Seeds of the long pod (LP) and the short pod (SP) forms of D. unguis-cati were collected during the fruiting months of 2013 from various sites around the greater Brisbane area in southeast Queensland, Australia. SP seeds were obtained from the following infestation sites: South Bank (27˚55'S, 153˚01'E), Ipswich Forest Reserve (27˚32'S, 152˚42'E), Chelmer (27˚47'S, 152˚58'E), Bardon (27˚30'S, 152˚41'E) and Boonah (27˚60'S, 152˚41'E). LP seeds plants were collected from Carindale (27˚30'S, 152˚41'E), Bardon (27˚30'S, 152˚41'E) and Sherwood (27˚30'S, 152˚59'E). Fewer sites were sampled for LP because this form is less prevalent than SP, and (and consequently fruit formation) does not occur every year [32] . Seeds were collected from fruits at the time of dehiscence to maturity of the seeds [24] . Once collected, seeds were stored for two weeks at room temperature in paper envelopes placed in containers with silica gel to ensure they remained dry before the commencement of germination assays. For the purposes of this experiment, seeds from different sites were pooled together to ensure adequate sample sizes, although we recognise that there could be differences in germination behaviour between individual plants and amongst sites for a given form.

Seed germination: Seeds were physically screened to ensure they were firm and intact. Those that appeared not to have viable embryonic content and/or were damaged by insects were not included in the germination assays. Seeds were sterilised by soaking in 1% sodium hypochlorite (NaOCl) for 5 minutes, then rinsed in water for 3 minutes [14] . Sterilised seeds were placed in 15 cm diameter petri dishes lined with 2 - 3 layers of 15 cm Whatman filter paper (No. 1) moistened with distilled water. They were subsequently exposed to varying temperature regimes in growth chambers (model ADAPTIS A1000; Conviron Ltd., USA) at the Queensland University of Technology (QUT) in Brisbane, QLD-Australia. Germination conditions were set to 1) cool (10/20˚C), 2) moderate (15/ 25˚C), 3) warm (20/30˚C) and 4) hot (30/45˚C) temperature regimes for 12 hours at each alternate temperature. Seed germination took place in light/dark conditions (12-hour photo-period) or constant 24 hours dark. When a 12-hour photo-period was applied, the higher temperature corresponded with the presence of light. Most of the temperature regimes followed the conditions applied by Vivian-Smith and Panetta [35] and also reflect the night-day temperature fluctuations in the distribution range of D. unguis-catiin Australia. Two additional treatments included constant room temperature (25˚C) with 5) 12 hour light/dark and 24-hour dark. Fifteen (15) replicates of 20 seeds each were used for each pod form in each treatment. Germination data were recorded every seven (7) days for 12 weeks after which the assay was terminated due to logistic reasons, and because no more appreciable germination was observed. After the 12^th week, the ungerminated seeds were physically checked and found to be mostly rotten with no visible viable embryo; testing the embryo of ungerminated seeds confirms loss of viability, except for those in the low temperature (10/20˚C). A dim light was used to examine the seed germination in the continuous darkness treatment.

Each seed was considered to have germinated with the emergence of one or more radicles [15] . Total germination percentage was calculated from the total number of germinated seeds divided by the total number of seeds. Germination rate index (GRI) was calculated following the Equation (1) of Maguire [36] ,

where N_i represents germinated seeds on the i^th day and D_i is the number of days from the commencement of the germination assay to the i^th day (see also [14] ). GRI was determined at six and 12 weeks for each treatment. Cumulative mean germination data (%) were plotted against time (weeks) from which the following indices were extracted: time-to-initiation of germination (T₁), time to 50% (T₅₀), and time to maximum germination (65% (T₆₅)) in this study (see [13] ).

If only one radicle emerged during germination, seeds were considered mono-embryonic, but if two or more radicles were observed then the seeds were considered polyembryonic [20] . Total percentage polyembryony was calculated from the proportion of seeds showing two or more radicles at germination for a given treatment. Seeds showing polyembryony were grouped into classes (twins, triplets and quadruplets) depending on the number of radicles that emerged during germination [19] .

Confirmation of polyembryony in the two forms of D. unguis-cati was further determined through seedling establishment [19] [20] . After germination, polyembryonic siblings with at least one pair of leaves were taken out of the germination petri-dishes and transferred into 20 cm plastic pots filled with locally available commercial soil (Osmocote Multi-Purpose Potting Mix with trace elements). Seedlings were left to grow in a light environment (range: 60 - 250 µmol∙m⁻²∙s⁻¹) over a 1 year period to ascertain if the individual seedlings would establish independent of other siblings, and whether each sibling would develop separate roots and tubers or not. The same process was repeated for polyembryonic seedlings in which the siblings were physically separated.

