Floral freezing tolerance is tied to flowering time in North American woody plant species

Abstract

Background and Aims

As winter and spring temperatures continue to increase, the timing of flowering and leaf-out is advancing in many seasonally cold regions. This advancement could put plants that flower early in the spring at risk of decreased reproduction in years when there are late freeze events. Unfortunately, relatively little is known about floral freezing tolerance in forest communities. In this study, we examined the impact of freezing temperatures on the flowers of woody plants in a region where there is rapid winter warming in North America.

Methods

We subjected the flowers of 25 woody species to a hard (−5 °C) and a light freeze (0 °C). We assessed tissue damage using electrolyte leakage. In a subset of species, we also examined the impact of a hard freeze on pollen tube growth. To determine if the vulnerability of flowers to freezing damage relates to flowering time and to examine the responsiveness of flowering time to spring temperature, we recorded the date of first flower for our study species for 3 years.

Key Results and Conclusions

Across species, we found that floral freezing tolerance was strongly tied to flowering time, with the highest freezing tolerance occurring in plants that bloomed earlier in the year. We hypothesize that these early blooming species are unlikely to be impacted by a false spring. Instead, the most vulnerable species to a false spring should be those that bloom later in the season. The flowering time in these species is also more sensitive to temperature, putting them at a great risk of experiencing a false spring. Ultimately, floral damage in one year will not have a large impact on species fitness, but if false springs become more frequent, there could be long-term impacts on reproduction of vulnerable species.

INTRODUCTION

In recent years, many temperate and boreal habitats in the northern hemisphere have exhibited early spring bud burst because of warmer than average winter and spring temperatures (Schwartz et al., 2006; Jeong et al., 2011; Büntgen et al., 2022). One potential consequence of this early growth is an increase in the frequency of frost and freeze damage experienced by some plant species (Cannell and Smith, 1986; Hanninen, 1991; Augspurger, 2009; Liu et al., 2018; Richardson et al., 2018; Vitasse et al., 2018). When early spring growth is followed by damaging, cold temperatures, it is often referred to as a false spring. False springs are increasing and/or are predicted to increase in frequency in many, but not all seasonally cold regions (Bennie et al., 2010; Augspurger, 2013; Ge et al., 2013; Allstadt et al., 2015; Park et al., 2021). When freezing damage occurs to leaf tissue, plants can sometimes grow a second cohort of leaves and compensate for the loss of resources by extending their growing season (Zohner et al., 2019). However, it is less common for plants to produce multiple cohorts of flowers, and as a result false springs can lead to fewer inflorescences, a shorter flowering time and reduced reproductive output for the year (Miller-Rushing and Primack, 2008; Pardee et al., 2018).

Freezing temperatures can decrease reproductive output by directly damaging the male and female gametophyte or indirectly by impacting pollination (Pardee et al., 2018). In terms of direct tissue damage, the female reproductive tissue is often the most sensitive to freezing temperatures (Neuner et al., 2013), but temperature can also impact pollen germination and pollen tube growth in some species (Pigott and Huntley, 1981). In terms of pollination, freezing temperatures are known to impact pollinator attractants like petals (Ishikawa et al., 2015; Kaya and Kose, 2019) and rewards like nectar (Akšić et al., 2015) and sometimes the synchrony between plants and their pollinators (Forrest, 2015; Pardee et al., 2018). Any of these negative impacts of freezing temperatures could compromise the ability of a plant to produce seeds in a given year.

Most research on floral freezing tolerance is focused on horticultural plants, for example fruit trees (Parker, 1963; Rodrigo, 2000). It is less clear whether the same sensitivity to freezing temperatures exists in native species, especially perennial plants from seasonally cold climates. In horticultural systems, any floral damage can lead to a decrease in the aesthetic quality of the plants/flowers and potentially lead to economic losses. In native systems, early flowering is only problematic if the fitness benefits (Price and Waser, 1998; Ehrlén and Münzbergová, 2009; Munguia-Rosas et al., 2011; Cleland et al., 2012; Rafferty et al., 2016; Ehrlén and Valdés, 2020) outweigh the costs of periodic loss in reproductive tissue. A handful of recent articles on subalpine and alpine habitats have highlighted the complexity that might exist when examining floral freezing tolerance in native plant communities. These studies suggest that freezing tolerance may depend on microhabitat (Ladinig et al., 2013), life history (Morales et al., 2020) and possibly flowering time (CaraDonna and Bain, 2016). There is also evidence that some (Kuprian et al., 2014) but not all floral tissue can supercool (Wagner et al., 2021).

