Samples were obtained from four regions in Alaska, USA, between September 2021 and February 2022. Nineteen lichens were collected from rock and tree bark substrates in Juneau, three in Kodiak, five in Utqiaġvik, and two in Fairbanks. These locations encompassed a range of biomes, from coastal rainforests in Juneau and Kodiak, to Arctic tundra in Utquiġvik, and to the boreal forest in Fairbanks (Simpson et al., 2007). Lichen samples were collected based on availability in the natural environment and feasibility of access to collection sites. We exclusively sampled foliose and fruticose lichens. These species protrude off their substrates in leaf- and hair-like structures, which enabled accurate identification based on physical features. Species were identified using vegetation identification guides (Pojar and MacKinnon, 1994). Samples were collected in sterile containers and stored at -18 ∘C prior to analysis.

To prepare extracts for quantitative analysis and temperature experiments, all samples were washed with pure water (Millipore Milli-Q^® Simplicity 185 Water Purification System, Merck KGaA, Darmstadt, Germany) to minimize contamination from external sources of INs. Extracts were prepared for ice nucleation assays and stability experiments using a standardized procedure adapted from Kieft and Lindow (Kieft and Lindow, 1988). First, a 5 mL aliquot of pure water was added to 2 g of lichen. Secondly, the samples were ground to a fine pulp using a mortar and pestle. This procedure was chosen to ensure that the aqueous extract contained all the molecules present in the lichen samples, both the molecules bound to the cell wall and any secreted molecules (Kieft and Lindow, 1988). Samples were then centrifuged at 5000 rpm for 10 min. The supernatants were filtered through 0.22 µm pore diameter syringe filters (Millex^® Syringe Filter, Merck KGaA, Darmstadt, Germany) to remove cellular debris. The aqueous extracts were immediately tested for ice nucleation activity using the Vali-type (Vali, 1971) apparatus and stored at -18 ∘C until systematic analysis. Lichen extracts remained frozen for up to 9 months.

For initial screenings of lichen ice nucleation activity, a Vali-type apparatus was used, which consists of a temperature-controlled aluminum cold plate (Linkam Scientific Instruments LTD, United Kingdom). For each sample, 20 extracted 1 µL droplets were cooled at a rate of 3 ∘C min-1, and freezing events were identified through the optical change in droplet appearance consistent with freezing. The freezing temperature for each droplet was recorded, and the temperature at which half of the droplets are frozen, T50, was used as a direct measure for the efficacy of the INs. Snomax (inactivated P. syringae; Snomax Int) was used as a positive control, resulting in a T50 of -3.5 ∘C at 1.0 mg mL-1, while pure water had a T50 of -15 ∘C.

The three most active lichen species, Platismatia herrei, Sphaerophorus globosus, and Peltigera britannica, were analyzed using the high-throughput twin-plate ice nucleation assay (TINA) (Kunert et al., 2018). In TINA measurements, a liquid handling station (epMotion ep5073, Eppendorf, Hamburg, Germany) was used to serially dilute the aqueous extracts in 10-fold increments with pure water, each dilution consisting of 96 (3 µL) droplets in two 384-well plates. The 384-well plates were cooled at a continuous rate of 1 ∘C min-1 from 0 to -30 ∘C. Freezing temperatures for each droplet were extracted, and the fraction of frozen droplets was calculated for the different temperatures. The cumulative number of active INs per mass unit of sample (Nm) in each dilution was calculated using Vali's equation (Vali, 1971).

The heterogeneous underlying-based (HUB) method and associated HUB-backward numerical code (de Almeida Ribeiro et al., 2022) were used to interpret the heterogeneous ice nucleation temperatures obtained from TINA measurements of the lichen samples. The HUB method relies on the assumptions of Vali (1971); e.g., that each IN has a distinct nucleation temperature, and that the IN with the highest nucleation temperature is responsible for the freezing of the droplet in the cooling experiments. The HUB-backwards stochastic optimization procedure was used to extract the differential spectrum Nm(T) and determine the underlying distribution of ice nuclei subpopulations in the cumulative freezing spectrum Nm(T) of lichen. The interpretability of the results in terms of subpopulations provides an advantage over polynomial fitting and differentiation of Nm(T).

