A subcellular atlas of Toxoplasma reveals the functional context of the proteome

Author's response

Konstantin Barylyuk and Ross Waller shared

Open questions 

1.Starting on a more general note, this is an exciting piece of work for the entire parasitology field. I imagine its relevance to other parasites has enormous potential. What are key limitations you envisage in trying to translate the application of the methodology you used, to other intracellular parasites such as Plasmodium, and to extracellular parasites such as Trypanosoma brucei?

Indeed, the hyperLOPIT approach has a great potential for characterising spatial proteomes of other parasitic protists. In fact, our research group is currently applying the method to a range of new organisms, including Plasmodium falciparum and several species of marine protists from lineages closely related to Apicomplexa that engage in a spectrum of symbiotic relationships with animals. The preliminary results indicate that hyperLOPIT is equally powerful in these cases. Furthermore, we are aware that several projects on spatial proteome mapping by hyperLOPIT are underway in other research groups, including the work of our colleagues from the Department of Biochemistry in Cambridge on African trypanosomes. Again, the coverage of the proteome, the resolution, and the level of detail achieved by hyperLOPIT in these cases is truly spectacular. We are convinced that the method has broad applicability in protists.

The main limitation of hyperLOPIT is the requirement of a large amount of starting material. This restricts the method’s applicability to organisms that are available in reasonably scalable cell culture or can be obtained in large quantities from natural sources. Another limitation is the steady-state nature of the spatial proteome map generated by hyperLOPIT. When the input population of cells is highly heterogeneous, which is frequently the case as it is often difficult to tightly synchronise a culture, the resolution of the spatial proteome map may be significantly reduced due to averaging over a large number of distinct cells captured in different stages of the cell cycle or even organism’s life cycle. Of course, learning the biology of the target organism and developing methods for maintaining and controlling it in the culture allows one to circumvent the limitations of hyperLOPIT mentioned above. However, it is not always possible. Hence, further development of the hyperLOPIT technique to improve its sensitivity, both the spatial and temporal resolution, and throughput will enable spatial proteome mapping in less accessible systems. And, indeed, such efforts are constantly underway in the laboratory of Prof Kathryn Lilley here in Cambridge (e.g. 3).

2.Also on a more general note, how easily can you explore the proteome of the parasite throughout each and every stage of its life cycle, particularly given that many of them show sometimes asynchronous development – for instance the jump between one Plasmodium stage and the next (eg. ring and trophozoite) is not a discrete step, but a continuum, assumingly with important changes occurring in the transition?

As mentioned above, hyperLOPIT requires a significant amount of starting material and produces a steady-state, population-averaged map of protein subcellular locations. The protocol is also relatively complex, time-consuming, and labour-intense. All these factors limit hyperLOPIT’s suitability to fast dynamic processes in asynchronous populations of cells. Typically, only a small fraction of cells undergo a rapid and dynamic developmental transition at any given moment, whereas the majority of the population persists in one or several differentiated states. A snapshot of such a mixed population taken by hyperLOPIT will be an average of all the cells, dominated by the signal from the most abundant cell type. The Bayesian machine-learning method used in our study to assign the unknown proteins to subcellular niches can handle this problem by quantifying the uncertainty of classification – but only to a certain extent. So, capturing these dynamic processes such as cell differentiation will be largely limited to the level of synchronicity of the cell population that can be achieved. Having said that, optimising hyperLOPIT to apply it to dynamic processes is currently an area of intense research. Stay tuned!

3.Do you think your method will allow enough resolution to study and capture differences during phenomena such as commitment to latent stages for example, in Toxoplasma and Plasmodium?

We are optimistic that applying hyperLOPIT to study the differentiation of apicomplexan parasites into dormant, persistent stages, such as the Toxoplasma bradyzoite, will provide extremely useful new data and insight into this process. In Toxoplasma, an in vitro model for bradyzoite differentiation is accessible, and the timeframe of events potentially allows taking a few snapshots of the spatial proteome during this process, even though the limitations of hyperLOPIT mentioned above will have to be carefully considered for interpreting the observations. Olly Crook, whose expertise in biological statistics and machine-learning methods have been key to successful analysis of the data reported in our paper, has been developing new powerful and flexible machine-learning algorithms for spatial proteomics (5). These new algorithms will enable the discovery of new, previously uncharacterized subcellular niches and detection of proteins that are differentially localized between the spatial proteomes of two developmental states. Both advances are extremely important in the context of investigating dynamic processes. We will be very excited to analyse the dynamic changes of the spatial proteome of Toxoplasma upon differentiation into bradyzoites – it is definitely on our to-do list.

4.Having focused in an intracellular parasite, and having explored in higher detail the cluster related to plasma membrane, in the future, can you simultaneously study the host and the parasite proteome to see the changes occurring during parasitism on both? More work across various cell biology fields has focused on understanding the relevance of membrane contact sites, and the parasitophorous vacuole membrane seems an exciting prospect.