Many temperate forests in North America are experiencing rapid winter warming, including those in northern Minnesota, USA. In this region, the average winter temperature increased by 2.1 °C from 1901 to 2011 and is expected to increase by an additional 4–6 °C before the end of the 21st century (Handler et al., 2014; Leiss et al., 2022). Forests in this region are dominated by deciduous trees and shrubs that flower early in the growing season, potentially making their flowers vulnerable to false springs. To determine whether this is the case, we need a better understanding of floral vulnerability to freezing temperatures. In this study, we assessed the floral freezing tolerance of 25 common, native woody/semi-woody species in Minnesota. We examined differences in floral damage caused by two freezing treatments (hard and light freeze) and tested whether plants that flower earlier in the growing season exhibit a higher floral freezing tolerance than those that flower later in the year. Freezing tolerance was quantified by measuring damage to bulk floral tissue and examining the impact of temperature on pollen tube growth. Our goal was to increase our understanding of the potential impact of climate change on temperate forests by determining whether plants that flower early in the growing season could have decreased reproductive output due to an increase in the frequency of false springs.

MATERIALS AND METHODS

Climate data and floral phenology

This study was completed on the University of Minnesota campus (latitude 46.8239, longitude −92.0870) and nearby at Chester Creek Park (latitude 46.8129, longitude −92.0963) in Duluth, MN, USA. The average last day when temperatures are at or below 0 °C in Duluth is 13 May, based on weather data from the last 30 years (1991–2020, NOWData, NOAA Online Weather Data Resource). During the month of May, the average minimum temperature is 5 °C, which is five degrees higher than in April, and five degrees lower than in June.

For three years (2020–23), the date of first flower was noted for 25 species (Table 1). A plant was considered in flower or at anthesis when flowers were open (petals reflexed) and/or there was pollen being released. Plants were monitored weekly. The date of first flower was not captured for seven species in 2020 (Apocynum androsaemifolium, Cornus rugosa, Diervilla lonicera, Rosa blanda, Solanum dulcamara, Tilia americana and Viburnum lentago) and instead the first flowering date was determined by regular iNaturalist observations recorded in the area. To examine how flowering time changed from year to year, we completed a linear regression of average day of the year (DOY) of first flower against the standard deviation of the DOY across years. We estimated the responsiveness of each plant to temperature by comparing the DOY of first flower between the warmest and coolest years. We determined which were the warmest and coolest years based the average temperature in the first quarter of the year and growing degree days accumulated by the end of March.

Table 1.

Species sampling, plant size and phenology. Quantitative traits are reported as average ± standard deviation.

Species	Family	Sex	Growth form	First flower (DOY)a	Height (m)b	Basal area (cm2)	Hard freeze	Light freeze	Pollen tube freeze
Acer negundo	Sapindaceae	F	Tree	134 ± 8	12 ± 4	500 ± 400	Yes	No	No
Acer negundo	Sapindaceae	M	Tree	134 ± 8	9 ± 5	300 ± 300	Yes	No	No
Acer rubrum	Sapindaceae	M	Tree	123 ± 11	4.5 ± 0.4	110 ± 40	Yes	No	Yes
Acer saccharum	Sapindaceae	M	Tree	139 ± 6	19 ± 10	800 ± 500	Yes	No	No
Acer spicatum	Sapindaceae	M/F	Shrub	158 ± 9	5 ± 2	10 ± 10	Yes	Yes	No
Alnus incana	Betulaceae	M	Shrub	107 ± 16	4 ± 1	60 ± 30	Yes	No	No
Amelanchier spicata	Rosaceae	M/F	Shrub	146 ± 8	2 ± 1	10 ± 10	Yes	Yes	Yes
Apocynum androsaemifolium	Apocynaceae	M/F	Herb	185 ± 5	0.9 ± 0.1	0.4 ± 0.2	Yes	No	No
Betula papyrifera	Betulaceae	M	Tree	136 ± 4	14 ± 5	1000 ± 700	Yes	No	No
Clematis virginiana	Ranunculaceae	M	Vine	207	3 ± 1	0.05 ± 0.03	Yes	No	No
Cornus rugosa	Cornaceae	M/F	Shrub	175 ± 7	1.7 ± 0.5	3 ± 1	Yes	Yes	No
Cornus sericea	Cornaceae	M/F	Shrub	160 ± 5	1.9 ± 0.4	10 ± 10	Yes	No	No
Corylus cornuta	Betulaceae	M	Shrub	110 ± 14	3 ± 2	10 ± 4	Yes	No	No
Diervilla lonicera	Caprifoliaceae	M/F	Shrub	178 ± 2	0.5 ± 0.2	0.3 ± 0.3	Yes	Yes	No
Populus tremuloides	Salicaceae	M	Tree	115 ± 17	13 ± 4	600 ± 300	Yes	No	No
Prunus americana	Rosaceae	M/F	Shrub	151 ± 13	3 ± 1	12 ± 6	Yes	No	No
Prunus virginiana	Rosaceae	M/F	Shrub	153 ± 8	3 ± 1	2 ± 3	Yes	Yes	Yes
Rosa blanda	Rosaceae	M/F	Shrub	179 ± 5	0.8 ± 0.3	0.6 ± 0.4	Yes	Yes	No
Rubus parviflorus	Rosaceae	M/F	Shrub	165 ± 6	1.0 ± 0.2	0.8 ± 0.7	Yes	Yes	Yes
Salix discolor	Salicaceae	F	Shrub	127 ± 14	6 ± 3	300 ± 400	Yes	No	No
Salix discolor	Salicaceae	M	Shrub	129 ± 10	6 ± 2	200 ± 200	Yes	No	Yes
Sambucus racemosa	Caprifoliaceae	M/F	Shrub	151 ± 5	2.6 ± 0.4	3 ± 2	Yes	Yes	Yes
Solanum dulcamara	Solanaceae	M/F	Herb	180 ± 6	1.0 ± 0.2	200 ± 90	Yes	Yes	Yes
Sorbus decora	Rosaceae	M/F	Tree	158 ± 9	8 ± 2	0.2 ± 0.1	Yes	Yes	No
Tilia americana	Tiliaceae	M/F	Tree	195 ± 4	10 ± 3	1000 ± 400	Yes	Yes	No
Viburnum lentago	Adoxaceae	M/F	Shrub	168 ± 3	2.7 ± 0.5	100 ± 100	Yes	Yes	Yes
Viburnum trilobum	Adoxaceae	M/F	Shrub	165 ± 7	2.4 ± 0.9	6 ± 7	Yes	Yes	No

