeds were placed in a 50-mL beaker, and distilled water was added (volume ratio of seeds: distilled water = 1:2). The full seeds sinking to the bottom were removed and dried in air. All full seeds were placed into a 500-mL beaker. Acid (98% H₂SO₄) was added to the beaker (seed volume: acid ratio = 1:1.5) and then the mixture was stirred continuously for 15 min with a glass rod. Gauze was tied around the mouth of the beaker, and then all of the acid was poured off. The beaker containing the seeds was rinsed with tap water for 20 min and the seeds placed into a yarn bag and soaked in gibberellin (GA₃) solution (500 mgL⁻¹) in a 500-mL beaker (seed: GA volume ratio = 1:1.2). The beaker was wrapped in a black plastic bag and placed in the dark for 12 h, and then the GA₃ solution was poured off. Seeds were mixed with wet sand (45% humidity) at a 1:3 (w/w) ratio and stratified at 4°C for 75 days in darkness. When the radicle length reached 2 mm, the seeds were considered to have germinated. Final germination is presented as mean ± SD, based on four replicates of 100 full seeds. Manually hulled seeds served as the control.

Paraffin sections were prepared and observed to characterize pericarp cell morphology. Eight samples were collected from the control (manually hulled) and liquid N (40 s) groups. An 1-mm³ piece was cut from the pericarp of each fruit, and then fixed for 24 h in a 70% formalin-acetic acid-alcohol fixative solution (Servicebio, Wuhan, China). Samples were then dehydrated, using an ethanol series, and embedded in a wax block. The wax blocks were cut into 10-μm sections using an HistoCore autocut sliding microtome (Leica, Guangzhou, China) equipped with a tungsten steel slicer. The sections were stained with safranine and fast green. All sections were scanned using a section scanner (Panoramic desk/midi/250/1000, 3DHISTECH Company, Budapest, Hungary). The images were viewed at 1× to 400× magnification using CaseViewer 2.4 software (https://www.3dhistech.com/solutions/caseviewer/). Target areas of the tissue was selected for 200× imaging. The areas of lignin and cellulose were measured using Image-Pro Plus 6.0 analysis software (two-and three-dimensional image processing and analysis) using mm as the standard unit length. The area percentages of lignin and cellulose were calculated based on 100 fields of view (each being 0.48864 mm²).

Eight fruit coat samples were collected from the control and liquid N (40 s) groups. A 2-mm² piece of the pericarp was mounted with double-sided tape on the sample stage of a scanning electron microscopy (SEM) (Thermo Fisher Scientific, Waltham, MA, USA). Samples were then gold-coated using a gold sputter coater (HITACHI E-1010, Tokyo, Japan) and observed by SEM in the high vacuum mode. Images were captured at 15 kV.

Atomic force microscopy (AFM) (Dimension Edge, Bruker, Germany) was used to observe the 3D structure of the ventral suture of the pericarp. Eight fruit samples were collected from the control and liquid N (40 s) groups. Tissue from the carpel junction of the pericarp was cut into 1-mm³ pieces and fixed immediately in 2.5% v/v glutaraldehyde aqueous fixative solution (Servicebio) for 2 h, then transferred to 1% w/v OsO₄ for 5 h, dehydrated in an ethanol series, and embedded in resin blocks. The resin blocks were cut using an ultra-thin microtome (Leica UC7, Wetzlar, Germany) equipped with a diamond slicing knife (Ultra 45°, Daitome, Bienne, Switzerland). The resin slices were fixed onto slides with double-sided tape.

Atomic level profile topography images by controlled probe piercing of the sample surface were acquired at a scan rate of 1.0 Hz and scan range of 5 µm.

Thirty pericarps from the control and liquid N (40 s) groups were analysed. The cellulose content was determined using the anthrone colorimetric method (Dubois et al., 1956); the lignin content by the acetyl bromide spectrophotometric method (Fukushima and Hatfield, 2001); and the hemicellulose content the by 3, 5-dinitrosalicylic acid (DNS) assay (Deshavath et al., 2020).