Germination percentage data were arcsine transformed before analysis to improve normality of residuals. Analysis of variance (ANOVA) was used to test the effects of light, temperature regimes, and pod form of D. unguis-cati, as well as their interactions on germination indices and occurrence of polyembryony. The form of D. unguis-cati, light, and temperature regimes were used as fixed effects on the ANOVA model. A Tukey HSD post-hoc test was performed to assess the germination and polyembryony differences between the temperature treatments. When no significant interactions were detected, a Pearson’s c² statistical test was used to compare the frequency of polyembryonic seeds of LP and SP plant forms. All statistical tests were carried out at α < 0.05 using the R statistical program on R version 3.1.0 [37] .

At all the temperature regimes, the seeds of the short pod (SP) plant form showed rapid germination whilst those of the long pod (LP) form were gradual and slower (Table 1; Figure 1). Across the temperature range and photoperiod cycles tested, it took seeds of SP plants an average of 11.5 days to initiation of germination (T₁), except for the cool 10/20˚C regime where up to 28 days was required to T₁ (Table 1). It took seeds of LP plants significantly longer period (an average of 20.2 days) to initiation of germination (T₁) across temperatures, other than 10/20˚C in which no germination was recorded (Figure 1(a)). The inhibitory effect of low temperatures (10/20˚C) on germination was more pronounced on seeds of the LP form than those of SP plants, since the former did not germinate at all at this temperature. When the experiment was terminated, seed germination in SP at 10/20˚C had reached about 42.9% (Figure 1(a)). However, the gradient of the curve indicate that had the experiment continued, more seeds would have germinated with time, and therefore a higher total germination percentage could have been attained at this temperature too.

Time to 50% germination (T₅₀) was also significantly lower for seeds of SP plants (range: 14 - 34 days) when compared to those of LP plants (range: 30 - 84 days) across the temperature regimes tested (Table 1). Light did not have any significant effect on T₁, T₅₀ or T₆₅ for either LP or SP plants at 20/30˚C (Figure 1(c)), but that was not the case at constant 25˚C for LP plants (Figure 1(d)). Light had a significant germination effect on T₅₀ on LP plants at 25˚C, but not on seeds of SP plants. At 25˚C, it took LP 30 days to reach T₅₀ under light conditions but >84 days to reach T₅₀ in constant darkness (Table 1 and Figure 1(d)). In contrast, at 25˚C, T₅₀ for SP was 14 - 16 days, irrespective of the light conditions.

The rate (speed) of germination estimated by Maguire’s germination rate index (GRI) showed significantly higher germination rates for seeds of SP plants than LP plants at all temperature regimes (Figure 2). The magnitude of the difference in germination rate between the two pod forms was also greater at 6 weeks compared to 12 weeks, especially for the moderate (15/25˚C) and warm (25˚C and 20/30˚C) temperature regimes (Figure 2(a), Figure 2(b) and Table 2). At 6 weeks, GRI values for all interaction effects (pod form x light; pod form x temperature, light x temperature, and form x light x temperature) were significant (P < 0.05; Table 2), suggesting that GRI response values for seeds of each plant pod form varied significantly depending on light and/or temperature conditions. In contrast, at 12-weeks, only the interaction effect of form x light was significant (F_{4, 218} = 9.05; P < 0.0001) (Table 2), implying a greater role of light than temperature regime on this germination index.

Polyembryony occurs in both plant forms of D. unguis-cati as demonstrated by emergence of two or more radicles from individual seeds during germination (Figure 3(a)). However, there was a significant difference in the

frequency of polyembryony between LP and SP plants (χ² = 71.730, df = 1, p < 0.002). SP plants displayed a significantly higher frequency of polyembryony than LP plants (SP plant: 38.52% ± 2.74%; LP plant: 4.68% ± 1.13%) (see Figure 4 and Figure 5). The polyembryonic seedlings lack any connections and were easily separated from each other, with each shoot system detaching with a corresponding radicle or root system (Figure 6).

Three classes (twin, triplets and quadruplets) of polyembryonic seedlings were observed in this study (Figure 3). All classes were observed in SP: single (60.4%; N = 628/1040), twin (25.5%; N = 265/1040), triplet (11.7%; N = 122/1040) and quadruplet (2.01%; N = 21/1040) (Figure 4). All polyembryonic seedlings in LP were twins, and constituted only 4.68% (N = 42/897) of germinated seeds (Figure 4).

Polyembryonic seedlings were easily separated from each other, with each shoot system detaching with a corre

sponding radicle or root system (Figure 6). Separated polyembryonic seedlings were able to develop and establish individually when transplanted into growth media and developed their own root system with tubers (Figure 3(c)). When left intact and transferred to growth media, polyembryonic siblings still developed individually with independent root and tuber systems. We also observed that for each set of polyembryonic siblings, one seedling was more robust and had a larger subterranean tuber than that of other siblings (Figure 3).

This study established that SP plants exhibited higher germination indices and higher frequency of polyembryony than LP plants, irrespective of temperature regimes (Figure 5). There were no significant interaction effects of temperature and/or light regime on the occurrence, frequency or classes of polyembryony (temperature x plant form: F-ratio = 0.594, df = 1, p = 0.443; light x plant form: F-ratio = 0.021, df = 1, p = 0.886), implying that these environmental resources did not influence the dynamics of polyembryony. Nonetheless, both LP and SP plants showed their highest frequency of polyembryony (15.9% and 41.2%, respectively) at 15/25˚C (Figure 5). At 30/45˚C, only a few incidents (0.3%) of germination (and no incidence of polyembryony) was observed for SP and no germination whatsoever for LP.