^aAverage day of first flower (± standard deviation) from 2020 to 2023. For C. virginiana data were only collected for 1 year.

^bTraits that were significantly different between plants that flowered early (before mid-May) and late in the season based on a t-test.

Species selection and floral traits

Every week in the spring and summer of 2021, we selected two woody species that were flowering in our study area from early April to late July. We sampled ten individuals per species and measured the size of each plant (total basal area and height). To determine the average number of flowers per inflorescence, we counted the number of flowers in three inflorescences per plant (30 per species). Then we calculated the total number of flowers per plant by estimating the number of inflorescences and multiplying it by the average number of flowers per inflorescence. For plants with over 100 flowers per inflorescence, we estimated the number of flowers based on the weight of the floral tissue (i.e. petals, sepals, carpels, stamens and receptacle). We did not count the number of individual flowers in catkins and reported our data for these species in terms of inflorescence number.

We collected three inflorescences per plant when they were at anthesis and measured wet and dry floral weight along with floral water content. We once again considered floral tissue to include petals, sepals, carpels, stamens and the receptacle. We also estimated the average floral surface area for each of our species based on measurements of three to five inflorescences per individual (high sample number for small inflorescences). Floral surface area is important because objects with greater surface area can cool more rapidly. For flowers with large petals, sepals and tepals, we cut open the flowers, laid them on a flatbed scanner and measured the projected area of the floral organs. We considered their total surface area to be twice the projected area. For catkins, we estimated the surface area by assuming the organs were a cylinder and measuring the diameter and length of the catkins. We considered the plant floral investment to be the average dry mass of a flower multiplied by the number of flowers on the plant. In a few instances when we had incomplete data on individual plants, the same or similar sized plants were resampled in subsequent years at the time of flowering.

Assessing floral freezing damage

To compare floral freezing tolerance among species, we studied floral freezing damage after exposing cut branches with flowers to a hard freeze (−5 °C) in 2021. For a subset of species, we also examined their response to 0 °C in 2022 (Table 1). We selected these temperatures based on preliminary data on a subset of species (Savage et al., 2024b). We only used open flowers that were at anthesis because this is often the most vulnerable stage of flower development to cold temperatures (Quamme et al., 1982; Ladinig et al., 2013; CaraDonna and Bain, 2016). Freezing trials were completed in a controlled temperature freezer (Tenney TUJR, Thermal Product Solutions, White Deer, PA, USA) with a temperature ramping controller (Watlow F4S/D, Winona, MN, USA). The cycle started at 19 °C and ramped at four degrees per hour down to the minimum temperature (0 or −5 °C). This is a standard rate of cooling for freezing trials (Kovaleski and Grossman, 2021). After holding at that temperature for 3 h, the freezer warmed at four degrees for per hour until it returned to 19 °C. Beaded thermocouples (Type T, Oneset, MA, USA) were taped to four samples per cycle to record floral tissue temperature. Data were logged using a four-channel Hobo Thermocouple Data Logger (Oneset, MA, USA) and used to monitor exotherms (spikes in temperature of >0.3 °C, which indicate ice nucleation) during the treatment.