To determine the total sugar contents, the dry pericarp (200 mg for each of 6 replicates) was repeatedly extracted (four times) with 20 mL 80% ethanol at room temperature and centrifuged at 10,000 × g for 20 min. The supernatant was collected and evaporated at 50°C under vacuum. The residue was dissolved to 20 mL with distilled water with 0.5 g poly (vinypolypyrrolidone), and then the mixture was centrifuged at 10,000 x g for 10 min. The supernatant was made up to a known volume and was assayed for total sugar by the dinitrosalicylic acid method (Gong et al., 2014). Total sugar content (%) = (C × V × N)/(10⁶ ×V_S × W) ×100, where C = sugar content converted from standard curve (μg), V = total volume of constant volume solution (mL), N = dilution multiple, Vs = volume of constant solution absorbed during determination (mL), and W = weight of sample (g).

Gas chromatography-mass spectrometry (GC-MS; 7890A-5975C, Agilent, American) was used to determine the content of six sugar components (arabinose, fructose, glucose, sorbitol, sucrose, and trehalose) in pericarps (Bianchi et al., 1993). The dry pericarp was extracted with 1 mL of 80% methanol and concentrated by drying. An aliquot of 30 µL of pyridinium methoxylate hydrochloride solution and 70 µL N, O-bis (trimethylsilyl) trifluoroacetamide was added to the dried sample before centrifugation. The supernatant was analysed by GC-MS using a HP-5 column (30m × 0.25mm × 0.25 µm). The transfer line temperature was 280 °C and the ion trap temperature 220°C. Other conditions for GC and MS were as follows: injector temperature: 220°C; detector temperature: 28°C; scanning range (m/z): 50-1000 amu at a rate of 1.5 scan/s; electron ionization energy: 70eV. The substance types under different peaks were searched in the NIST database through the data processing system of the X-caliber workstation to determine each sugar component. The relative content of each component (%) was calculated by area normalization, thus: Content of each sugar component (%) = (peak area of each sugar component)/sum of peak areas of sugar components × 100.

Nanoindentation is a relatively new technique that has been used to measure the nanomechanical properties of surface layers of bulk materials and thin films (Tarefder et al., 2010). Eight seeds from the control and liquid N (40 s) groups were analysed. Samples collected from the carpel junction of the pericarp were cut into 1-mm³ pieces and fixed immediately in 2.5% v/v glutaraldehyde aqueous fixative solution (Servicebio) for 2 h, then transferred to 1% OsO₄ w/v for 5 h, dehydrated in an ethanol series, and embedded in resin blocks. The resin blocks were sliced using an ultra-thin microtome equipped with a diamond slicing knife as described above. The surface of the pericarp was polished. A Nanoindenter (iMicro, Nanomechanics, Oak Ridge, TN, USA) was used to test the hardness and elastic modulus of the pericarp. The resin blocks were fixed on a round slide with glue. A Berkovich indenter was used in the ‘load control’ mode, with the loading rate/load constant kept constant during the loading process. The conditions were: a load resolution of 50 nN, displacement resolution of 0.01 nm, 0.4 mN maximum applied load, 100 μN/s loading and unloading rate, 150 nm pressure depth, 5 s duration at maximum load, and a Poisson’s ratio (nu) of 0.28. The optical microscope of the instrument was used to randomly choose the indentation area of 10 spots around the mesocarp. Each data item was the average of three test results. At the same time, the morphology of the indentation was observed. When calculating the elastic modulus and hardness by the Oliver & Pharr method (Li et al., 2020), the hardness (H) and equivalent modulus (Er) can be defined by the following formulae: H = (P_max)/Ac and Er = (S × √π)/(2β× √Ac), where P_max is the maximum loading load, Ac is the surface area of the indentation, β is a parameter related to the shape of the indenter (for the Berkovich indenter, the value of β is 1.034), and S is the contact stiffness.

