Fieldwork on Apoballis mutata was carried out along the forest edge adjoining farmland in Cameron Highlands, Pahang, Malaysia. The study population comprised approximately 150 mature terrestrial plants, typically occurring in mesophytic habitats and occasionally near muddy stream banks. Observations on flowering biology were conducted between February and April 2016, with peak flowering likely in late February, as inferred from the abundance of developing infructescences in March.

Anthesis was recorded and floral visitors of A. mutata were observed during the pistillate, transitional, and staminate phases. Detailed observations focused on floral traits such as spathe movements, the presence or absence of perceptible floral scent, and the behaviour of pollinators and other floral visitors. These qualitative observations were conducted during field visits across all three floral phases in the 10 inflorescences monitored for flowering and pollinator behaviour. Particular attention was given to whether spathe constriction or staminode expansion occurred during anthesis, based on qualitative visual observations. Spathe opening was measured only at the onset of anthesis, while subsequent changes in spathe and staminode form were monitored observationally, as these features in some aroids restrict insect movement.

In total, 59 inflorescences were randomly selected from different individuals, which were temporarily marked to ensure that each plant was sampled only once. Of these, 10 were monitored for flowering and pollinator behaviour during both daytime and nocturnal field observations, while 49 (27 at pistillate anthesis and 22 at staminate anthesis) were bagged for insect sampling. During this study, brief field observations indicated approximately equal occurrence of creamy green and creamy melon-red spathe morphs within the population. For each morph, insect visitation rates were recorded and analysed using generalized linear models equivalent to a two-way ANOVA in R v.4.5.0 (R Core Team 2025).

To test whether the upper spadix alone attracted the same insect species, three pistillate-phase inflorescences had their upper spadices and spathe limbs excised and placed approximately 100 m from the mother plants. All insects collected from the bagged inflorescences and excised spadices were preserved in 70% ethanol and identified to species level. Fly specimens were sorted into morphotypes and gender and compared to the reference material and descriptions in the DrosWLD database (https://evolgen.biol.se.tmu.ac.jp/DrosWLD/modules/stdb/). Undescribed morphotypes were assigned tentative species names (see also Sultana et al. 2006; Fartyal et al. 2013; Toda unpublished database) pending formal taxonomic treatment.

Infructescence development was monitored in 20 post-staminate inflorescences, comprising 10 from the intensively observed group and 10 randomly selected in the field, each from a separate individual, until seed dispersal. For seed germination trials, 600 seeds taken randomly from three infructescences were cultivated on cotton wool in sealed Petri dishes.

Because flowering and fruiting were not synchronous, reproductive success was assessed using separate sets of individuals: 20 pistillate inflorescences and 20 infructescences. Fruit set was calculated as the ratio of berries to pistillate flowers. As pistillate flower and berry counts were obtained from different individuals, calculated fruit set values could exceed 100%; therefore, fruit set values were normalised to a maximum of 100% to facilitate comparative analyses across individuals. Seed counts per berry and infructescence were not determined because the infructescences were only partially mature. All collected flowers and infructescences were preserved in 70% ethanol.

Five randomly selected staminate flowers of A. mutata from the 49 bagged inflorescences were dissected under a stereomicroscope and dehydrated through an ethanol series to absolute ethanol (Low et al. 2016). Samples were transferred to pure acetone and dried using a critical point drier (BALTEC CPD030) with CO₂ as the transitional medium. Dried specimens were mounted on aluminium plate, sputter-coated with gold using an Auto Fine Coater (JEOL JFC-1600), and imaged with a JEOL JSL-6390LA SEM (JEOL, Peabody, MA, USA). Images were captured at 10 kV accelerating voltage, with a working distance of 10–13 mm and magnifications ranging from 30× to 10,000×.

To visualise pollen distribution on pollinators, two to three randomly selected specimens of Colocasiomyia spp. collected at pistillate and staminate anthesis were similarly mounted, gold-coated, and scanned under SEM.