We assessed freezing damage using visual inspection and electrolyte leakage (Savage et al., 2024b). Electrolyte leakage is a common method used to assess freezing tolerance and provides an estimate of cellular and membrane damage (Flint et al., 1967; Burr et al., 1990). It requires the comparison of a frozen sample with a non-frozen control sample (kept in a refrigerator at 8 °C) and also an estimate of complete tissue death, which is made by killing the samples in a −80 °C freezer. The output parameter, index of injury (I_t) is calculated using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle I_{\mathrm{t}}=(\frac{L_{\mathrm{t}}}{L_{\mathrm{k}}}-\frac{L_0}{L_{\mathrm{d}}})/(1-\frac{L_0}{L_{\mathrm{d}}})}\end{array}}$$

where L_t is the conductivity of the frozen sample, L_k is the conductivity of the killed sample, L₀ is the conductivity of the non-frozen control, and L_d is the conductivity of the non-frozen control killed sample. We used an overnight −80 °C treatment to determine L_k and L_d because this temperature has been shown to be effective in determining maximum freezing injury (Lim et al., 1998; O’Connell and Savage, 2020; Savage et al., 2024b). The index of injury is bound between 0 and 1, with 1 indicating complete tissue death.

For each freezing trial, we cut at least four terminal branches with flowers from each plant. Samples were cut underwater and put in plastic floral tubes filled with tap water. Flowers were kept at room temperature (19 °C) for a couple of hours to allow the air-exposed parts of the flower to dry and minimize premature ice formation on the surface of the flowers. In the early spring, this temperature was warmer than the average outside temperature. All material was processed within 12 h of the start of the trial. After samples were given their respective treatments, any browning or visible damage to the flowers was noted (Supplementary Data Fig. S1). Then a set number of flowers were cut off the stem (between 1 and 20 depending on the species and flower size) and placed into 20 mL of Mili-Q water in a test tube and incubated in a shaking water bath held at 22 °C. After 24 h, the conductivity of the solution was measured using an Oakton PC 7000 (Environmental Express, Charleston, SC, USA). A round piece of mesh was placed in the tubes to keep the flowers submersed. Next, the samples were placed in the −80 °C treatment for ~18 h, incubated for an additional 24 h in a shaking water bath and measured a final time.

Pollen tube growth

We tested the impact of a hard freeze (−5 °C) on pollen tube growth in a subset of eight species (Table 1) using a previously established protocol (Wagner et al., 2016). We germinated pollen on small circular pads of medium (3% agarose, 0.02% boric acid, 10 mm potassium chloride, 10 mm calcium chloride dihydrate, and 2 mm magnesium sulfate, pH 8). Pads were formed by warming the medium and aliquoting an 80-μL drop into the lid of a 96-multiwell plate. After the pads solidified, they were transferred onto a microscope slide and pollen was applied to the surface of the pad either using a paint brush or by rubbing the anthers of the flowers directly on the pad. For each species, we prepared three pads per plant from seven plants per species. One pad was used as a control and kept at room temperature (19 °C). Two pads were subjected to the hard freeze treatment (as described above) after the pollen had started germinating (Supplementary Data Fig. S2). Of the two pads, one was removed after the samples had been held at −5 °C for 3 h and one was allowed to warm and recover after the treatment (30 h). The control treatment was allowed to run until the end of the recovery treatment. After each treatment, pollen tube growth was stopped by applying 80 % glycerol to the pads and covering them with a coverslip.

Pollen grains were imaged using a compound microscope and the length of the ten longest pollen tubes was measured on each pad. We checked whether pollen tubes were actively growing during our experiment by comparing the average length of pollen tubes on the pad removed right after freezing (L_F) with those that were collected after recovery (L_R). We excluded samples that did not grow between the end of the freeze treatment and the end of the other two treatments. To determine the impact of a hard freeze on pollen tube growth, we calculated the reduction (R) in pollen tube length in the recovery sample compared with the control sample (L_C) as a proportion of the total growth after the freeze treatment using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle R=\frac{L_{\mathrm{C}}-L_{\mathrm{R}}}{L_{\mathrm{C}}-L_{\mathrm{F}}}}\end{array}}$$

We also noted when there was a significant change in pollen tube growth between the freezing treatment and the recovery sample because this difference indicates when the freezing treatment terminated pollen tube growth.

Statistical analyses

We used ANOVAs to test for differences among species and t-tests to examine differences between plants that bloomed early in the season (before mid-May) and those that bloomed later in the season (mid-May to late July) at species level. Linear regression was used to compare traits. Male and female plants of Salix discolor and Acer negundo were treated like separate species in the analysis to capture differences in their floral traits. Both index of injury and water content data were arcsine transformed in all analyses. All analyses were completed using JMP Pro (ver. 17.0.0). All data for this paper is archived in the DRUM data repository (Savage et al., 2024a).

