Spicoid morphology of Mapanioideae (Cyperaceae): an evolutionary perspective

Ancestral character
reconstruction was performed using the
Bayesian Inference molecular phylogenetic hypothesis of Cyperaceae from Semmouri et al. (2019) restricted to Mapanioideae
species. Cyperaceae inflorescences were previously classified based on the typology
of the inflorescence architecture found in Cyperaceae. This resulted in the “scirpoid”,
“cyperoid”, “rhynchosporoid” and “caricoid” type of spikelets proposed by Dahlgren
et al. (1985). However, we believe that considering a typological classification
for the spicoid morphology may hamper elucidating the evolution of its morphological
variation, since typologies represent a combination of several characters that
can be shared by different types. Therefore, instead of evaluating it as a
single unit, the spicoid morphology was tackled in a reductive coding through
an iterative process of decomposition. A similar approach was successfully used
by Semmouri et al. (2019) to study the evolution of embryo morphology in
Cyperaceae.

Spicoid morphology was initially partitioned into eight characters: 1. presence/absence
of subtending bract; 2. number of lateral scales; 3. presence/absence of keel in the lateral scales; 4. number of lateral stamens, 5. number of inner scales,
6. number of inner stamens; 7. position of the pistil; 8. Pistil merosity (related to the number of stigmas branches). From these
characters, five were shown to be variable, potentially phylogenetically
informative, and therefore selected for downstream comparative analyses. The
variable characters are: 1. The presence of keels on the lateral scales (a: present, b: absent); 2. Number of lateral stamens
(a: 0, b: 2); 3. Number of inner scales (a: 0, b: 1-6, c: 7-12, d: 13-19, e: >19);
4. Number of inner stamens (a: 0, b: 1-6, c: 7-14, d: >14); 5. Pistil merosity
(a: dimerous, b: trimerous). The coding arrangements were performed to optimize the
character reconstruction analyses. The parameter “type=discrete” was used since the morphological features
were coded as discrete. Bayesian stochastic character (BSC) mapping (Huelsenbeck
et al., 2018) was performed using the ‘make.simmap’ function available in
phytools v0.5–64 (Revell, 2012). The analyses were implemented in R (R Core Team, 2018) using the model
that best fits to the data (“equal rates” or “all rates different”) to recover the evolutionary history of the morphological characters
under investigation and recover ancestral character state probabilities at the
nodes. Outgroups, terminals with more than one accession per taxon and
polytomies were removed from the analysis using the function ‘drop.tip’ in ape (Paradis
et al., 2004) to prevent bias in character state reconstruction. Stochastic character mapping
was performed using 1000 simulations. BSC provided transition rates which were
used to build a transition matrix among features as input to the heatmaps
depicted using pheatmap v1.0.12.