The synflorescence of Apoballis mutata produced up to four sequential inflorescences blooming consecutively. Inflorescences were borne on erect peduncles; and the spathe showed a constriction aligned with the sterile zone between the pistillate and staminate flowers zones. The spathe comprised an upper spathe limb and a lower pollination chamber (Fig. 1D, E). Two spathe colour morphs occurred at roughly equal frequency within the population: creamy green and creamy melon-red, which matched the petiole colour of the respective plant. In the creamy green morph, prominent white, stalk-like staminodes larger than the pistils were interspersed among the pistillate flowers (Fig. 1B). These staminodes varied in form between clavate to mushroom-shaped or globose, were arranged 3–4 rows above the pistillate zone. In contrast, the creamy melon-red morph lacked staminodes within the pistillate region, showing only a partial row above it (Fig. 1C). Both morphs contained a naked sterile interstice with scattered atypical bisexual flowers (ABF), usually occurring in cluster of 2–4 flowers. The staminate zone consisted of dense stamens with truncated, rectangular to dumbbell-shaped surfaces and slightly grooved connectives. The spadix terminated in an appendix bearing irregular polygonal staminodes (Fig. 1B, C).

Stigma receptivity in Apoballis mutata typically began around 08:30 (local time, UTC +8.00), coinciding with initial spathe opening (approximately 5 cm × 5 mm) at the intermediate spadix zone (Fig. 1D–F). This marked the onset of the pistillate phase. During this period, the appendix emitted a sweet fruity scent that may help to attract pollen-bearing Colocasiomyia flies. These flies entered along the spathe margin into the lower chamber, where they contacted receptive pistillate flowers (Fig. 1D–G). Occasionally, the Colocasiomyia flies were observed feeding on the inner spathe surface, possibly on secretions of tissues. Cycreon beetles were also present (Fig. 1H), but they remained on the spathe margin or upper inner spathe and did not access the pollination chamber. By the afternoon the transitional phase began, stigma receptivity was lost, and floral scent diminished, and Colocasiomyia flies moved upward to the inner spathe surface. No spathe constriction or staminode expansion were observed during the pistillate or transitional phases; the spathe remained inflated and open, allowing free insect movement.

Anthesis of A. mutata lasted approximately 26 h. The staminate phase began on the following morning. Around 08:00, the spathe began to tighten (Fig. 1I, J), accompanied by faint scent release, and pollen dehiscence commenced from the staminate flowers. At this stage, Colocasiomyia flies congregated near the staminate zone or on the outer spathe, where powdery pollen was released from the thecae pores, covering their bodies (Fig. 1I, J).

After anthesis, the upper spathe and spadix abscised within five days, initiating fruit development within the lower spathe (Fig. 1K). Berries of approximately 2 mm in length matured over approximately one month, turning from light green to light yellow as the spathe split open (Fig. 1L).

Randomly selected seeds of A. mutata were cultivated immediately after infructescence dehiscence on cotton wool in sealed Petri dishes (Fig. 1M). Germination began after three weeks (Fig. 1N), indicated by the emergence of the first leaf. The majority of the seeds germinated successfully, indicating high viability.

From the 49 bagged A. mutata inflorescences, a total of 3041 Colocasiomyia sp. 2 aff. bogneri (mean ± SD: 62.06 ± 38.65 individuals per inflorescence; Fig. 2B), 320 C. sp. 17 aff. bogneri (6.53 ± 6.66; Fig. 2C), 729 C. sp. 34 aff. bogneri (14.86 ± 10.52; Fig. 2D), and 2 C. sp. 1 aff. bogneri (0.04 ± 0.20; Fig. 2A) were collected (Table 1; Suppl. material 1). The overall sex ratio of Colocasiomyia spp. was approximately 1:1 (1919 ♀; 2173 ♂; Fig. 2G; Table 1). Based on counts summarised in Table 1, C. sp. 2 aff. bogneri occurred in comparable proportions during pistillate (53.24%) and staminate (46.76%) phases. In contrast, C. sp. 17 aff. bogneri and C. sp. 34 aff. bogneri declined markedly from the pistillate to the staminate stage: C. sp. 17 aff. bogneri from 79.69% to 20.31%, and C. sp. 34 aff. bogneri from 74.86% to 25.24%. Colocasiomyia sp. 1 aff. bogneri was incidental, represented by a single individual at each anthesis phase. Per-inflorescence abundances across species, anthesis phases, and spathe colour are illustrated in Fig. 2E and are consistent with these patterns. Three dissected pistillate-phase inflorescences showed the predominance of C. sp. 2 aff. bogneri both in the lower chamber and in the upper spathe (Table 1). Cycreon beetles were rare visitors (38 in total; Table 1; Fig. 2E; Suppl. material 1).