RESULTS

Freezing damage to floral tissue

The amount of floral damage after a hard freeze (−5 °C treatment) was not the same across species (ANOVA, F_26,243 = 35, P < 0.0001), with the earliest blooming plants exhibiting the least freezing damage based on electrolyte leakage data (Fig. 1A). There was also higher freezing damage in flowers that bloomed after 13 May (t-test, t = 8.6, d.f. = 18, P < 0.0001), which is the average last day when temperatures are ≤0 °C according to the last 15 years of weather data in Duluth, MN, USA (NOWData, NOAA Online Weather Data Resource). The average index of injury before 13 May was 18 ± 5 % (s.d.) and after this date it was 78 ± 19 % (s.d.). A similar pattern emerged from the visual assessment of floral tissue damage as only the late blooming species showed visible browning after the treatment (Fig. 1A). Damage to flowers was apparent in both display tissue (e.g. petals) and female reproductive organs (e.g. ovaries). An average index of injury >40 % corresponded to the presence of visual damage. We selected 14 species that exhibited damage in our hard freeze treatment and tested their ability to survive a light freeze (0 °C) treatment. After this treatment, the average index of injury was low, 8 ± 11 % (s.d.), and there was no evidence of visual tissue damage (Supplementary Data Table S1).

Fig. 1.

(A) Boxplot of floral damage after a hard freeze. Damage was assessed using electrolyte leakage and estimated with the index of injury. Species names are abbreviated using the first two letters of the genus and species. They are ordered based on when they bloomed in 2021. Male and female plants are noted with the letter M and F, respectively. Box colour indicates whether flowers exhibited visual damage. (B) Relationship between reduction in pollen tube growth after a hard freeze and day of the year of flower collection. The reduction in pollen tube growth is reported as a proportion of growth in the control treatment. Each point represents a species. The line is a best-fit linear regression of the data. The grey background delineates species that bloom after the average last day of freezing temperatures in the region (13 May).

Based on our thermocouple data, floral tissues reached an average minimum temperature of −4.6 ± 0.3 °C (n = 60) in our hard freeze treatment and 0.3 ± 0.1 °C (n = 6) in the light freeze treatment. Ice was visible in 83 % of the tubes holding flowers after the end of the hard freeze treatment (after samples were gradually rewarmed) but none of the samples in the light freeze treatment. We only successfully detected exotherms in about half our hard freeze samples because of challenges with keeping the thermocouples in contact with the flowers when they wilted. The average low-temperature exotherm recorded from the floral tissue was −3.5 ± 0.9 °C (n = 31). We did not detect any exotherms in the light freeze treatment.

Freezing-induced disruption of pollen tube growth

The hard freeze treatment led to decreased pollen tube growth in the subset of species that were studied. There was a significant relationship between the reduction in pollen tube length and flowering time (linear regression, F_1,6 = 13.3, P = 0.01, r² = 0.69, Fig. 1B), with less reduction occurring in species that flowered earlier in the growing season. Two of the later-flowering species exhibited no pollen tube growth after the hard freeze treatment (Rubus parviflorus and S. dulcamara). There was also a positive correlation between reduction in pollen tube growth and total floral tissue damage (linear regression, F_1,6 = 12.2, P = 0.01, r² = 0.67, Supplementary Data Fig. S3).

Variation in phenology from year to year

We found that plants that bloomed earlier in the year had a higher standard deviation in flowering date from year to year than plants that bloomed later in the growing season (linear regression, F_1,23 = 35.9, P < 0.0001, r² = 0.61, Supplementary Data Fig. S4). This variation was driven by two years, 2021 and 2022, the warmest and coldest years during this period. We used the difference in flowering time between these two years as an estimate of a species’ responsiveness to temperature. We found a significant relationship between responsiveness of flowering to temperature and flowering time, with species that flower earliest in the growing season being the most responsive to changes in temperature (linear regression, F_1,23 = 34.8, P < 0.0001, r² = 0.60, Fig. 2).

Fig. 2.

Relationship between estimated responsiveness of flowering to temperature and the first day of flowering. Responsiveness of flowering is determined as the difference in first flower between the coldest and warmest years in our study, 2022 and 2021, respectively. The grey background delineates species that bloom after the average last day of freezing temperatures in the region (13 May).

Relationship between floral freezing damage, phenology and floral/plant traits

There were large differences among species in their floral traits (e.g. inflorescence mass, surface area and water content; Table 2 and Supplementary Data Table S2), and distinct differences between species that bloomed early and late in the spring. First, most of the species that bloomed early in the year had catkins and/or unisexual and non-petalous flowers, while plants that bloomed later in the year tended to have bisexual and petalous flowers (Table 2). This shift in flower structure led to a significant difference between late- and early-blooming plants in their inflorescence surface area (t-test, t = 3.4, d.f. = 17, P = 0.004). Second, there was a significant difference between late- and early-blooming plants in terms of their height (t-test, t = 11.7, d.f. = 12, P = 0.004, Table 1) because tree species made up the entirety of the plants that bloomed early, while most late-blooming species were shrubs, vines or herbs. The main exceptions were Sorbus decora and T. americana, which were both trees that flowered after mid-May. Third, there was a decrease in floral tissue investment on a whole-plant basis throughout the spring across species, although it did not lead to a significant difference in total floral biomass between early- and late-blooming species (t-test, t = 0.31, d.f. = 17, P = 0.8). Seven out of the eight species that bloomed before mid-May produced >0.05 kg of dry floral mass per plant (average 0.4 ± 0.4 kg, n = 10 when male and female flowers of A. negundo and S. discolor were considered separately; Table 2). The one exception to this trend was C. cornuta, which produced only 0.0008 ± 0.0003 kg per plant. Of the plants that bloomed after mid-May, 15 of the 17 species had <0.05 kg of dry floral tissue. The two exceptions were the tree species, S. decora and T. americana. When these two species were excluded, the average floral investment per plant was 0.007 ± 0.008 kg per plant late in the season. There was no significant difference between early- and late-blooming species in their inflorescence dry mass or water content (t-test, α = 0.01, Supplementary Data Table S3).

