Chloroplasts alter their morphology and accumulate at the pathogen interface during infection by Phytophthora infestans

Upon immune activation, chloroplasts switch off photosynthesis, produce anti-microbial compounds, and associate with the nucleus through tubular extensions called stromules. Although it is well-established that chloroplasts alter their position in response to light, little is known about the dynamics of chloroplasts movement in response to pathogen attack. Here, we report that chloroplasts accumulate at the pathogen interface during infection by the Irish potato famine pathogen Phytophthora infestans, associating with the specialized membrane that engulfs the pathogen haustorium. Chemical inhibition of actin polymerization reduces the accumulation of chloroplasts at the pathogen haustoria, suggesting this process is partially dependent on the actin cytoskeleton. However, chloroplast accumulation at haustoria does not necessarily rely on movement of the nucleus to this interface and is not affected by light conditions. Stromules are typically induced during infection, embracing haustoria and interconnecting chloroplasts, to form dynamic organelle clusters. We found that infection-triggered stromule formation relies on BRASSINOSTEROID INSENSITIVE 1-ASSOCIATED KINASE 1 (BAK1) mediated surface immune signaling, whereas chloroplast repositioning towards haustoria does not. Consistent with the defense-related induction of stromules, effector mediated suppression of BAK1 mediated immune signaling reduced stromule formation during infection. On the other hand, immune recognition of the same effector stimulated stromules, presumably via a different pathway. These findings implicate chloroplasts in a polarized response upon pathogen attack and point to more complex functions of these organelles in plant-pathogen interactions.


Introduction
Phytophthora infestans is an oomycete pathogen that causes potato late blight, one of the most historically important and economically devastating crop diseases. The pathogen penetrates host cells via haustoria, infection structures that extend from its intercellular invasive hyphae. Haustoria are surrounded by the plant-derived extrahaustorial membrane (EHM), across which effectors secreted by the pathogen translocate inside the host cell (Wang et al., 2017;Whisson et al., 2007Whisson et al., , 2016. This interface is key to the success or failure of infection and is therefore targeted by focal immune responses of the plant (Bozkurt et al., 2011;Dagdas et al., 2018;Kwon et al., 2008). This includes the deposition of callose, redirection of autophagy, and movement of the nucleus towards the site of penetration (Dagdas et al., 2018;Griffis et al., 2014;Jones & Dangl, 2006). While continuous with the plasma membrane, there is a stark difference in the biochemical composition of the EHM and the plasma membrane (Bozkurt et al., 2014. The EHM typically lacks the surface localized pattern recognition receptors (PRRs), which activate downstream immune responses through recognition of pathogen associated molecular patterns (PAMPs) (Bozkurt et al., 2014. Once PPRs detect PAMPs, downstream signaling is triggered, often in co-ordination with a co-receptor such as BRASSINOSTEROID INSENSITIVE 1-ASSOCIATED RECEPTOR KINASE1 (BAK1), to induce an immune response (Chaparro-Garcia et al., 2011;Heese et al., 2007). To counteract this, pathogens typically deploy host-translocated effectors to subvert surface mediated immunity. For example, P. infestans host-translocated RXLR effector AVR3a suppresses BAK1 mediated surface immune responses . However, presumably, plant basal responses still contribute to immunity against adapted pathogens, as immune suppression by effectors is often partial, as in the case of AVR3a . Moreover, downregulation of BAK1 gene expression in solanaceous model plant Nicotiana benthamiana leads to significantly enhanced pathogen growth, further highlighting the importance of surface mediated immunity against P. infestans . This makes N. benthamiana an excellent model for studies to dissect the functional principles of basal plant immunity against P. infestans. Furthermore, N. benthamiana lacks the specialized nucleotide-binding leucinerich repeat (NLR) type of immune receptors that can sense P. infestans effectors intracellularly. This is also advantageous because it allows for live cell imaging of P. infestans infection, as NLR mediated immunity often triggers a form of programmed cell death at the site of infection known as the hypersensitive response (Wu et al., 2017).
Activation of immunity at the cell surface stimulates chloroplasts to shut down photosynthesis, synthesize defense hormone precursors, and generate reactive oxygen species (ROS) (Padmanabhan & Dinesh-Kumar, 2010;Su et al., 2018), indicating that chloroplasts are major components of the plant defense system. Pathogens are known to target chloroplasts with effector proteins, further highlighting their importance in immunity (Jelenska et al., 2007;Pecrix et al., 2019;Petre et al., 2016;Zabala et al., 2015). Interestingly, several genes associated with resistance to oomycete pathogens were found to encode chloroplast-localized proteins (Belhaj et al., 2009;Van Damme et al., 2009). Chloroplasts also produce stroma filled tubules (stromules) in response to a range of elicitors, including phytohormones, ROS and the bacterial PAMP, flg22 (Brunkard et al., 2015;Caplan et al., 2015;Gray et al., 2012). While the exact function(s) of stromules is still unclear, they have been implicated in immunity, chloroplast movement, and connection to the plant cell nucleus (Caplan et al., 2015;Kumar et al., 2018). Immune stimulation by PAMPs and ROS also induce the association between chloroplasts and the nucleus, hinting at potential defense-related roles of chloroplast re-distribution during infection (Ding et al., 2019). However, the molecular and physiological mechanism of how chloroplast immunity is launched against invading pathogens is unclear.
Here, we used quantitative confocal microscopy to investigate the spatial dynamics of chloroplasts in living plant cells infected by P. infestans. We show that chloroplasts accumulate around haustoria in a dynamic fashion, but this process does not necessarily rely on movement of the nucleus towards the haustorium. We found that the actincytoskeleton, but not light conditions, are critical for chloroplast positioning around the haustorium. Our microscopy analyses using optical tweezers suggest association between chloroplasts and the EHM. Finally, we demonstrate that chloroplasts also alter their morphology by induction of stromules as a defense response, whereas effectors can counteract this process.

