Autophagy mediates temporary reprogramming and dedifferentiation in plant somatic cells

Abstract

Somatic cells acclimate to changes in the environment by temporary reprogramming. Much has been learned about transcription factors that induce these cell‐state switches in both plants and animals, but how cells rapidly modulate their proteome remains elusive. Here, we show rapid induction of autophagy during temporary reprogramming in plants triggered by phytohormones, immune, and danger signals. Quantitative proteomics following sequential reprogramming revealed that autophagy is required for timely decay of previous cellular states and for tweaking the proteome to acclimate to the new conditions. Signatures of previous cellular programs thus persist in autophagy‐deficient cells, affecting cellular decision‐making. Concordantly, autophagy‐deficient cells fail to acclimatize to dynamic climate changes. Similarly, they have defects in dedifferentiating into pluripotent stem cells, and redifferentiation during organogenesis. These observations indicate that autophagy mediates cell‐state switches that underlie somatic cell reprogramming in plants and possibly other organisms, and thereby promotes phenotypic plasticity.

Autophagy facilitates cellular proteome adjustment to new stimuli and allows coordinated transition to a new cell state by erasing previous cellular programs.

Introduction

Somatic cells in multicellular eukaryotes are relentlessly exposed to diverse physiological and environmental stimuli including changes in temperature, nutrients, hormones, and pathogen load (Cherkasov et al, 2013; Chovatiya & Medzhitov, 2014). At certain levels, such stimuli become stressful and provoke adaptive cellular responses (Galluzzi et al, 2018). To survive, eukaryotes have evolved sophisticated acclimation mechanisms that mediate temporary reprogramming of somatic cells. In both animals and plants, these mechanisms include alterations in transcriptional activities and epigenetic signatures (Davière & Achard, 2016; Xu et al, 2017; Koo & Guan, 2018; Zhang et al, 2018; Hafner et al, 2019).

Somatic cells can also undergo directional reprogramming through dedifferentiation and can form pluripotent cells. This allows somatic cells to redifferentiate into other cell types, organs, and even whole organisms in plants (Takahashi & Yamanaka, 2006; Papp & Plath, 2013; Ikeuchi et al, 2015; Li & Belmonte, 2017; Li et al, 2017). Similar to temporary reprogramming, reprogramming into other cell types is orchestrated by evolutionarily conserved processes and involves major changes in the transcriptome and epigenetic landscape (Roche et al, 2017; Sang et al, 2018; Iwafuchi‐Doi, 2019). Despite the wealth of knowledge on initial transcriptional and epigenetic changes driving somatic reprogramming events, how proteostasis mechanisms delete current cellular states to allow installment of new programs remains largely unknown.

Macroautophagy (hereafter autophagy) is a conserved quality‐control pathway that facilitates cellular adaptation by removing superfluous or damaged macromolecules and organelles (Popovic & Dikic, 2014; Liu & Klionsky, 2015; Ho et al, 2017). Although initially discovered as a starvation‐induced survival mechanism in yeast (Yang & Klionsky, 2013), many studies have now shown that autophagy plays crucial roles in a variety of stress responses (Bassham et al, 2006; Mizushima et al, 2008; Munch et al, 2014; Rui et al, 2015; Katheder et al, 2017; Kumsta et al, 2017; Dikic & Elazar, 2018) and may act as both positive and negative regulator of programmed cell death (Gutierrez et al, 2004; Nakagawa et al, 2004; Liu et al, 2005; Berry & Baehrecke, 2007; Hofius et al, 2009). Autophagy has also been implicated in induced pluripotent stem cell (iPSC) formation, cellular regeneration and stem cell survival in mammals (Saera‐Vila et al, 2016; Boya et al, 2018; Calvo‐Garrido et al, 2019), and cell fate determination of embryo suspensor in plants (Minina et al, 2013). However, some of these studies have contrasting conclusions. For example, autophagy was shown to have opposite functions in mammalian cells during reprogramming into pluripotency (Wang et al, 2013; Wu et al, 2015) and stem cell maintenance in mice (Mortensen et al, 2011; Ho et al, 2017). So, how can we reconcile these functions and discrepancies regarding the function of autophagy? And is autophagy involved in iPSC formation in plants?