An X-ray diffractometer (Ultima IV, Rigaku, Japan) was used to test the crystallinity of the pericarp. Four replicates of 30 seeds from the control and liquid N (40 s) groups were used in these analyses. The pericarps were incubated at 60°C for 4 h, and then crushed into a powder using a crusher (DE-300 g, Hongjingtian, China). The powder was put through a 320-mesh sieve, spread out in the sample preparation frame, and then pressed flat with a knife. The test parameters were as follows: Horizontal Goniometer, generator power = 3 KW, voltage = 40 kV, current = 30 mA. Crystallinity was calculated from X-ray diffraction (XRD) spectra using Jade 6.5 software. Crystallinity values were determined using the formula: Xcw = (Ic)/(Ic + Iw) × 100%, where Xcw is X crystallinity %, Ic is the intensity of the crystalline diffraction peak, and Iw is the intensity of the amorphous dispersion peak. The grain size of the crystal plane was calculated by Scherrer’s formula: D = (K×λ)/(COSθ × β), where K = Scherrer constant, D = average thickness of the crystal grain perpendicular to the crystal plane, B = half-height width of the diffraction peak of the sample, θ = diffraction angle, and γ = X-ray wavelength (0.154056 Å).

The hardness of the pericarp was measured using a Vickers hardness machine (Falcon 507, Innovatest, Maastricht, The Netherlands). Eight seeds from the control and liquid N (40 s) groups were used in these analyses. After the outer surface of the pericarp was polished, the fruit was fixed with a clamp. A rhombus-shaped indentation was pressed into the pericarp’s outer surface using a diamond square pyramid with a vertex angle of 136°. A loading force of 10 gf was used and maintained for 10 s. The pressure was calculated based on the pressure per unit surface area of the indentation. The length of the two diagonal lines of the indentation was measured using the microscope on the instrument, and the software automatically displayed the hardness value.

Pericarp toughness was evaluated at room temperature using a universal testing machine (WDW-10, Bairuo, China). Eight seeds from the control and liquid N (40 s) groups were analysed. The seed was fixed on the fulcrum. Force was applied to the surface of the pericarp until it broke, and the fracture toughness recorded. Measurements were conducted in ‘displacement control’ mode, and the loading rate was 3 mm/min. The crack opening displacement (COD) was measured using a YYJ-10/6-N clip extensometer. The morphology of the fracture surface was observed under a fluorescent stereo microscope (M205FA, Leica).

Brittleness was evaluated using a low temperature brittleness testing machine (JMDW-2C, Shanghai, China) in eight fruits from the control and liquid N (40 s) groups. The brittleness temperature is the highest low temperature when the pericarp is damaged by impact. The freezing medium (industrial ethanol) was injected into the cold well, and the intact seed was clamped vertically on the holder. The temperature was lowered in 1°C steps towards −40°C and the seed was held for 3.0 ± 0.5 min at each temperature point. The holder was lifted and the pericarp was impacted once within 0.5 s and each fruit observed to determine whether any damage had occurred. If the fruit was damaged, the temperature was increased back to room temperature. If the fruit was not damaged, the temperature was decreased and tested until the pericarp showed a brittle fracture upon impact.

SPASS 25.0 (SPASS Software Inc., SPSS, Chicago, IL, USA) was used to analyze the significance of the correlation between hulling time and percentage of damaged seeds, cellulose, hemicellulose, and lignin contents, total sugars and sugar components, X-ray diffraction spectrum, cellulose crystallinity, grain, and full width at half-maximum (FWHM), hardness, toughness, and brittleness temperature. Results were considered statistically significant at P ≤ 0.01. All figures were drawn with Origin Software (Originlab, Northhampton, MA, USA).

In this study, in both the manually hulled and liquid N-treated groups, the pericarp split all along the ventral suture. However, the morphology of the split surface differed between the two groups. In the manually hulled group, the spiral vessel of the vascular bundle was pulled out after it separated from cells under an impact load. In the liquid N-treated group, the exocarp and mesocarp were separated, and the surface of the mesocarp was covered with small holes. Similarly, perilla seeds subjected to freeze-thaw (-20°C) had damaged seed coats, with signs of splitting at the seams, that facilitated seed processing and the efficient extraction of perilla seed oil (Lee et al., 2020). In another industrial application, liquid N freeze-drying can help to fabricate cellulose nanofiber (CNF). Treatment results in the destruction of the 3-Dcellulose network such that the CNF produced had a porous layered structure (Wang et al., 2020).