Chloroplasts accumulate at the host-pathogen interface in an actin dependent manner but irrespective of light conditions
While the immune-related roles of chloroplasts in producing anti-microbial compounds and defense signaling molecules are well established, little is known on the subcellular dynamics of chloroplast movement and stromule induction during infection. To gain insights into this, we monitored the live infection of N. benthamiana by P. infestans, specifically targeting cells containing a pathogen haustorium (haustoriated cells). In this infection model, haustoria are generally visible in the host epidermal cells, where the chloroplasts are smaller and less abundant than in the mesophyll. We reasoned that the more sparsely distributed chloroplasts of these epidermal cells may redistribute towards the site of infection to partake in a localized intracellular immune response.
During infection of N. benthamiana with the fluorescently tagged strain of P. infestans, 88069td, haustoria are easily visible in host cells. Confocal microscopy of infected leaf epidermal cells stably expressing GFP in chloroplast stroma (CpGFP herein) revealed that chloroplasts associate with 40% of haustoria (N = 280 haustoria) (Fig. 1A, Video S1). Haustoria were often associated with multiple chloroplasts, and these chloroplasts were often mobile around the site of infection (Fig. 1A, Video S2).
We investigated whether positioning of chloroplasts around haustoria was a response to infection or due to the result of chance encounter, i.e., because of haustoria coincidentally penetrating the cell at the position where there is a chloroplast. To check whether this association occurred at a greater frequency than would be expected by random chance encounter, we developed an unbiased method to position mock haustoria throughout micrographs of infected tissue (Fig. S1A). For each real haustorium in an image, a straight line was drawn from the point at which the haustorium entered the cell until the opposite edge of the same cell was reached; this end point was imagined as the position of a mock haustorial penetration. We then categorized whether each mock haustoria was immediately adjacent to, or in contact with, a chloroplast in that position (Fig. S1A). If chloroplast accumulation at haustoria was random, we would expect rates of actual and mock haustoria to be the same on opposite edges of the cell. Compared to the actual haustoria, there were significantly fewer instances contact between chloroplasts and mock haustoria (Fig. S1B), suggesting the association between haustoria and chloroplasts was not due to random encounter during cell penetration.
To gain further insights into chloroplast positioning around the haustorium, we monitored chloroplast dynamics in haustoriated cells with time lapse-microscopy. Here, we observed several instances of chloroplast movement towards the haustoria, suggesting an active relocation of chloroplasts towards the site of intracellular infection (Video S3, S4). Similar observations were made during wild-type P. infestans infection, visualizing the EHM with RFP:REM1.3 (Video S5). Additionally, we observed some instances where haustoria that were associated with chloroplasts collapse during the acquisition of the time-lapse (Video S6, S7). This indicates that some of the haustoria-chloroplast associations could be missed out during image quantification as it is not always possible to identify collapsed haustoria.
The cytoskeleton, and particularly actin, has known roles in chloroplast movement (Wada & Kong, 2018), nucleus movement (Higa et al., 2014), stromule interactions (Erickson et al., 2018;Kumar et al., 2018), and cell polarization towards the sites of pathogen penetration (Kobayashi & Hakuno, 2003;Opalski et al., 2005). Therefore, we next tested whether chemical inhibition of actin polymerization by Latrunculin-A (LatA) would influence the accumulation of chloroplasts at haustoria. To monitor successful disruption of actin, a GFP tagged actin chromobody was transiently expressed during 1 μM LatA treatment of infected wild-type plants by P. infestans 88069td (Fig. 1B). We also tested a range of LatA concentrations for ability to visually disrupt actin filaments before completing the infection microscopy, in order to use a minimal concentration of LatA (Fig. S2). Treatment with LatA significantly reduced the frequency of chloroplast-haustoria associations compared to the control condition (34% of N = 148 haustoria, and 47% of N = 93 haustoria respectively) (Fig. 1C). This points to a role for actin in the accumulation of chloroplasts at haustoria. However, we are cautious of overinterpreting this data due to the non-specific nature of the drug and its plausible ability to influence the growth and virulence of the pathogen (Ketelaar et al., 2012). By using the GFP marker of actin to ensure inhibition in the cells imaged, we sacrificed the ability to monitor stromule induction during this experiment.
Finally, we also tested whether light exposure has any impact on accumulation of chloroplasts at haustoria. For the dark condition, we kept P. infestans 88069td infected CpGFP leaves in the dark for 2 days prior to imaging (dark-dark), while we kept the control group in regular day-night light cycling conditions (light-dark). Live cell infection microscopy revealed no difference in chloroplast accumulation between the two lighting conditions (Fig. S3). Collectively, these findings indicate that chloroplasts actively position around haustoria during P. infestans infection through a process that requires actin polymerization, whereas this process is not affected by changes in light conditions.