Visitation patterns to the two spathe colour morphs showed minimal differences across anthesis phases. In creamy melon-red spathes, C. sp. 2 aff. bogneri was abundant and occurred in broadly comparable proportions during pistillate (1172 individuals; 58.08%) and staminate (846 individuals; 41.92%) phases (Fig. 2E; Table 2; Suppl. material 1). Colocasiomyia sp. 17 aff. bogneri (79.82%) and C. sp. 34 aff. bogneri (77.71%) both decreased substantially in the staminate phase (20.18% for C. sp. 17 aff. bogneri, and 22.29% for C. sp. 34 aff. bogneri) irrespective of spathe colour (Fig. 2E; Table 2; Suppl. material 1). In creamy green spathes, C. sp. 2 aff. bogneri increased slightly from pistillate (447 individuals; 43.70%) to staminate (576 individuals; 56.30%) phases. Generalised linear models revealed no significant effects of spathe colour on visitation rates for any Colocasiomyia species or Cycreon beetles (all p values > 0.05; Fig. 2F). Predicted mean numbers of insect visitors were similar between creamy melon-red and creamy green spathes at both anthesis phases, indicating that spathe colour does not significantly influence Colocasiomyia abundance in A. mutata.

Reproductive success was assessed using separate sets of individuals, comprising 20 pistillate inflorescences and 20 infructescences. Across 20 pistillate inflorescences, a total of 7746 pistillate flowers were counted, with a mean 387 ± 82 (SD) pistillate flowers per inflorescence. Twenty infructescences yielded 7164 berries, averaging 358 ± 73 (SD) berries per infructescence. Based on these averages, overall fruit set was 92.49% (Table 3; Fig. 2G; Suppl. material 2).

When analysed by spathe colour, creamy melon-red spathes averaged 425.8 ± 80.99 pistillate flowers per infructescence and 331.7 ± 56.79 berries per infructescence, corresponding to a fruit set of 77.90%. Creamy green spathes averaged 348.8 ± 64.72 SD pistillate flowers and 384.7 ± 81.00 SD berries (Fig. 2G), although the unnormalized berry-to-flower ratio resulting in a calculated fruit set of 110.29%, values were normalised to a maximum of 100% (Table 3).

SEM revealed similar stamen morphology in both spathe morphs in A. mutata (Fig. 3A–E). Stamens exhibited smooth papilla surfaces with few stomata (Fig. 3C). Thecae pores were membranous and sealed during pistillate anthesis (Fig. 3B), rupturing prior to pollen release in the staminate phase (Fig. 3D). Dense spiny pollen extruded from the pore slits adhered to papillae or lodged between papilla cells (Fig. 3E). Pollen grains were spiny, globose, and approximately 15 μm in diameter (Fig. 3F).

During staminate anthesis, spiny pollens were found deposited on multiple parts of Colocasiomyia flies’ bodies, including the mouthparts, thorax, abdomen, wings, legs, and head (Fig. 3G–O).

Consistent with other unisexual-flowered genera in the tribe Schismatoglottideae (Low et al. 2014, 2016; Hoe and Wong 2016; Hoe et al. 2018; Low and Wong 2022), A. mutata exhibits a short protogynous anthesis sequence. The brief overlap of pistillate and staminate phases, approximately 30 minutes, minimizes the gap between pollen release and stigma receptivity, improving pollen transfer efficiency, a key feature in tropical Araceae (Soonthornkalump et al. 2020). Pollen longevity varies within the family. For example, in neotropical Araceae with bisexual flowers, such as Anaphyllopsis americana (Engl.) A.Hay, small pollen grains remain viable for several days due to long flowering cycles (Barabé et al. 2008). In contrast, in unisexual species with short flowering cycles, such as Montrichardia arborescens Schott, pollen viability drops by half within 24 h (Barabé et al. 2008). Apoballis mutata also possesses unisexual flowers and exhibit a comparatively short anthesis lasting approximately 26 h. The relatively large, spiny pollen grains of A. mutata and its short anthesis suggests an adaptation for rapid and effective pollen transfer within a narrow temporal window. Similar associations between short flowering duration, pollen morphology, and fly pollination have been reported in other aroids pollinated by dipterans, including Arisaema Mart. and Colocasia Schott species, where pollen is transferred during brief staminate phases by flower-breeding or visiting flies (Sultana et al. 2006; Takenaka 2006).