Table 2.

Species floral traits. Quantitative traits are reported as average ± standard deviation.

Species	Sex	Petals	Inflorescence type	Infl. dry mass (g)	Flower water content (g g−1)	Inflorescence surface area (cm2)	Total flower mass per plant (kg)
Acer negundo	Female	No	Raceme	0.05 ± 0.04	0.76 ± 0.10	8 ± 2	0.1 ± 0.3
Acer negundo	Male	No	Raceme	0.06 ± 0.02	0.78 ± 0.04	12 ± 3	0.6 ± 0.9
Acer rubrum	Male	Yes	Fascicle	0.05 ± 0.005	0.76 ± 0.02	4 ± 1	0.6 ± 1
Acer saccharum	Male	No	Corymb	0.06 ± 0.01	0.68 ± 0.02	14 ± 3	0.4 ± 0.2
Acer spicatum	Bisexual	Yes	Panicle	0.08 ± 0.02	0.72 ± 0.01	27 ± 6	0.005 ± 0.003
Alnus incana	Male	No	Catkin	0.16 ± 0.05	0.40 ± 0.10	12 ± 2	0.2 ± 0.2
Amelanchier spicata	Bisexual	Yes	Raceme	0.05 ± 0.01	0.73 ± 0.03	43 ± 7	0.01 ± 0.01
Apocynum androsaemifolium	Bisexual	Yes	Cyme	0.04 ± 0.002	0.85 ± 0.02	4 ± 1	2E-4 ± 2E-4
Betula papyrifera	Male	No	Catkin	0.17 ± 0.05	0.71 ± 0.02	15 ± 3	1.3 ± 0.7
Clematis virginiana	Male	Yes	Panicle	0.10 ± 0.01	0.84 ± 0.02	55 ± 8	4E-4 ± 4E-4
Cornus rugosa	Bisexual	Yes	Cyme	0.25 ± 0.05	0.77 ± 0.01	80 ± 10	0.007 ± 0.007
Cornus sericea	Bisexual	Yes	Cyme	0.13 ± 0.04	0.77 ± 0.02	39 ± 5	0.004 ± 0.005
Corylus cornuta	Male	No	Catkin	0.02 ± 0.004	0.47 ± 0.06	3.1 ± 0.5	8E-4 ± 3E-4
Diervilla lonicera	Bisexual	Yes	Cyme	0.04 ± 0.004	0.82 ± 0.03	9 ± 1	3E-4 ± 2E-4
Populus tremuloides	Male	No	Catkin	0.10 ± 0.02	0.60 ± 0.10	31 ± 6	0.5 ± 0.3
Prunus americana	Bisexual	Yes	Umbel	0.02 ± 0.002	0.79 ± 0.02	14 ± 3	0.007 ± 0.005
Prunus virginiana	Bisexual	Yes	Raceme	0.14 ± 0.04	0.79 ± 0.05	30 ± 10	0.02 ± 0.02
Rosa blanda	Bisexual	Yes	Singular	0.08 ± 0.01	0.77 ± 0.02	34 ± 6	4E-4 ± 3E-4
Rubus parviflorus	Bisexual	Yes	Corymb	0.06 ± 0.01	0.84 ± 0.02	150 ± 40	3E-4 ± 1E-4
Salix discolor	Female	No	Catkin	0.05 ± 0.01	0.70 ± 0.10	16 ± 5	0.2 ± 0.2
Salix discolor	Male	No	Catkin	0.08 ± 0.02	0.77 ± 0.02	17 ± 3	0.06 ± 0.02
Sambucus racemosa	Bisexual	Yes	Cyme	0.4 ± 0.2	0.86 ± 0.01	140 ± 40	0.03 ± 0.01
Solanum dulcamara	Bisexual	Yes	Cyme	1.1 ± 0.4	0.78 ± 0.01	200 ± 30	0.3 ± 0.2
Sorbus decora	Bisexual	Yes	Cyme	0.14 ± 0.06	0.86 ± 0.01	38 ± 6	9E-4 ± 3E-4
Tilia americana	Bisexual	Yes	Cyme	0.9 ± 0.5	0.65 ± 0.02	10 ± 2	9 ± 5
Viburnum lentago	Bisexual	Yes	Cyme	0.27 ± 0.08	0.75 ± 0.02	52 ± 7	0.01 ± 0.01
Viburnum trilobum	Bisexual	Yes	Cyme	0.34 ± 0.09	0.80 ± 0.10	80 ± 30	0.01 ± 0.01