Optical tweezers reveal the association between chloroplasts and the haustoria
Accumulating evidence points to the importance of organelle membrane contacts in response to various physiological or stress conditions (Helle et al., 2013;Liu & Li, 2019;Silva et al., 2020). To investigate the nature of chloroplasts association with haustoria/EHM, we used optical tweezers in combination with Total Internal Fluorescence Microscopy (TIRF) in CpGFP plants, infected with the wild-type P. infestans 88069, and transiently expressing the extrahaustorial membrane (EHM) marker, RFP:REM1.3 (Bozkurt et al., 2014). Relocation of chloroplasts further than a 10 μm threshold was considered a successful movement by optical tweezers. This threshold was set to ensure consistency between experiments; 10 μm was chosen as this distance was small enough to avoid side effects from moving the chloroplast towards the vacuole, such as pushing the chloroplast into the tonoplast.
Using optical tweezers, we successfully trapped and moved 17% of chloroplasts (N = 29) in non-haustoriated cells beyond the threshold using the automated trapping routine. In comparison, we were unable to trap and move any chloroplasts (0%, N = 18) neighboring haustoria past 10 μm, suggesting an association or connection between the chloroplasts and the EHM. Consistent with this, we recorded instances where these chloroplasts were initially pulled away from the EHM, but before they passed 10 μm, they escaped the trap and sprang back towards their former position (22%, N = 18) (Fig. 1D, Video S8). These results suggest that chloroplasts may establish secure contacts with the EHM, however, further genetic and biochemical evidence supporting these findings are required to reach definitive conclusions.

Chloroplasts can associate with haustoria independently of the host nucleus
Relocation of the plant nucleus towards pathogen penetration sites was reported as a hallmark of plant focal immune responses (Griffis et al., 2014). Given recent reports showing the association of chloroplasts with the nucleus during plant stress responses (Caplan et al., 2015;Ding et al., 2019;Erickson et al., 2018), it is possible that chloroplasts are dragged to haustorium along with the nucleus. However, this is unlikely because there is only a single nucleus per plant cell, and P. infestans can form multiple haustoria that are in contact with chloroplasts within a single cell (Fig. 1A).
Nevertheless, we quantified the numbers of haustoria associated with chloroplasts alone, compared to chloroplasts and nuclei together, to determine the extent to which chloroplast positioning around the haustoria correlates with the presence of nucleus. To track the nucleus, we used leaf epidermal cells stably expressing GFP in the endoplasmic reticulum (ER-GFP herein), to visualize the silhouette of the nucleus (Fig. 2). Confocal microscopy followed by manual quantification of 463 haustoria revealed that 10% were associated with both a chloroplast and the plant nucleus ( Fig. 2A), while 26% were associated with just a chloroplast but not a nucleus (Fig. 2B). Very rarely (2%), a haustorium was in contact with a nucleus without also being in contact with a chloroplast (Fig. 2D).
Using the same methodology in Figure S1, we compared the frequency of actual chloroplast/nuclear accumulation at haustoria with hypothetical rates expected by chance encounter. We found significantly fewer cases of chloroplast association alone with haustoria than in the mock haustoria data, as well as fewer cases of chloroplast-nucleus co-association with haustoria, but not with nucleus alone association (Fig. S4). Taken together, these results suggest the chloroplast positioning towards haustoria can occur independently of nuclear migration, and that both chloroplast and nuclear accumulation at haustoria occurs at a greater frequency than expected by chance.