Toxoplasma is propagated in culture in the so-called lytic cycle that replicates the acute infection in human and animal tissues. During this cycle, the parasite alternates between the two major states: the extracellular tachyzoite that is committed to invasion, and the intracellular tachyzoite that grows and replicates inside the host cell. For the first application of hyperLOPIT, the fully differentiated extracellular tachyzoite that is arrested in the cell cycle was a more practical choice. Naturally, the intracellular stage is the next logical target. The hyperLOPIT approach was originally developed using a mammalian cell model rendering the analysis of Toxoplasma-infected human cells fairly straightforward. In our ongoing work, we are aiming to address questions on both the parasite and the host cell sides. What are the differences between the spatial proteomes of extra- and intracellular tachyzoites? What are the responses of the host cell’s spatial proteome to the infection? What are the repertoire and the subcellular destinations of the proteins secreted by Toxoplasma in the host cell? Since both the parasite and the host cells are present in the sample, in this case, the main challenge is how to differentiate the two. Luckily, the protein sequences of the two species are different enough to distinguish them in the mixture, and the conditions of the mammalian cell lysis are a lot milder than what is required for breaking Toxoplasma cells open. The parasitophorous vacuole membrane, as a major interface between host and parasite, and already the known destination of many Toxoplasma secreted proteins, is indeed an exciting compartment that we hope to capture. As mentioned above, we will soon have access to new machine-learning methods tailored specifically to address challenges emerging in the analysis of such complex data.

5.A lot of the work in parasitology has assumed good translation between parasites- namely in the form of orthologues/homologues, to guide exploration of proteins and genes. Throughout your work, did you find any surprises on this respect, of proteins previously assumed to be elsewhere or previously misclassified?

Parasites evolve under the constant pressure from the host’s defence mechanisms. Contrary to the common point of view that the reduction of complexity is the major theme in their evolution, parasites gain a lot of novelty in the process of adaptation to their ever-changing and unwelcoming habitat. This rapid divergence limits the possibility to infer functions of unknown proteins by orthology to comparisons of closely related lineages. For example, according to our in-house orthology analysis, T. gondii and Plasmodium falciparum share only 2,712 orthologous groups between 8,322 and 5,303 protein-coding genes, respectively. Even in these cases, we are looking at the more conservative parts of the cell, such as the cytosol, the mitochondrion, the ER and Golgi, the IMC. In our paper, we discuss several examples of Toxoplasma proteins found in subcellular locations other than what would be expected based on the knowledge available from other organisms or inferred by orthology. One such example is the localisation of an apicomplexan-specific RNA-binding protein containing the so-called RAP domain in the apicoplast, whereas all the other members of this protein family are targeted to the mitochondrion. The functions of RAP-domain proteins remain elusive, and it is even more intriguing to find one of them in a different organelle. Other examples are dense-granule proteins originally annotated as belonging to rhoptries and differentiation of a subset of ER proteins including the canonical ER marker protein BiP from the rest of the ER. So, it is clear that while homology can be a useful guide to function, there are many processes driving proteins to new functions and behaviours and caution should be applied when making assumptions based on conservation.

6.Going to some specifics, I was wondering if you discuss in more detail the relevance of two of your findings: the first – that dense granules seem to have the highest novelty in proteins. The second: the unexpected mutation behaviour in mitochondrial soluble proteins and in the apicoplast, which you suggest could be a contributing factor for metabolic control, and even guide tropism, taxon preference, and even virulence across parasites and/or strains. For instance, for the latter, do you expect big changes in the proteome atlas based on the parasite’s tissue tropism?

As mentioned above, apicomplexan parasites are constantly challenged by the host’s defence mechanisms and must rapidly change to adapt to this evolutionary pressure. The secretory organelles and the parasite’s surface are most exposed to the host. Furthermore, Toxoplasma gondii exploits a remarkably wide range of intermediate hosts: virtually any warm-blooded animal can be infected. Toxoplasma must, therefore, arm itself with a broad arsenal of secretory effectors to enable successful invasion and survival within such a diversity of hosts. Hence is the accumulation of novel proteins in the secretory compartments of the parasite including the dense granules.

The mitochondrion and the apicoplast are different in this regard since they are never exposed to the host. These are endosymbiotic organelles that used to be free-living organisms in the distant past. They contain their own genomes, albeit heavily reduced, and maintain dedicated machinery for gene expression. Their main functions are associated with essential metabolic pathways, which can often be reduced in parasitic organisms to a bare minimum thanks to the availability of metabolites and nutrients from the host. Thus, we expected to find the proteomes of these organelles to be largely conserved and stable. Much to our surprise, we found a significantly higher than the average density of polymorphisms in the genes encoding the soluble mitochondrial and apicoplast proteins but otherwise neutral or only slightly elevated selection for changes. These observations suggest a particular relevance of codon usage to these compartments and their metabolism. The ability to modulate physiological states has been seen as relevant to other parasites also, such as Plasmodium which switches from anaerobic to aerobic metabolism (for review see 6). These switches have been suggested to enable the parasite to enter or exit from periods of dormancy or altered physiological states of their hosts. The population-level genetic data for Toxoplasma, mapped onto our hyperLOPIT data, suggest the hypothesis that tuning of metabolism might equally be relevant to Toxoplasma parasitism. Indeed, the Toxoplasma population data is drawn from isolates with differences in the host taxa and tissues that they favour, as well as virulence properties. However, rather than ‘big changes’ in the protein atlases, we’d anticipate these to represent subtle fine tuning and, if so, this points to further evidence of the exquisite adaptation of these successful parasites. This is an example of the exciting and intriguing hypotheses generated by the types of data that hyperLOPIT can achieve, and that will now need testing.