Unlike reprogramming in stem cells, temporary reprogramming events in somatic cells are reversible and provide phenotypic plasticity in response to various stimuli (Fusco & Minelli, 2010; Pfennig et al, 2010; Kelly et al, 2012; Oostra et al, 2018). Although autophagy possesses the degratory capacity to mediate rapid cell‐state switches, whether it is involved in temporary reprogramming of somatic cells remains unknown. Here, we find that autophagy functions in various cellular reprogramming events in plants. Stimuli as diverse as phytohormones, danger signals, and microbial elicitors all trigger rapid and robust activation of autophagy. Using quantitative proteomics, we show that autophagy mediates the switch between somatic cell programs by removing cellular components that are no longer required. At the same time, autophagic mechanisms ensure a controlled execution of the newly established programs. Accordingly, autophagic dysfunction leads to defects in organismal fitness, dedifferentiation of somatic cells into pluripotency, and redifferentiation of pluripotent cells into other cell types in plants.

Results

Autophagy is rapidly engaged upon perception of diverse stimuli

To examine whether temporary reprogramming engages autophagy, we exposed young seedlings of the model plant Arabidopsis thaliana expressing the autophagic markers GFP‐ATG8a or YFP‐mCherry‐NBR1 (Svenning et al, 2011) to an array of treatments. We evaluated autophagic flux in response to a selection of microbial elicitors, danger signals, and hormones known to induce temporary reprogramming: peptide‐1 (PEP1, a small peptide produced during wounding) and ATP, which are perceived as danger‐associated molecular patterns (DAMPs); abscisic acid (ABA, a hormone commonly associated with abiotic stress responses); 1‐aminocyclopropane‐1‐carboxylic acid (ACC, precursor of the gaseous hormone ethylene involved in development and senescence); brassinolide (BL, a steroid hormone involved in growth); 1‐naphthalene acetic acid (NAA, a synthetic auxin involved in growth modulation); and 6‐benzylaminopurine (6‐BA, a synthetic cytokinin involved in cytokinesis and growth). All of these treatments induced rapid accumulation of GFP‐ATG8a (Fig 1A and B) and YFP‐mCherry‐NBR1 foci (Fig EV1A and B). GFP‐ATG8a vacuolar degradation produces free GFP fragments that can be detected by immunoblotting to measure autophagic flux (Mizushima et al, 2010). All of the treatments induced accumulation of free GFP, pointing to increased autophagic flux (Fig 1C). Further corroboration of autophagic flux increase came from immunoblotting against native NBR1, a well‐known autophagy receptor (Svenning et al, 2011) (Fig 1D). Because high NBR1 turnover complicates this analysis in wild‐type plants, we used atg2‐2 (Wang et al, 2011) instead and observed increased levels of NBR1 in all treatments (Fig 1D), further confirming the induction of autophagy during temporary reprogramming events. Taken together, our results indicate that regardless of the nature of the signal, autophagy is rapidly induced and may function as an intrinsic component in temporary reprogramming of somatic cells.

Figure 1. Autophagy is rapidly engaged upon perception of diverse stimuli.

GFP‐ATG8a expressing seedlings in Murashige and Skoog (MS) growth medium or 30 min after treatment with MS containing ACC, ABA, ATP, BL, 6‐BA, Flg22, NAA, or PEP1.

1. Representative maximum intensity projection images of 10 Z‐stacks per image. Scale bar: 10 μm.
2. Quantification of GFP foci per 0.0025 mm². Values are presented as mean ± standard deviation of the mean and were calculated from at least three independent experiments with three individuals per replicate. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.
3. GFP‐ATG8a cleavage immunoblot for plants exposed to the same treatments as in (A). Numbers below the blots represent ratio for given sample normalized to input and relative to non‐treated control. Experiments were repeated minimum 3 times with similar results.
4. NBR1 immunoblot for atg2‐2 samples for given treatments. Numbers below the blots represent ratio for given sample normalized to input and relative to non‐treated control. Experiments were repeated minimum three times with similar results.

Source data are available online for this figure.

Figure EV1. Autophagy is rapidly induced upon recognition of a wide range of stimuli.

YFP‐mCherry NBR1 accumulation before or 30 min after treatment with ACC, ABA, ATP, BL, 6‐BA, Flg22, NAA, and PEP1.