Chloroplasts alter their morphology and contact each other via induction of stromules in response to infection
During live cell imaging of CpGFP plants infected by P. infestans 88069td, we also noted an increase in the frequency of stromulesfrom 4% in the mock infected tissue, to 19% in the infected tissue ( Fig. 3A-B). As previously reported, stromules varied in shape and size (M. Schattat et al., 2011); some of these stromules extended towards and wrapped around haustoria (Fig. 3C, Video S9). Further, stromules often extended between different chloroplasts, occasionally even bridging multiple haustoria (Fig. S5, Video S9 & S10). We also monitored stromule-haustorium associations by transiently expressing the EHM marker protein RFP REM1.3 in leaves and infecting with the wild-type P. infestans 88069. We noted close association of stromules with the EHM under these conditions (Fig. 3C, Video S11). These results are in agreement with the earlier reports that stromules can be induced by PAMPs (Caplan et al., 2015), and hint at defense-related roles of stromules possibly through mediating chloroplast-chloroplast and chloroplast-EHM associations, as well as the previously reported chloroplast-nucleus associations (Caplan et al., 2015).
During our attempts to move chloroplasts in infected cells, we once attempted to move a chloroplast that was seemingly connected to another by a stromule like extension. Here, we observed co-migration of the chloroplasts by moving only one of the pair (Fig. 3D, Video S12), indicating that chloroplasts may be linked by their stromules, or that chloroplasts can follow one another when bridged by a stromule. Co-migration between non-stromule connected chloroplasts has also been previously observed (Caplan 2018). However whether plastids can fuse to form a continuous stromal compartment that enables macromolecule exchange thorough stromules is still under debate (Hanson & Hines, 2018;M. H. Schattat et al., 2015).

Stromules are induced upon surface immune activation
PAMPs trigger a range of immune responses when recognized by surface localized immune receptors (Jones & Dangl, 2006). Flg22, a peptide PAMP from bacterial flagellin, was previously shown to induce stromules (Caplan et al., 2015). We replicated this result with flg22, while also testing other elicitors: chitin, a polysaccharide PAMP of fungal microbes and arthropod pests; and INF1, an extracellular P. infestans protein that, unlike most PAMPs, elicits host cell death upon perception by the plant (Kamoun et al., 1998). Flg22, chitin, and INF1 all induced stromules after 24 hr compared to the water control (Fig. 4A), suggesting stromule production is a general response to a range of microbes. PAMP activity was confirmed by detection of phosphorylated mitogen-activated kinases (MAPKs) by Western blot (Fig. S6).
Extending this, we tested whether silencing of BAK1 through VIGS would alter chloroplast and nucleus accumulation at the haustorium. To do this, we used live cell infection microscopy of BAK1 silenced ER-GFP plants. We found no significant difference in chloroplast and/or nuclear accumulation at haustoria during infection compared to silencing controls (N = 168 and N = 142 haustoria respectively) (Fig. 4C). This indicates that BAK1 dependent signaling for stromule induction is independent from that of chloroplast accumulation at the host-pathogen interface.
To complement the BAK1 silencing phenotypes on stromule formation, we then tested whether AVR3a, a host-translocated effector of P. infestans that suppresses BAK1mediated immune signaling (Chaparro-Garcia et al., 2011, can perturb pathogen induced stromule development. Following transient expression of AVR3a, stromule formation during infection decreased significantly, from 23% to 10% (Fig. 4D). This indicates pathogen effectors can perturb stromule induction, further supporting a defenserelated role of stromules.
Functionality of AVR3a was confirmed by observing cell death when transiently coexpressed with the cognate R3a NLR immune receptor (Fig. S8). We also observed that in the absence of the pathogen, transient co-expression of AVR3a with R3a led to an induction of stromules compared to the control, where R3a was transiently co-expressed with EV instead of AVR3a (Fig. 4E). This points towards the induction of stromules by HR cell death triggered by effector recognition, despite the functionality of the effector in suppressing stromules when it goes undetected by the plant. Taken together, these results show importance of cell-surface immune perception of PAMPs in the induction of stromules and suggests this process can be suppressed by pathogen effectors but can be elicited upon the detection of those effectors (Fig. 4F).

HyPer ROS reporter displays increased reactivity in chloroplasts upon infection
Previously, Caplan et. al. (2015) proposed that close proximity of chloroplasts to the plant nucleus could allow for ROS transfer to the nucleus via stromules. We reasoned that chloroplast positioning around the haustorium may serve to increase localized reactive oxygen species (ROS) production by chloroplasts at the site of infection. To test this, we visualized ROS in live cell infection using the HyPer ROS sensor fused to the chloroplast transit peptide of A. thaliana RecA, cTP-HyPer herein (Caplan et al., 2015). In the presence of ROS, the ratio between emission intensity from 405 nm and 488 nm excitation changes, giving a detectable readout of ROS from confocal microscopy (Belousov et al., 2006).
The epidermal plastids of infected cells showed noticeably more signal than the uninfected cells (Fig. 5). However, within a single haustoriated cell, we could not see a discernable difference in ROS signal between the different plastids, including between those associated and those unassociated with haustoria (Fig. 5). We speculate that chloroplast ROS production and visualization may be highly variable due to differences in stages of infection and the changes in the microenvironment as well as potential manipulation by effector secreted by the pathogen.  . Visual representation of the cTP-HyPer signal intensity ratio between excitation by 405 and 488 nm lasers, shown with the ImageJ 'Fire' LUT. Haustoria marked by white arrow heads. All scale bars are 10 µm. All confocal settings were kept identical between infected and uninfected during acquisition and all channels were displayed, hence why some green background signal is visible in the uninfected image.

