Stochastic organelle genome segregation through Arabidopsis development and reproduction

Abstract

Organelle DNA (oDNA) in mitochondria and plastids is vital for plant (and eukaryotic) life. Selection against damaged oDNA is mediated in part by segregation - sorting different oDNA types into different cells in the germline. Plants segregate oDNA very rapidly, with oDNA recombination protein MSH1 a key driver of this segregation, but we have limited knowledge of the dynamics of this segregation within plants and between generations. Here, we reveal how oDNA evolves through Arabidopsis thaliana development and reproduction. We combine stochastic modelling, Bayesian inference, and model selection with new and existing tissue-specific oDNA measurements from heteroplasmic Arabidopsis plant lines through development and between generations. Segregation proceeds gradually but continually during plant development, with a more rapid increase between inflorescence formation and the next generation. When MSH1 is compromised, the majority of observed segregation can be achieved through partitioning at cell divisions. When MSH1 is functional, mtDNA segregation is far more rapid; we show that increased oDNA gene conversion is a plausible mechanism quantitatively explaining this acceleration. These findings reveal the quantitative, time-dependent details of oDNA segregation in Arabidopsis. We also discuss the support for different models of the plant germline provided by these observations.

Keywords: Arabidopsis; bottleneck; chloroplasts; development; inheritance; mitochondria; organelle DNA; segregation.

Figure 1.. Models and data for heteroplasmy segregation in plant development.

(A) Developmental models for observations of heteroplasmy (h, the proportion of mutant organelle DNA type in a sample) in Arabidopsis thaliana. MSi and CSi are the unobserved (latent) ancestral cells at different developmental stages (O, original precursor state; EL, early leaf; LL, late leaf; INF, inflorescence) in Mother and Child shoot apical meristem (SAM). The blue horizontal bars denote the generation of sex cells and establishment of a new generation. Greyed-out elements are unidentifiable given our observations and play no role in our model. n_i correspond to the number of effective segregation events (model cell divisions) at each developmental stage. (B) Example of model for heteroplasmy h within the linear developmental model in (A). The SAM at the CS2 stage includes cell with a distribution of heteroplasmy levels. In this example, three cells a, b, and c from this distribution, with different heteroplasmy levels, go on to be the ancestors of two late leaves (LL₁ and LL₂) and part of the future SAM at stage CS3. Segregation increases heteroplasmy variance as the descendants of a, b, and c develop, leading to new distributions. These may be sampled (the mean of LL₁ and LL₂ are recorded) or unseen (the CS3 distribution plays a latent role in our model). (C-F) Observed heteroplasmy data through development in different heteroplasmic plant families from Broz et al. [2022]: (C) mtDNA in mutant msh1 background; (D) mtDNA in wildtype background (no within-plant data was taken here in Broz et al. [2022]); (E) ptDNA in mutant msh1 background; (F) no heteroplasmic samples for wildtype ptDNA available in Broz et al. [2022]. Samples are taken between generations (lower stages; comparing mother EL to a range of offspring EL) and within offspring (upper stages; measuring EL-LL-INF within offspring); within-offspring measurements were not taken for wildtype mtDNA in Broz et al. [2022]. Different colours correspond to different families (with different founder mothers). The “fanning out” of individual sample heteroplasmies over time corresponds to increasing sample-to-sample variance.

Figure 2.. Posteriors from inference process.

Posterior distributions, inferred across models, for the effective segregation events from a precursor state (O for child, OM for mother) to different tissue precursors (EL, early leaf; LL, late leaf; INF, inflorescence), and between generations (OM → O) in Arabidopsis thaliana. (A) msh1 mtDNA ($N_e=50$), (B) msh1 ptDNA ($N_e=7$); (C) wildtype mtDNA ($N_e=50$, different scale); no within-plant data was taken here in Broz et al. [2022], so the details of vegetative segregation cannot be inferred.

Figure 3.. New data and predicted segregation behaviour.

(A-B) Previous (“old”, left) and new (right) oDNA observations for (A) wildtype mtDNA and (B) msh1 ptDNA in Arabidopsis thaliana. Data are displayed as heteroplasmy levels h measured at different developmental stages. Different colours correspond to different families (with different founder mothers). (C) Within-plant segregation dynamics for wildtype mtDNA, plotted as the probability of a given number of effective segregation events between different developmental stages. Predictions (blue) from scaling the msh1 observations seven-fold to match between-generation observations; (red) inferred effective segregation events from new data. (D) Segregation dynamics of msh1 ptDNA; previous observations (blue); new observations (grey); and refined estimates inferred from the joint dataset (red). Developmental stages: O, original precursor state; EL, early leaf; LL, late leaf; INF, inflorescence.

Figure 4.. Patterns and models of segregation through development inferred from combined heteroplasmy profiles.

(A) Evidence for progressive vegetative segregation through development in Arabidopsis thaliana. Each plot asks whether the extent of segregation over one period is greater than that over another. LL > EL corresponds to late leaf segregation exceeding early leaf segregation; INF > LL corresponds to inflorescence segregation exceeding late leaf segregation; OM->O > O->INF corresponds to whether between-generation segregation (early mother to early offspring) exceeds vegetative segregation (early offspring to inflorescence). The probability for yes/no answers to these questions is given, with two independent computational estimates plotted to demonstrate numerical convergence. (B) Probabilities of different model structures from reversible jump MCMC. Models 0–2 are respectively the linear germline, separate soma, all separate lineage models from Fig. 1. Rows correspond to different organelle-mutation combinations: the final row is the mtDNA msh1 mutant with one potentially outlier lineage removed (see text). The two colours correspond to results from different RJMCMC simulation to demonstrate convergence (see also Supplementary Fig. S4). Developmental stages: OM, original precursor state of mother plant; O, original precursor state; EL, early leaf; LL, late leaf; INF, inflorescence.

Figure 5.. Summary of inferred segregation dynamics within plants and between generations.

Illustrative distributions of heteroplasmy in Arabidopsis thaliana, corresponding to the inferred mean segregation magnitude ($n$ segregating events, for $N_e=50$ mtDNAs or $N_e=7$ ptDNAs; and $N_b$, effective bottleneck size). Distributions at each developmental stage, and an initial heteroplasmy of 0.5, are shown for mtDNA (MT) and ptDNA (PT) in wildtype and msh1 mutants (all wildtype PT observations are homoplasmic, so no inference is possible; see Discussion for hypotheses). Continuous segregation is supported by inference in all systems except PT; model selection suggests most support for a picture where separate developmental lineages are involved for each developmental stage.