We used AFM to show that the mesocarp crack along the ventral suture was smooth in the manually hulled group, but undulating and rough in the liquid N treated group with an uneven distribution of raised and recessed parts. These findings suggest that the pericarp cells that contracted during liquid N treatment did not fully expand after rewarming. When wheat seeds were given a prolonged cold plasma treatment in an attempt to improve seed wettability, the seeds had enhanced seed coat roughness (as shown by AFM) but the germination of seedlings was slowed (Starič et al., 2022). In contrast, liquid N-treated T. miqueliana fruits had rougher morphology without comprised seed viability.

In the present study, the size of the intensities of the characteristic peaks, cellulose grains and overall crystallinity were reduced in the pericarp after LN treatment. Crystallinity is an important factor in determining the cellulose hydrolysis efficiency (Alvira et al., 2010). Shen et al. (2019) found that the intensities of the characteristic peaks decreased after liquid N and ball milling (LN–BM) pretreatments. The crystallinity index (CI) of cellulose significantly decreased to 41.7% after only 2 h of LN−BM treatment and further decreased to 39.4% after 3 h of LN−BM treatment. LN−BM was effective at destroying the hydrogen bond networks within cellulose (Shen et al., 2019). It is possible that liquid N-treatment also reduced the hydrogen bond networks within cellulose in the pericarp of T. miqueliana.

Low temperature brittleness refers to the impact absorption energy of the material as the temperature decreases. Below a certain temperature, the impact absorption energy decreases significantly and the material changes from a ductile state to a brittle state (Diao et al., 2017). After the liquid N treatment, the hardness and modulus of elasticity at the ventral suture, and the hardness and brittleness temperature of the pericarp were significantly increased, while the toughness was significantly decreased. This response is similar to the low temperature processing of walnut fruits, such that the maximum effect on the destruction of the hazelnut shell is achieved by freezing it in nitrogen for 15 minutes, the load being about 208.3 N, which differ much from the load without freezing and is F = 261 (N). Similarly, hazelnut shells become fragile and easily destroyed when the nuts are exposed to shock deformations, whilst the integrity of the nut kernels is preserved as much as possible (Neverov et al., 2022). Thus it seems that liquid N treatment dramatically alters T. miqueliana pericarp mechanical properties, including changing the fracture type from ductile to brittle.

Pericarps are rich in plant secondary cell walls polymers, such as cellulose, hemicelluloses, and lignin (Landucci et al., 2020). These ubiquitous organic polymers are non-toxic, biodegradable (in time), provide mechanical properties and have slightly different water holding properties, such that hemicellulose > cellulose > lignin (Santos et al., 2013).

The main building blocks of lignin are the hydroxy-cinnamyl alcohols (or monolignols) coniferyl alcohol and sinapyl alcohol (Vanholme et al., 2010). Cellulose is composed of glucose units (Hu et al., 2015). Hemicelluloses are heterogeneous polymers of pentoses (xylose, arabinose), hexoses (mannose, glucose, galactose), and sugar acids (Gunnarsson et al., 2008). In this study, the composition of the pericarp on a per weight basis in the control (manually dehulled) group was 10.4% cellulose, 16.3% hemicellulose, 39.1% lignin, and 1.1% pectin. Although these compounds are deemed to be relatively stable, after liquid N treatment, the pericarp contents of cellulose and hemicellulose decreased significantly, whilst that of lignin increased significantly, and that of pectin did not change. It is known that the polymorphic structures of cotton cellulose can change instantaneously when ammonia and alkali-impregnated fabrics are immersed in liquid N, e.g., by conversion to cellulose III or to cellulose II (Jung et al., 1977). Similarly, when cellulose is pretreated with liquid N coupled with ball milling (LN–BM), the glucose yield from cellulose increases almost two-times compared to that obtained from untreated cellulose (Shen et al., 2019). We assume, therefore, that the observed low temperature-induced modifications of cellulose and glucose contents in T. miqueliana pericarp may reflect similar degradation processes.