Discussion
Here, we demonstrate that during host colonization by P. infestans, chloroplasts accumulate at the pathogen interface ( Fig. 1-2) and alter their morphology through induction of stromules (Fig. 3-4). Additionally, we show that nuclei are almost exclusively localized to haustoria in the company of a chloroplast, but that chloroplasts can accumulate independently (Fig. 2). Stromules occasionally embrace the EHM and link chloroplasts to each other, forming dynamic organelle clusters around the pathogen interface (Fig. 3C, Fig. S5) that are reminiscent of mitochondrial networks with poorly understood functions (Hoitzing et al., 2015). Notably, infection-triggered stromule development relies on surface immune signaling, whereas the pathogen can subvert this process remotely interfering with these pathways (Fig. 4). Our results implicate chloroplasts in cell polarization upon pathogen attack and point to more complex functions of these organelles in plant-pathogen interactions.

How do chloroplasts position at the pathogen interface?
It is well-established that chloroplasts alter their subcellular localization in response to light in an actin dependent manner (Kadota et al., 2009;Wada & Kong, 2018). We found that infection induced chloroplast positioning around the haustorium relies to an extent on actin polymerization ( Fig. 1B-C), whereas the process is not affected by light conditions (Fig.  S3). Previous studies have shown that modulation of the actin cytoskeleton can affect cell penetration by filamentous plant pathogens (Kobayashi & Hakuno, 2003;Tang et al., 2016), and others have noted the re-organization of actin around the site of pathogen penetration (Opalski et al., 2005). Our finding, that actin polymerization inhibitor LatA reduces chloroplast positioning at haustoria (Fig. 1C) is in agreement with previous studies that implicate actin in the focal immune response, as well as the known importance of actin in both chloroplast movement (Higa et al., 2014;. While this is a promising first step in altering chloroplast positioning during infection, we also heed caution in the use of pharmacological treatment of live infection systems with potential off target effects in both host and pathogen. Ideally, work following up to this would move towards using targeted genetic approaches or identifying effectors that can target this process more specifically. Additionally, we monitored chloroplast association to haustoria in the context of nuclear association (Fig. 2). Although the movement of nuclei towards plant-pathogen interfaces appear to be under varying spatio-temporal dynamics in different patho-systems (Griffis et al., 2014;Scheler et al., 2016), evidence suggest that this process may contribute to plant immunity (Daniel & Guest, 2006). However, nuclear movement during cell penetration by filamentous microbes is not exclusively uni-directional, with the nucleus moving first towards and then away from penetration sites in many interactions (Genre et al., 2005;Schmelzer, 2002). Our results regarding the accumulation of nuclei provide a snapshot at a single time point of infection, therefore our conclusions are derived from the observed state of the nucleus at that time point, not considering whether the nucleus is in the process of moving towards or away from the haustorium. Further dissection of specific nuclear movement components over extended time courses is required to address the intricacies and impact of nuclear movement in plant focal immunity.

Why do chloroplast alter their morphology during infection?
Our findings on PAMP induction of stromules (Fig. 4A) validate and expand upon those from Caplan et. al. (2015), who previously showed that flg22 can induce stromules. The reduction of stromules during infection by systemic silencing of BAK1 (Fig. 4B) suggest that a major sub population of stromules (if not all) induced during infection rely on BAK1 mediated immune signaling initiated at the cell surface. Here, we used AVR3a, an effector protein, as a tool to cross-examine the role of BAK1 (Fig. 4B) since it is known to suppress BAK1-mediated defense-signaling (Chaparro-Garcia et al., 2011. Furthermore, we hypothesize that other effector proteins are likely able to inhibit stromules indirectly by targeting similar cell surface signaling pathways. Interestingly, BAK1 silencing did not affect positioning of chloroplasts and/or the nucleus around haustoria (Fig. 4C), hinting at differential signaling pathways between stromules and organelle positioning. As of yet, we do not know what signaling takes place to re-route plant defenses towards the host-pathogen interface. The decoupling of stromule induction and chloroplast accumulation at haustoria is further supported by our time-lapse imaging which shows accumulation of chloroplasts at haustoria can occur independently of stromules, although we observe both stromule led and stromule independent movement towards haustoria (Video S3 & S4). This is in keeping with observations by Kumar et. al. (2018), who found that chloroplast movement is mostly, but not exclusively, stromule directed.
Induction of stromules by immune signaling during pathogen attack strongly points to defense-related functions of these tubular organelle extensions. But how could stromules contribute to immunity? While our understanding of stromules is still limited by a lack of stromule specific inhibitors/inducers, a model of stromules as signaling conduits is emerging. Caplan et. al. (2015) suggest that the increased surface area provided by a stromule aids in the transfer of chloroplast synthesized pro-defense molecules to the cytosol and nucleus where they function. The increase in chloroplast-nucleus contact, facilitated by stromules, and triggered during an HR response, is thought to amplify the progression of HR in a positive feedback loop. Similarly a more recent study tracked the spatiotemporal redox state of chloroplasts, and stromule induction, in potato during potato virus Y challenge (Lukan et al., 2021). Lukan et. al. (2021) conclude that, in this pathosystem, stromules are involved in signaling on the virus multiplication front.
We observed that chloroplasts establish network-like interactions via stromules (Fig. S5), and that some chloroplasts intimately associate with the haustorium interface through stromules (Fig. 3C). These chloroplast clusters and stromule extensions around the haustorium could plausibly aid the coordination of defense-related functions of chloroplasts by, for instance, mediating deployment of pro-defense molecules at the pathogen interface. Our findings further support the notion that stromules are induced to contribute to pathogen defense (Caplan et al., 2015;Erickson et al., 2018), but the function of stromules still remains to be determined (Hanson & Hines, 2018).