1. Representative maximum intensity projection images of 10 Z‐stacks per image. Scale bar: 10 μm.
2. Quantification of YFP/mCherry foci for given treatments per 0.0025 mm². Values given are mean ± standard deviation of the mean and are based on three independent experiments, with three individuals per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.

Autophagy facilitates temporary reprogramming

We hypothesized that the primary function of autophagy in temporary reprogramming is to assist cellular “clean‐up” to allow a new program to unfold before returning to basal levels. If so, (i) a second reprogramming stimulus should re‐activate autophagy, and (ii) establishment of the second program should be concurrent with a rapid decay of the first program. To test this, we applied consecutive stimuli and examined reprogramming from ABA (abiotic stress proxy) to flg22 (immunity stress proxy) and NAA (growth and development) to 6‐BA (growth and development; Figs 2A and D, and EV2A and C). We quantified GFP‐ATG8a foci (Fig 2B and E), YFP‐mCherry‐NBR1 foci (Fig EV2B and D), and free GFP via the cleavage assay (Fig 2C and F). All of these assays demonstrated that autophagic flux resets to basal levels after 16 h of ABA or NAA treatment, contrasting with the rapid induction seen before (Fig 1). Transferring those seedlings to flg22‐ or 6‐BA‐containing medium caused reactivation of autophagy as demonstrated by significant accumulation of GFP and YFP‐mCherry‐positive foci (P < 0.05, Figs 2B and E, and EV2B and D) and free GFP (Fig 2C and F), in comparison with the control treatment. Hence, our data indicate that autophagy is engaged to clean‐up, is reset after the clean‐up, and can be reactivated upon perception of new stimuli.

Figure 2. Autophagy is reactivated upon contrasting stimulus perception.

1. Seedlings were acclimated for 16 h in MS containing ABA and then imaged 30 min after being swapped to MS containing ABA/NAA (control), MS, or MS containing flg22. Images are representative maximum intensity projection of 10 Z‐stacks per condition. Experiments were repeated three times independently with similar results. Scale bar: 10 μm.
2. Quantification of GFP foci per 0.0025 mm², for samples treated as described in (A). Values are presented as mean ± standard deviation of the mean and are based on three independent experiments, with three individuals per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.
3. GFP‐ATG8a cleavage immunoblot for given treatments. The ratio of free GFP to loading control, normalized to the 16‐h pre‐treated sample (set to 1), is provided below each band.
4. Seedlings were acclimated for 16 h in MS containing NAA and then imaged 30 min after being swapped to MS containing ABA/NAA (control), MS, or MS containing 6‐BA. Images are representative maximum intensity projection of 10 Z‐stacks per condition. Experiments were repeated 3 times independently with similar results. Scale bar: 10 μm.
5. Quantification of GFP foci per 0.0025 mm², for samples treated as described in (D). Values are presented as mean ± standard deviation of the mean and are based on three independent experiments, with three individuals per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.
6. GFP‐ATG8a cleavage immunoblot for given treatments. The ratio of free GFP to loading control, normalized to the 16‐h pre‐treated sample (set to 1), is provided below each band.

Source data are available online for this figure.

Figure EV2. Autophagy is reactivated upon contrasting stimulus perception.

1. Seedlings were acclimated for 16 h in MS containing ABA and then imaged 30 min after being swapped to MS containing ABA/NAA (control), MS, or MS containing flg22. Images are representative maximum intensity projection of 10 Z‐stacks per condition. Experiments were repeated 3 times independently with similar results. Scale bar: 10 μm.
2. Foci quantification for given treatments per 0.0025 mm². Values given are mean ± standard deviation of the mean and are based on three independent experiments, with three individuals per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to a t‐test.
3. Seedlings were acclimated for 16 h in MS containing NAA and then imaged 30 min after being swapped to MS containing ABA/NAA (control), MS, or MS containing 6‐BA. Images are representative maximum intensity projection of 10 Z‐stacks per condition. Experiments were repeated 3 times independently with similar results. Scale bar: 10 μm.
4. Foci quantification for given treatments per 0.0025 mm². Values given are mean ± standard deviation of the mean and are based on three independent experiments, with three individuals per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to a t‐test.