The sweet fruity fragrance emitted during anthesis appears to be the main attractant for Colocasiomyia flies. This differs from the benzaldehyde almond oil-like scent reported in Apoballis acuminatissima (Ulrich et al. 2010), however, whether this scent also specifically attracts Colocasiomyia flies remains unknown. Such variation in scent likely explains differences in pollinator assemblages between species. For example, in Arum maculatum L., population-level scent differences were associated with shifts in dominant pollinator guilds (Gfrerer et al. 2023). Thus, scent chemistry may play a central role in mediating pollinator specificity in Apoballis. In A. mutata, multiple Colocasiomyia species were recorded sharing inflorescences. This synhospitalic behaviour, where several fly species occupy a single inflorescence, has also been reported in other aroids such as Alocasia odora (Rob. ex Lodd., G.Lodd. & W.Lodd.) Spach (Toda et al. 2022). However, the scent profile of A. odora differs from that of A. mutata, and synhospitality in these systems appears to be driven by shared use of floral resources rather than similarity in floral odour.

Like Phymatarum M.Hotta and Schottarum P.C.Boyce & S.Y.Wong (Low et al. 2016), the pale spathe colours of A. mutata are unlikely to be primary visual attractants. Instead, the strong floral scent likely the key attractant (Díaz Jiménez et al. 2019; Gfrerer et al. 2023). Nonetheless, intraspecific variation in spathe colour may reflect genetic or ecological diversity, potentially influencing subtle aspects of pollinator interaction.

The interaction between A. mutata and Colocasiomyia flies is best described as mutualistic, aligning with findings in related taxa (Sultana et al. 2006; Takenaka 2006; Takenaka Takano et al. 2012). In these plants, the flies obtain resources while ensuring pollen transfer. Although experimental test is lacking, repeated visits to pistillate flowers during anthesis, combined with revisits to staminate flowers, strongly suggest cross-pollination and gene flow (Hu and He 2006). It is possible that Colocasiomyia sp. 2 aff. bogneri is a pollinivorous species consuming pollen while facilitating dispersal (Fig. 3J).

The adaptations observed in A. mutata, such as pollen morphology, and the timing of pollen release relative to pollinator activity, are consistent with strategies reported in other Araceae to maximise pollen attachment and transfer. In Anthurium acutifolium Engl., bees (Paratetrapedia chocoensis Anguiar & Melo, 2011) carry pollen on their abdomens and legs (Etl et al. 2017), whereas in Anthurium acutangulum Engl., gall midge (Cecidomyiidae Newman) transfer pollen via head contact (Etl et al. 2022b). In Arisaema, fungus gnats (Mycetophilidae Newman and Sciaridae Billberg) are dusted with pollen as they escape the spathe tube (Vogel and Martens 2000; Suetsugu et al. 2021, 2022; Zeng et al. 2023). Heavy pollen loads have also been documented in Parastasia beetles (Amorphophallus napalensis (Wall.) Bogner & Mayo; Chaturvedi 2017), Neella bugs (Syngonium hastiferum (Standl. & L.O.Williams) Croat; Etl et al. 2022a), and Colocasiomyia flies (Schottarum; Low et al. 2014). The pollen of these species varies in ornamentation: A. acutifolium has small, smooth pollen; A. acutangulum has slightly ornamented pollen; Arisaema species produce fine, smooth pollen; and beetle- or bug-pollinated species such as A. napalensis and S. hastiferum typically have granular to slightly sticky pollen. Likewise, Colocasiomyia visitors of A. mutata are often coated with spiny pollen during staminate anthesis, suggesting efficient attachment. In some aroids, sticky resinous exudates further aid adhesion (Hoe et al. 2016; Gibernau et al. 2021), though none were observed here.