Right time, right place: Chloroplast position at pathogen interface is a host defense or pathogen strategy?
It is unclear whether the association of chloroplasts with the haustoria of P. infestans is a plant defense or a pathogen virulence strategy given the arsenal of immune chemicals produced by chloroplasts, it is plausible that their presence at haustoria may enhance the effectiveness of their deployment. However, we cannot discount the possibility that chloroplasts accumulate at haustoria to the benefit of the pathogen, perhaps serving to nourish the parasite. Specific, genetic strategies that impair chloroplast positioning around the haustorium are necessary to reach definitive conclusions on this. Expanding this work to other patho-systems, particularly those distantly related or those with a different lifestyle (i.e. completely biotrophic), as well as using alternative strains of P. infestans with different effector repertoires would also help resolve the physiological and evolutionary relevance of this observed response. We believe this work will lay the foundation for future studies regarding chloroplast movement towards and association with intracellular pathogen structures. By further defining the role of surface immune signaling in stromule induction and showing how effectors can be used to manipulate this process, we believe these tools will help accelerate research into the stromule function and signaling pathways.

Biological Material
Nicotiana benthamiana plants grown in a growth chamber at 25°C under high light intensity (16-h-day/8-h-dark photoperiod). Transplastomic GFP-expressing N. benthamiana plants, accumulating GFP in the chloroplast stroma (Stegemann et al., 2012), and transgenic GFP-expressing N. benthamiana plants, accumulating GFP in the endoplasmic reticulum were maintained in the same conditions as wild-type N. benthamiana. Phytophthora infestans isolate 88069  and 88069td (Whisson et al., 2007), a transgenic strain expressing the red fluorescent marker tandem dimer RFP (tdTomato), were used. Both isolates were cultured on plates with rye sucrose agar (RSA) for 12-16 days at 18°C in the dark, as described elsewhere (Song et al., 2009) prior to use for infection of N. benthamiana.

Virus induced gene silencing (VIGS)
Agrobacterium was prepared as above carrying TRV1 and the appropriate TRV2 construct and mixed to a final OD600 of 0.4 or 0.2 respectively, in agroinfiltration buffer supplemented with 100 µM acetosyringone (Sigma) and left in the dark for 2 h prior to infiltration to stimulate virulence. 14-day old N. benthamiana seedlings were infiltrated in both cotyledons and any true leaves that had emerged. N. benthamiana plants were infiltrated with TRV1 and TRV2-BAK1 for BAK1-silencing and TRV1 and TRV2-EV for the empty vector control. TRV2 containing the N. benthamiana sulfur (Su) gene fragment (TRV2-NbSU) was used as a positive control to indicate viral spread. Plants were left to grow under standard conditions until experiments could be carried out four weeks later.

Phytophthora infestans infection
Zoospores were harvested from sporangia by addition of cold distilled water and collected after 2h of incubation at 4°C, adjusting dilution to 50,000 spores/ml. Infections were performed by the addition of 10 μL of zoospore droplets to the abaxial side of the leaf. The infected leaves were maintained in plastic boxes on damp paper towels at 18°C under 16h-day/8-h-night conditions (except in the dark/dark experimental condition where there was no day period, Fig. S3).

PAMP treatments
Flg22 and INF1 purified peptides were provided by The Sainsbury Laboratory (Norwich). Chitin was prepared from powdered shrimp shell (Sigma-Aldrich) as per manufacturers protocol. Working concentrations of 1 μM (flg22 and INF1) and 100 μg mL-1 (chitin) were used unless otherwise stated. Dilutions were made in water and infiltrated into the underside of leaves using a needless syringe.