Freeze–thaw cycles and heat treatments were performed on selected lichen extracts and Snomax, to assess the stability of INs exposed to temperature changes. For freeze–thaw cycles, aliquots of lichen extract were consecutively frozen by gradually cooling to -30 ∘C at 1 ∘C min-1 and thawed to room temperature 12 times over the course of 12 h. After each temperature cycle, the activity was determined using TINA. For heat treatment experiments, the aliquots of the lichen extracts were incubated at 98 ∘C for 1 h, and the ice nucleation activity before and after heating was determined using TINA.

Table 1 shows the freezing temperatures of 29 lichen extracts as determined in initial studies by a Vali-type droplet freezing assay and 16 of the lichen extracts as measured by TINA. Freezing temperatures are shown as T50 values and are defined as the temperatures at which 50 % of the undiluted lichen extract droplets are frozen. We find that all Alaskan lichen species collected between September 2021 and February 2022 froze at temperatures ranging from -5 to -14.5 ∘C. These findings are consistent with previous reports of widespread ice nucleation activity in lichen (Kieft and Lindow, 1988; Henderson-Begg et al., 2009; Moffett et al., 2015; Creamean et al., 2021). The ice nucleation activity of the collected lichen does not seem to increase from warmer (e.g., Juneau) to colder (e.g., Utqiaġvik) climate zones, and none of the lichen species collected in the Arctic tundra (Utqiaġvik) showed freezing temperatures warmer than -7.6 ∘C. Moreover, lichens with the highest ice nucleation activity (e.g., Platismatia herrei) coexist in the same coastal rainforest environment and even on the same tree substratum as lichens with relatively low ice nucleation activity (e.g., Usnea longissima). Among the 29 lichen, Platismatia herrei, Peltigera britannica, and Sphaerophorus globosus showed notably high IN activity in initial screenings, with T50 values of -5.2, -5.7, and -6 ∘C, respectively. To better characterize the INs of these lichen species, we investigated their INs using the TINA setup, which allows the simultaneous measurement of a complete dilution series with robust statistics (Kunert et al., 2018).

Figure 1 shows the results of TINA measurements of untreated aqueous lichen extracts of P. britannica, S. globosus, and P. herrei. The initial solutions had a concentration of 0.4 g mL-1 and were then serially diluted 10-fold. The cumulative IN number concentration (Nm) was calculated using Vali's formula and represents the number of INs per unit weight that are active above a certain temperature (Vali, 1971). We find that for the three lichen species, the freezing spectra show two strong increases in Nm(T) around -4.5 and -13 ∘C, with plateaus between ∼-7 and ∼-13 ∘C and below ∼-16 ∘C. The two rises in the spectra suggest the presence of two classes of INs with different activation temperatures, while the plateaus indicate that fewer INs are active within those temperature ranges (Budke and Koop, 2015; Lukas et al., 2020). The presence of two IN subpopulations is notable, given that quantitative measurements of fungi-associated INs typically display only one IN population (Fröhlich-Nowoisky et al., 2015; Kunert et al., 2019). Two rises and subsequent plateaus in the freezing spectra are more characteristic of those observed in the spectrum of P. syringae (Fig. 3b). For P. syringae, the increases at ∼-2.8 and -7.5 ∘C were assigned to two classes of INs that consist of different aggregate sizes of the same INPs (Turner et al., 1990; Qiu et al., 2019; Schwidetzky et al., 2021a). The classes and molecular composition of lichen INs have not been identified, but experimental results of Kieft and Ahmadjian (1989) have previously identified the lichen mycobiont as responsible for freezing above -5 ∘C, while the photobiont has much lower activity. We therefore tentatively assign the two increases in the freezing spectra of the lichens to INs of the mycobiont and photobiont partners. However, we cannot exclude the possibility that lichen INs are exclusively mycobiont-derived macromolecules that aggregate similarly to bacterial INs, and thus induce freezing at different activation temperatures depending on the aggregate size. Given the uncertainty, we refer to the first IN subpopulation active at -4.5 ∘C as class 1 and the second class active at -13 ∘C as class 2.