The principal pollinators of A. mutata were three Colocasiomyia species: Colocasiomyia sp. 2 aff. bogneri (74.32% of visits; Fig. 2B), C. sp. 34 aff. bogneri (17.82%; Fig. 2D), and C. sp. 17 aff. bogneri (7.82%; Fig. 2C). Colocasiomyia sp. 1 aff. bogneri was an incidental visitor, with only two observed visits (Table 1; Fig. 2E, F). Cycreon beetles were occasional visitors and likely acted as food foragers rather than pollinators, a role seen in Phymatarum and Schottarum (Low et al. 2016). Visitation rates of Colocasiomyia flies did not vary between spathe colours morphs (Fig. 2F), indicating that spathe colour does not influence pollinator attraction or visitation intensity in A. mutata. Instead, pollinator behaviour appears to be driven by inflorescence architecture, anthesis timing, scent emission, access to reproductive or nutritive resources, factors known to be more important than visual cues in many fly-pollinated Araceae (Gibernau 2011).

Inflorescences in the Schismatoglottideae typically attract insect visitors from Hydrophilidae (Cycreon), Chrysomelidae (Coleoptera), and Drosophilidae (Diptera, genus Colocasiomyia) (Takenaka 2006; Gibernau 2016; Arriaga-Varela et al. 2018). Multiple pollinator species can co-occur within the same inflorescence but remain reproductively isolated through behaviours such as mate discrimination (Tanaka et al. 2022). During pistillate anthesis, Colocasiomyia flies were seen feeding on stigmatic secretions, moving freely between floral structures and facilitating pollen transfer. They later collected food from the inner spathe and pollen from staminate flowers (Fig. 3J). Observations were consistent across visits, although fine-scale behaviour of each species requires more targeted study.

Host specificity is well known in Colocasiomyia (Sultana et al. 2006; Xiao et al. 2022). Geographic isolation of aroid hosts may promote specialisation through host-associated ecological divergence (Matsubayashi et al. 2010), although overlap among host ranges occurs. For example, the dominant Colocasiomyia sp. 2 aff. bogneri is also recorded from Anadendrum microstachyum (de Vriese & Miq.) Backer & Alderw, Aglaonema simplex Blume, and Schismatoglottis rupestris Zoll. & Moritzi ex Zoll. Colocasiomyia sp. 17 aff. bogneri also visits S. rupestris, while C. sp. 1 aff. bogneri occurs on Aglaonema pictum Kunth and Schismatoglottis sp. (Takenaka 2006). In contrast, C. sp. 34 aff. bogneri has not yet been recorded as a pollinator of other aroids.

The high fruit set observed (> 90%) indicates efficient pollination in A. mutata. Creamy melon-red spathes averaged 77.90% fruit set, while creamy green spathes exceeded 100% (Table 3). Values exceeding 100% likely reflect sampling differences, as pistillate flowers and infructescences were counted on separate individuals.

Given that Colocasiomyia flies represented 99% of all recorded insect visitors (Table 1) and were present exclusively in pollination chamber, their role as primary pollinators is strongly supported.

The pollen of A. mutata is echinate (spiny) and ~15 μm in diameter, slightly larger than that of A. acuminatissima (~10–12.5 μm; Ulrich et al. 2012). This echinate structure contrasts with the smooth (psilate), inaperturate pollen found in other Schismatoglottideae (Thanikaimoni 1969; Grayum 1992; Hesse 2005; Ulrich et al. 2012; Low et al. 2016; Low and Wong 2022).

Pollen traits are often linked with pollination systems and pollinator types (Sannier et al. 2009). Spinose pollen in Apoballis is frequently linked to fly pollination (Grayum 1992). SEM confirmed that Colocasiomyia flies carried pollen during staminate anthesis, with pollen grains found on multiple body parts and no other pollen types detected. Distribution appeared incidental rather than targeted, with spines aiding attachment, supporting the interpretation of Colocasiomyia as specialised but opportunistic pollen carriers.