On a species level, there was no relationship between the index of injury in the hard freeze treatment and inflorescence dry mass (linear regression, F_1,25 = 1.2, P = 0.3) or inflorescence surface area (linear regression, F_1,25 = 1.2, P = 0.3). In contrast, floral water content was positively correlated with the index of injury (linear regression, F_1,25 = 8.4, P = 0.008, r² = 0.25, Supplementary Data Fig. S5A). However, this relationship was strongly influenced by two species with catkins (A. incana and C. cornuta). When these species were excluded from the analysis, the relationship was not significant (linear regression, F_1,23 = 4.3, P = 0.05, Supplementary Data Fig. S5B).

DISCUSSION

Previous studies have suggested that there could be a relationship between floral freezing tolerance and flowering time (Kader and Proebsting, 1992; Miller-Rushing and Primack, 2008; CaraDonna and Bain, 2016), but this is one of first studies to document the relationship across a large group of species (Fig. 1). This relationship is very similar to what is seen in leaves where greater freezing damage is observed in species with later leaf-out (Lenz et al., 2013, 2016; Savage and Cavender-Bares, 2013; Vitasse et al., 2014). The relationship in leaves is often attributed to a trade-off between freezing tolerance and growth (Koehler et al., 2012; Savage and Cavender-Bares, 2013). The thought is that freezing tolerance should be costly and require delayed leaf-out, resulting in less growth during the growing season (Loehle, 1998; Gömöry and Paule, 2011; Vitasse et al., 2014) and ultimately lower reproductive output (Boinot et al., 2022). As a result, natural selection should favour high freezing tolerance in leaves that are produced when temperatures are cold but favour low freezing tolerance later in the spring. It is possible that a similar trade-off exists in flowers, especially if the trade-off is driven by the impact of cold temperatures on critical processes like cell wall growth (Kutsuno et al., 2023) and xylem transport (Davis et al., 1999; Hacke and Sperry, 2001; Pittermann and Sperry, 2003; Savage et al., 2022). It is also possible that there is natural selection for coordination in freezing tolerance and de-acclimation across organs in a plant.

When our results are considered in the context of false springs, it is clear that the predicted impact of warming will not be consistent across species. False springs often occur because warm temperatures trigger premature flowering at a time when there is still risk of freezing. Similar to other studies (Ho et al., 2006; Munguia-Rosas et al., 2011), we found that plants that flower earlier in the growing season are more likely to flower prematurely in a warm year than species that flower later in the growing season (Fig. 2). If these species demonstrate earlier flowering as temperatures continue to warm, the probability that they will experience freezing temperatures may increase. For species that currently flower when there is a risk of freezing, this may not be an issue because they are already freezing tolerant. However, species that flower mid-season (mid to late May) that have a low freezing tolerance might be forced to flower when there is still a risk of freezing damage. As a result, these species may be more vulnerable to false springs than those that flower earlier and later in the year.

The male and female gametophyte are both impacted by freezing temperatures

Most research on freezing tolerance has emphasized that ovaries (which contain the female gametophyte) are often the most sensitive part of the flower to freezing (Neuner et al., 2013; Kaya and Kose, 2019; Hillmann et al., 2021) but few studies have compared the impact of freezing temperatures on the female and male gametophyte (pollen). It is known that pollen tube germination and growth can be impacted by temperature (Zamir et al., 1981; Weinbaum et al., 1984) and that pollen’s functional temperature range varies depending on the climate and possibly flowering time (Kakani et al., 2002; Steinacher and Wagner, 2013; Rosbakh and Poschlod, 2016; Wagner et al., 2016). In our study, a hard freeze impacted both the male and female gametophytes of flowers that bloomed late in the season. In the female gametophyte there was tissue damage and visible browning, and in the male gametophyte there was a decrease in pollen tube growth. In both situations, the impact of freezing was related to flowering time. What remains unclear is the extent of the observed reduction in pollen tube growth on reproduction in the study species. In vitro pollen tube growth experiments are a great way to examine how pollen tube growth is affected by environmental factors, but they do not always reflect the growth observed in vivo (Kato et al., 2022). More work is needed to determine if the observed reduction in pollen tube growth could compromise fertilization for the species in our study, thus, testing whether pollen tube growth and ovary damage could co-limit or differentially impact plant reproduction after a hard freeze in our system.