LatA treatment
Latrunculin A (abcam) was diluted to a stock concentration of 100 μM in 100 % DMSO. Water control was prepared with matching final concentration of DMSO (v/v). LatA was infiltrated into leaf tissue by needless syringe, 24 hours before microscopy.

Visualization of chloroplast ROS
Live cell imaging of chloroplast ROS was imaged using the HyPer ROS sensor (Belousov et al., 2006) fused to the chloroplast transit peptide of Arabidopsis thaliana RecA. This ROS sensitive fluorescence-based marker is imaged by fast line switching between 405 nm (channel 1) and 488 nm excitation (channel 2), detecting emission in the range of 491-543 nm. The ratio between emission from channel 1 and channel 2 gives the final signal by dividing signa from channel 2 by channel 1 in ImageJ using the inbuilt 'Math' functions (Schneider et al., 2012). Lookup table was set to Fire in ImageJ for better visualization of intensity.

RT-PCR assay
60 mg of leaf tissue was excised from 5-week old leaves (VIGS experiments) and frozen in liquid N2. RNA was extracted from the leaf tissue using the Plant RNA Isolation Mini Kit Protocol (Agilent Technologies). RNA quality and concentration was measured using a NanoDropTM Lite Spectrophotometer (Thermo Scientific). cDNA was synthesized using as a template 2 μg of RNA following the SuperScript II RT protocol (Invitrogen). To amplify the cDNA, a standard PCR (RT-PCR) was then performed using DreamTaq DNA polymerase (5 U/μL) (Thermo Scientific). VIGS BAK1 silencing was confirmed as previously described (Chaparro-Garcia et al., 2011).

Confocal microscopy
All microscopy analyses were performed on live N. benthamiana epidermal cells 2-6 days post agroinfiltrations and infections. Leaf discs were excised and imaged on either a Leica SP5 or SP8 resonant inverted confocal microscope (Leica Microsystems) using 63X, or 40X respectively, 1.2NA Plan-Apochromat water immersion objective. Specific excitation wavelengths and filters for emission spectra were set as described previously (Koh et al., 2005). The Argon laser excitation was set to 488 nm and the Helium-Neon laser to 543 nm and their fluorescent emissions detected at 495-550 and 570-620 nm to visualize GFP and RFP fluorescence, respectively. To avoid bleed-through from different fluorophores, images were acquired using sequential scanning and Maximum Intensity Projections were created from the Z-stacks. 3D images and videos were generated with confocal files in 12-bit TIFF format imported into NIS-Elements (Version 4.50, Nikon Instruments, UK) and processed with Advanced Denoising. Videos were made using the Volume View and Video Maker modules.

Optical trapping setup
Optical trap for chloroplast/stromule capture was setup as described by Sparkes et al 2017, Chapter 13 (Sparkes et al., 2018. An optical trap with a two-channel TIRF microscope (TIRF-M) was combined with a Nikon Ti-U inverted microscope. Optical trapping was performed using a near infrared trapping laser at 1070 nm using a Nikon 100x, oil immersion, NA 1.49 TIRF objective lens. For GFP and RFP chromophores fused to the proteins of interest were excited using 488 and 561 nm laser diode, respectively. Their Fluorescent emissions were detected using two electron multiplying charge-coupled device (EMCCD, iXon, Andor) cameras. The sample (~5 mm 2 leaf tissue) was mounted on a computer-controlled variable speed (Märzhäuser) stepper motor stage. The associated computer-controlled hardware was interfaced using National Instruments LabVIEW which provides full automation for each trapping routine. The power of the optical trap laser transmission was set to 40.7 mW. The TIRF image was recorded from 0 s, the trap was turned on at 1 s, the translation stage movement of 10 μm at 2 μm/s begins at 5 s and ends at 10 s, the trap was deactivated at 11 s, and the image recording stops at 22 s (relating to 11 s recovery periods). A 10 μm distance threshold was chosen to ensure consistence between experiments by which chloroplasts were moved; in particular, 10 μm was chosen as it is not so far as to generate potential side effects such as pushing the chloroplast into the tonoplast, yet large enough to move chloroplasts a substantial distance from their original position.

Micrograph quantification
Chloroplast stromule quantification was done automatically using a MATLAB script. Stromules were manually counted using a semi-automated MATLAB script. Percentage of chloroplasts with stromules were calculated by dividing the number of chloroplasts one (or more) stromule(s) by the total number of chloroplasts.
Quantification of haustorial-chloroplast-nucleus accumulation was done manually from original confocal micrographs by looking through each layer of the z-stack with only the brightfield and P. infestans 88069td channels active to identify haustoria without bias. Association of chloroplasts and or nuclei to these marked haustoria was then counted, assessing each layer of the z-stack as opposed to viewing a z-projection that could cause false-positives. For Fig. 1C & S3B, two individuals independently quantified chloroplasthaustoria association, with discrepancies re-checked to reach a consensus quantification.