Seed banking under dry (c. 3–7% moisture content) and cold (i.e., −18°C or in liquid N) conditions underpins ex situ plant conservation; it is cost effective and requires less space compared to in-situ methods (Li and Pritchard, 2009). To survive transfer to sub-zero temperature, seeds must be dried below the high moisture freezing limit (HMFL), which is related to seed oil content (SOC), being about 9% moisture content (MC) for a 50% SOC and just over 20% MC for very low SOC (Pritchard, 1995). In this study, the extracted seeds’ MC was the same after of liquid N-treatment (8.59 ± 0.1%) and manually hulling (8.61 ± 0.1%). Although the SOC was not determined in T. miqueliana, Tilia species tend to produce desiccation tolerant seeds (Seed Information Database; http://data.kew.org/sid/) with low SOC, e.g. 18% average for three species (Plant FA database; https://plantfadb.org/). It is likely, therefore, that T. miqueliana seeds at 8.5% MC were below the HMFL when exposed to liquid N and that there was insufficient water in the seeds for damaging ice crystals to form. This is supported by observation that the seeds germinated normally after liquid N treatment. This response is consistent with the successful cryopreservation of Tilia cordata seeds at 7 – 18% MC (Chmielarz, 2002). Such cryotolerance in seeds has been observed in many species. For example, seed viability of 89% of 103 plant species from 33 families from five regions of the Russian Far East did not decrease in the course of cryogenic storage (Kholina and Voronkova, 2008). Similarly, it was concluded that a large proportion of native Western Australian species may be amenable to storage in liquid nitrogen based on the screening of 90 species from 33 families (Touchell and Dixon, 1993). Thus, transfer of seeds to liquid N can be an effective means of plant conservation.

In addition to inducing changes in the microstructure of the pericarp, liquid N treatment of seeds might cause the formation of cracks in the seed coat or covering structures that later facilitate the ingress of water and improved germination (Busse, 1930; Brant et al., 1971; Normah and Vengadasalam, 1992; Salomao, 1995). In tropical orthodox seed, liquid N treatment had no adverse effect on the germinability of seeds with heterogeneous levels of hardness (Salomão, 2002). In contrast, liquid N treatment had partially beneficial effects on B. basiloba, S. brasiliensis, S. mombin, Bowdichia virgilioides, Pterodon emarginatus (Fabaceae) and Apeiba tibourbou (Tiliaceae) seeds (Salomão, 2002). In corroboration, the liquid N treatment dehulling method did not have a negative effect on T. miqueliana seed germination, with germination being > 60% in every tested year. The widescale use of liquid N treatment/storage for biodiverse species suggested to us that the liquid N dehulling treatment might be a safe procedure to test on other forestry species with propagules that are difficult to process.

A nut is a fruit composed of a hard shell and a seed, where the hard-shelled fruit does not open to release the seed. For most of the nut fruits the common post-harvest processes are drying, shelling and storage (Ashok Raj, 1979). The most common traditional method for shelling the nut fruit is shelling by hand (manual), which is a time-consuming process and labour-intensive method. Based on our understanding of the impacts of liquid N treatment on pericarp structural properties, we hypothesized that a liquid N treatment could be used to facilitate the extraction of T. miqueliana, seeds, and those of other species. However, it was important to establish that the seeds would not be negatively affected by this treatment.

We selected three species that have seeds with mechanical dormancy: Sassafras tzumu (Lauraceae) (Zhou et al., 2006), Pinus bungeana (Pinaceae), which is suitable for landscaping because of its unique white bark (Guo et al., 2019), and Cerasus yedoensis (Rosaceae), which has brightly coloured flowers and is of high ornamental value (Wu, 1985). We also choose a wild apricot (Armeniaca sibirica, Prunus sp.), which is an important local fruit and potential industrial crop (Rai et al., 2016; Hansakon et al., 2019; Qin et al., 2019), and Pinus koraiensis (Pinaceae), which is one of the most economically important tree species in Northeastern China, with pine nut exports generating > US $ 250 million dollars each year (Lin et al., 2018) and other attributes also being of importance, such as the pine nut shell as a potential source of bioenergy (Kaviriri et al., 2021). The results for all six species studied show that the peeling time and percentage of damaged seeds is lower (<20% of seeds) after liquid N treatment than after manual hulling. Overall, this method has similar efficiency to the machine shelling of cashew (Ojolo et al., 2010). Nonetheless, liquid N shelling is more economical and greatly reduces the cost of hulling (see Table S2 available as Supplementary Data).