Floral structure is connected to freezing tolerance and pollination syndrome

There are many factors that can influence the freezing tolerance of plant organs, from the presence of apoplastic ice barriers to cellular level changes in membrane integrity and cell wall elasticity (Xin and Browse, 2000; Kuprian et al., 2014; Zhang et al., 2016). One factor that has been associated with high freezing tolerance is low organ water content (Weiser, 1970). Many plant organs lose water during cold acclimation (Savage, 2019; Hillmann et al., 2021; Savage and Chuine, 2021). Low water content is beneficial under freezing temperatures because it can result in increased viscosity and a higher solute concentration inside cells (Wolfe et al., 2002). These two changes can lead to greater supercooling and less damage from dehydration during freezing, respectively. There is also evidence that apoplastic water potential has a strong impact on the temperature when ice is nucleated in a plant. Therefore, tissue with a low water content and more negative water potential will freeze at a lower temperature than tissue with a high water content and less negative water potential (Lintunen et al., 2018), which could be important in minimizing ice propagation into living tissue (Kaku et al., 1980; Neuner et al., 2019). Previous research has shown a negative relationship between floral water content and freezing tolerance during floral de-acclimation in the spring (Kader and Proebsting, 1992; Rinne et al., 1994; Hillmann et al., 2021) and at flowering across species (Miller-Rushing and Primack, 2008). We found a similar relationship in our data, but it was strongly impacted by low-water-content catkin-bearing species.

The coordination of different floral types with measured traits like water content make it challenging to determine whether specific features impact floral freezing tolerance. The two largest morphological differences associated with freezing tolerance in our study were the presence of petalous structures in flowers and inflorescence surface area. All flowers that survived a hard freeze were non-petalous (Table 1). While it is possible that this feature is important for floral freezing tolerance, it might also be a side effect of the relationships between freezing tolerance, flowering time and pollination syndrome (flower traits associated with specific pollinators). Generally, wind-pollinated plants tend to flower early in the growing season (Heinrich, 1976; Kjell et al., 2003) when the canopy is open and pollen is easy to disperse (Rathcke and Lacey, 1985; Gougherty and Gougherty, 2018). These plants often have non-petalous flowers. Wind pollination is also common among many of our native tree species, and this could explain the relationship between floral freezing tolerance and plant height. Therefore, it remains unclear whether any of the noted differences between more and less freezing-tolerant species described above directly impact freezing tolerance or are a side effect of the relationship between freezing tolerance and flowering time.

Questions that remain about how warming will impact plant reproduction

A key factor that our study and other studies have emphasized when considering the impact of freezing temperatures on plants is the importance of the timing of freeze events (Proebsting and Mills, 1978; Salazar-Gutiérrez et al., 2014; Ishikawa et al., 2015). Flowers tend to be vulnerable to freezing temperatures after anthesis (Ashworth, 1984), but damage can occur during many stages of floral and gametophyte development (Hedhly, 2011). Depending on when a freezing event occurs, it could have a very different impact on plant reproduction. There are many phenological events that need to line up for successful fertilization and seed production. For example, the stigma needs to be receptive when pollen is dispersed. Previous work has shown that temperature changes can lead to a mismatch in these (Hedhly et al., 2004) and other important events, including synchrony between pollinators and flower anthesis (Rafferty and Ives, 2011; Kudo and Ida, 2013; Kharouba et al., 2018). It is also possible that there are interactions between freeze-damaged flowers and pollinators. For example, in a study on Prunus pumila, we found that insect preference towards viable flowers changed when a plant experienced a hard freeze event (Lake Diver, 2022). Taken together, these studies show that even when there is minimal tissue damage to flowers, like on the plants that bloom early in the season, there is still a chance that freezing temperatures might decrease plant reproduction. As a result, the current study serves only as a first step in understanding the potential impact of a false spring on plant reproduction in our region.

Conclusions

In this study, we found that a hard freeze can cause significant floral tissue damage and reduce the function of the male gametophyte in some species in our region. The most susceptible plants to late freeze events are those that flower when the risk of freezing is low. These plants often flower earlier in warmer years, and as winters continue to warm these species could be in danger of losing flowers to a false spring. The impact of floral loss during a false spring is largely determined by the frequency of these events. When false springs occur irregularly, as is observed under current climate conditions, the long-term impact of floral damage on perennial plants might be limited, but if the frequency of these events increases there could be a cumulative negative impact on plant fitness. Therefore, in regions that are experiencing high rates of winter and spring warming, attention should be paid to the vulnerability of flowers to false springs, because it could have a lasting impact on future forest recruitment.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following. Figure S1: damage to flowers resulted in browning and wilting of floral tissue. Figure S2: examples of non-germinated (top row) and germinated pollen (bottom row) from Rubus parviflorus and from Corylus cornuta. Figure S3: relationship between floral freezing damage after a hard freeze (index of injury) and reduction in pollen tube growth after a hard freeze. Figure S4: relationship between standard deviation of first flower date and first flower date. Figure S5: relationship between index of injury and water content across all species. Table S1: floral damage after light freeze. Table S2: statistical analysis of species level differences in floral traits. F-statistic and P-value are reported for each ANOVA. Table S3: statistical analysis of traits in early- and late-blooming species.

FUNDING

This work was supported by the National Science Foundation (IOS1656318) and the Office of the Vice President for Research, University of Minnesota (GIA 322644) to J.A.S.