We employed the heterogeneous underlying-based (HUB) stochastic optimization analysis (de Almeida Ribeiro et al., 2022) to better identify and characterize the underlying number of IN subpopulations in lichen. Figure 2a shows the experimentally obtained cumulative spectrum Nm(T) of P. herrei, together with the differential spectrum predicted by HUB from the optimized distribution of nucleation temperatures (e.g., the differential spectrum), shown in Fig. 2b. The two IN subpopulations in the differential spectrum are centered at - 6.8 and - 13.8 ∘C (Fig. 2b). We assign the mode at - 6.8 ∘C to the class 1 INs and the mode at - 13.8 ∘C to class 2 INs. The differential spectrum shows that freezing at - 6.8 ∘C is two orders of magnitude less likely to happen than freezing at - 13.8 ∘C. The robustness of these signals and their physical nature require further investigation.

The lichen INs are exposed to physical stressors such as low, high, or quickly changing temperatures in the natural environment or as aerosolized particles. To investigate the stability of lichen INs, we performed freeze–thaw cycles and heat treatments. Figure 3 displays ice nucleation measurements of aqueous extracts of P. herrei and a solution of Snomax (inactivated P. syringae) in water during 12 consecutive freeze–thaw cycles. We find that the freezing spectra of P. herrei look similar across all cycles, with nucleation events occurring consistently at ∼-4.5 and -13 ∘C and plateaus between ∼-7 and -13 ∘C. In contrast to lichen INs, the freezing spectra of P. syringae show a systematic trend in which the final cycle differs markedly from the first. We find that the rise at ∼-2.8 ∘C in the bacterial Nm spectrum is reduced and shifted to slightly lower temperatures with progressing cycles. The second rise at ∼-7.5 ∘C shifts not only to lower temperatures but also increases over time. Evidently, repetitive freeze–thaw cycles reduce the activity of the class A INs (larger INP aggregates active at ∼-2.8 ∘C) and transform them into the less active class C INs (smaller aggregates active at ∼-7.5 ∘C). These findings are consistent with the apparent instability of bacterial INs (Polen et al., 2016), and contrast the high stability of the lichen INs.

The repetitive freezing measurements further reveal that the highest activity of lichen IN was already developed during the initial preparation of the extracts and that no equilibration time was required. The storage of aqueous extracts at -18 ∘C prior to freeze–thaw cycles also did not impact the ice nucleation activity of the highly active class 1 INs, given that T50 values of the initial measurements and up to 9 months later are similar. These findings indicate that lichen INs can undergo multiple freeze–thaw cycles without losing significant activity and would be reusable in freeze-tolerant lichens and able to influence cloud glaciation.

As a second test of stability, Fig. 4 shows ice nucleation measurements of aqueous lichen extracts before and after 1 h of exposure to 98 ∘C. We find that the freezing spectra of all three lichen species are altered substantially after heat treatment. For P. britannica, we observe that the first rise at ∼-4.5 is lowered to -6.5 ∘C, while the second rise at -13 ∘C remains and becomes more apparent. Moreover, the total cumulative number of P. britannica INs decreases moderately from 107 to ∼106 g-1. For S. globosus, we observe a markedly different behavior: the first rise at ∼-4.5 decreases to -9.5 ∘C, while the second rise at -13 ∘C completely disappears. In addition, the total cumulative number of INs decreases significantly from 107 g-1 to less than 105g-1. For P. herrei, we find that the first rise at ∼-4.5 is only slightly lowered to -6 ∘C, while the second rise at -13 ∘C is absent. Clearly, the lichen INs are affected differently by heat treatment, suggesting distinct variation across the biomolecules responsible for ice nucleation in lichens. We conclude that the macromolecular composition of INs of class 1 and class 2 likely differ both within individual lichen and among lichen species.

The high stability of the class 2 INs after the 98 ∘C treatment points toward some non-proteinaceous ice nucleating molecules, and it has been proposed that some biological INs may consist of polysaccharides (Pummer et al., 2012; Dreischmeier et al., 2017). Given the lower stability of the class 1 INs upon heating, it is reasonable to assume that the ice nucleation activity is primarily due to proteinaceous compounds that denature under high temperatures. However, future experiments, including chemical analyses, are required to adequately characterize the macromolecules responsible for ice nucleation activity in lichen.