Mock haustoria
Mock haustoria were applied to each of the confocal micrographs that were used in the data sets that contributed to Fig. 1A and Fig. 2 using standard ImageJ line tools. All channels were turned off except the brightfield and the P. infestans 88069td channel to reduce bias. To position the mock haustoria, a straight line was drawn that bisected the base (where the haustoria enters the plant cell) to the tip of each actual haustoria, extending across the vacuole until the cell border opposite was hit (as seen from the brightfield). This end position was taken to be the position at which a mock haustoria penetrated the cell. The line width was set to 3 µm (based on the general observation that haustoria were 2.7 µm in width). All channels were turned back on and the region around the mock haustoria and the instances of chloroplast-nucleus presence adjacent to these mock haustoria were counted, note only the z-slices containing the actual haustoria from which the mock was positioned were counted.
This method was used to keep the number of mock and actual haustoria per cell similar for comparison, but in the following cases, a mock haustoria could not be placed: if the mock haustoria position overlaps with an actual haustoria, if the cell border in out of the field of view, if the actual haustoria has developed in the crook of a cell and is touching both cell borders. If no mock haustoria could be successfully placed, the image was not included in the pair-wise comparison.

Statistical analysis
Statistical significance of the differences observed when comparing means (stromule quantification) were assessed by t test when found to be normal by Shapiro-Wilk test. If data was found to be non-normally distributed Wilcox statistical test was implemented by R. A pair-wise Wilcoxon test was used to compare mock v actual haustoria accumulation (Fig. S1B, Fig. S2), with samples paired by data coming from the same micrograph. Counts of haustoria-chloroplast-nucleus association were pooled together to generate an overall proportion/percentage from all micrographs instead of treating each micrograph as technical repeat and its taking percentage association as a single data point; this was done to avoid data skew by micrographs that contained only one haustoria and would therefore generate many 100% and 0% values that skew the mean estimate. The proportions of each observation were compared using a Fisher's Exact test in R where statistical comparison was made.

Chloroplast automated counting algorithm through image processing
The image processing algorithms were used to calculate the gradient of the image to identify the boundaries of the puncta. Enclosed regions formed by the boundaries were algorithmically identified and counted. This procedure was done for each individual channel green (in chloroplast stroma) and blue (Chloroplast Auto-fluorescence). The chloroplast (GFP channel) containing stromule were counted in a semi-automated fashion.

Western blotting
Protein extraction, purification and western blot analysis steps were performed as described previously (Bozkurt et al., 2011). Anti-phosphorylated MAPK (Phospho-p44/42 MAPK, Cell Signaling Technology) was used as primary antibody, anti-rabbit (Sigma-Aldrich, UK) antibody was used as secondary. Figure S1: Mock haustoria positioned opposite real haustoria are less frequently adjacent to chloroplasts. (A) Confocal micrograph of N. benthamiana infected with P. infestans strain 88069td as an example of how mock haustoria were positioned and used to assess the rates of chloroplast-haustoria association that might be expected by chance encounter. White arrow heads show actual haustoria in the cell. Red lines show trajectory from an actual haustoria to the opposite site of the infected cell, the point at which a mock haustoria was then marked. Cross ( †) shows association of actual haustoria with a chloroplast, asterisk (*) shows the association of a mock haustoria with a chloroplast. Shaded area is another infected cell, removed to simplify the explanation of the method. (B) Paired samples scatter plot showing the image-by-image percentage of chloroplast association between the actual and mock haustoria (N = 39 images total). Black cross bar marks the mean. Samples are paired based on derivation from the same confocal image, paired samples are joined by a line that is colored based on the value difference between them. Asterisks denote p < 0.01 as determined by a paired samples Wilcoxon test.       Video S10: 3D image of Fig. S5A showing chloroplasts form long-distance stromule interactions that can bridge more than one haustorium. 3D visualization comprises Zstack of confocal images of leaf epidermal cells from transplastomic CpGFP (Yellow) N. benthamiana plants infected with red-fluorescent P. infestans strain 88069td (Magenta). 'H' indicates haustorium.

Supplementary information
Video S11: 3D image of chloroplast and stromules embracing a haustorium. 3D visualization comprises Z-stack of confocal images of leaf epidermal cells from transplastomic CpGFP (Yellow) N. benthamiana plants expressing PM and EHM marker RFP:REM1.3 (Magenta) infected with wild-type P. infestans 88069.
Video S12: Time-lapse series showing optical trapping of chloroplast in Fig. 3D and co-migration of a second chloroplast interacting via a stromule-like extension. TIRF microscopy combined with laser capture in leaf epidermal cells from transplastomic CpGFP (Grayscale) N. benthamiana plants. Scale bar is 10 µm.