Our results suggest that germination niche requirements of SP are broad and non-specific, while LP germinates optimally only under warmer temperature conditions (20/30˚C and 25˚C; Figure 1; Table 1). Flexible germination cues enhance the invasive capacity of plants by enabling them to spread and establish in novel climatic conditions of recipient communities [38] [39] . SP plants had a much higher germination rate (compared to LP plants) at the cooler temperature regime of 10/20˚C. Equally, although of low frequency, seeds of SP plants showed evidence of germination incidents at hot, 30/45˚C temperature regime while there was no germination at all for LP plants under this scenario. These wider germination amplitudes may indicate greater resilience in SP seeds, and may suggest a potential for this form of D. unguis-cati to spread further into the cooler state of New South Wales and Victoria as well as into the warmer/hotter areas of Australia (e.g., Northern Territory and western Queensland), especially under a climate change scenario. In general, the rapid germination behaviour of seeds of SP seeds (Figure 1) is typical of invasive species [38] [40] . Whilst the longevity of seeds of SP seeds is low (<12% by 1 year) after dispersal [35] , its rapid germination under a wider range of temperatures may confer a fitness advantage in terms of seedling establishment and spread.

The higher frequency of polyembryony in seeds of SP plants could also allow it to proliferate successfully in a variable environment in contrast with those of LP plants (Figure 4 and Figure 5). Thus for the same number of initial seeds introduced in a new environment, there is a greater likelihood that more SP rather than LP plants would establish. Considering that twins, triplets and quadruplets occur in polyembryonic SP seeds, while only twins occur in LP (and at a lower frequency), it can be assumed that a higher propagule pressure would be exerted by SP than LP plants upon introduction. Polyembryony may also increase invasiveness potentials by the bet-hedging strategy, which ensures that at least one individual seedling from a polyembryonic seed survives [41] [42] . Although mature LP fruits are known to have twice as many seeds per fruit as SP fruits [32] , higher germination and polyembryony exhibited by SP may be likely to increase propagule pressure leading to SP plants being the more invasive form than LP. Propagule pressure has previously been correlated with plant invasiveness [43] - [45] . Nonetheless, caution is needed in the interpretation of this finding because the polyembryony phenomenon may not necessarily be adaptive as it puts siblings into direct competition for environmental resources [18] [42] .

Some evidence suggests that polyembryony reduces seed germinability significantly [19] , but our results do not support this position. SP consistently had higher germination rates than LP at all temperature regimes whilst also exhibiting higher frequency of polyembryony than LP. Temperature did not significantly affect expression of polyembryony, suggesting a genetic basis for the phenomenon [46] , but both LP and SP appeared to show relatively higher polyembryony frequencies at 15/25˚C (their optimal growth condition) than at other temperature regimes. Our preliminary observations of 1-year old siblings from polyembryonic seeds indicate equal survival rates, but slower mean growth rates per individual as the number of siblings increase (Figure 3; JC Buru, unpublished data). However, it remains to be seen how differences in germination and polyembryony rates will translate to initial biomass gain per individual, and ultimately offspring fitness.

The differences in the invaded range distributions of the two forms could be a product of colonization events, with the LP form being a more recent arrival in Australia compared to the SP form. There may also be an underlying genetic basis responsible for the observed pattern [47] . Although largely untested, and therefore speculative, the prevalence of SP in Australia may potentially be a classic case of (i) recent whole genome duplication followed by diploidization during which genes are lost/modified/rearranged [48] [49] , and/or (ii) release from natural enemies in the novel environment. Both scenarios have the tendency to increase competitiveness, niche pre-emption, and ultimately spread and distribution [5] [50] [51] .

The differences in germination dynamics and frequency of polyembryony between SP and LP plants lends further credence to suggestions that the two forms of D. unguis-cati could be two extremes of the same species or even different species [33] . Anecdotal evidence suggests that the predominant form of D. unguis-cati in the native range is similar in appearance to the form being referred to as LP plants in Australia (Dhileepan K. personal comm.). Interestingly, an earlier study conducted within the native range found that D. unguis-cati did not exhibit any polyembryony [20] . This is in sharp contrast to what we have observed in SP plants (40%), but is closer to our results for LP plants (<5%). Whether the two forms are different species, products of two independent introductions from the native range of the weed, or arose from (auto-/allo-) polyploidy remains to be determined. Only a comprehensive phylogenetic study of the different forms of the species and other members of the genus Dolichandra will help clarify the status of the two forms (but see [34] ). Moreover, work on chromosome number (karyotype), level of polyploidy (a trait known to correlate significantly with polyembryony [20] [47] ), as well understanding the interplay of their breeding systems (sexual vs. asexual) variation [31] , will also help our understanding of their invasiveness potential.