To further support our observations, we performed comparative proteomics using Tandem Mass Tag labeling (TMT) Mass Spectrometry (MS/MS) on wild‐type (WT) and the autophagy‐deficient mutant atg2‐2 upon consecutive, temporary reprogramming inducing stimuli ABA and flg22 (Fig 3A). We detected 11,300 proteins, of which 1,241 responded to the treatments (Table EV1). Based on their behavior, we divided these proteins into fifty clusters. Validating our proteome profiling approach, various ATG8 isoforms and NBR1 clustered together and accumulated to higher levels in atg2‐2 (Fig EV3A). We then searched for proteins induced by ABA that decreased when switched to flg22 treatment in WT plants but failed to decrease in atg2‐2 (Fig 3B, correlation = 0.86, 10.5% of all responding proteins). Importantly, most proteins with this profile also decreased faster in WT plants when switched from ABA to flg22 than to control media (Fig 3C). Several proteins fitting this profile have been previously associated with ABA responses, among them TSPO, which is degraded through autophagy upon completion of the ABA program (Vanhee et al, 2011). Using a TSPO antibody (Guillaumot et al, 2009), we confirmed that TSPO follows the same pattern in another autophagy‐deficient mutant, atg5‐1 (Thompson et al, 2005), during consecutive reprogramming from ABA to flg22 (Fig EV3B), as well as during reprogramming from ABA to NAA (Fig EV3C). These results indicate that autophagy is activated to rapidly remove components of previous cellular programs.

Figure 3. Autophagy facilitates temporary reprogramming during perception of contrasting stimuli by removing old components and modulating the intensity of new responses.

• A
  Schematic representation of the strategy used for consecutive stress treatment.
• B
  Pattern correlation used to find proteins that accumulate upon ABA/NAA treatment and are removed in WT but not in atg2 after swapping to flg22/6‐BA.
• C–F
  Protein clusters obtained after quantitative proteomics of WT (green) and atg2‐2 (magenta) samples treated as described in (A). (C) Protein cluster fitting the pattern displayed in (B). (D) Protein cluster for proteins that accumulate to higher levels in atg2 than WT upon treatment with flg22. (E) Protein cluster of proteins which accumulate upon NAA treatment and are removed in WT but not in atg2 after swapping to 6‐BA. (F) Protein cluster for proteins that accumulate to higher levels in atg2 than WT upon treatment with 6‐BA.

Source data are available online for this figure.

Figure EV3. Protein level change upon treatments with consecutive stresses.

1. Proteins that accumulate in atg2 in comparison with WT.
2. TSPO immunoblot for WT, atg2‐2, and atg5‐1 samples treated as described in Fig 2A for the ABA to flg22 consecutive stress set.
3. TSPO immunoblot for WT, atg2‐2, and atg5‐1 samples treated as described in Fig 2A for the ABA to NAA consecutive stress set.
4. CAT immunoblot for WT, atg2‐2, and atg5‐1 samples treated as described in Fig 2A for the NAA to 6‐BA consecutive stress set.

Our clustering analyses revealed that some stress‐related proteins peaked to much higher levels in atg2‐2 upon switching from ABA to flg22 (Fig 3D, 4.6% of all responding proteins). As expected, many of these such as ATPXG2, PDF2.1, and its close homolog AT1G47540 have previously been associated with immune responses (Petersen et al, 2000; Tsiatsiani et al, 2013; Zhao, 2015). This indicates that autophagy also modulates the intensity of a new cellular program when it is being installed.

To confirm that autophagy functions as an intrinsic component in cellular reprogramming, we extended our proteomic analysis and examined reprogramming between the contrasting developmental phytohormones auxin (NAA) and cytokinin (6‐BA; Table EV2). Here we also observed major proteostatic dysregulation in atg2‐2 plants and identified a major cluster (Fig 3E, 10.2% of all responding proteins), comparable to our previous observations (Fig 3A). We identified catalase 2 (CAT2) in this cluster and used a catalase antibody to confirm the same pattern in both atg5‐1 and atg2‐2 (Fig EV3D). Interestingly, we observed that auxin‐responsive proteins accumulate in untreated atg2‐2 (Fig 3E), unlike ABA‐responsive proteins (Fig 3A). Since stress programs (ABA) are normally “off” under normal growth conditions, while growth and development programs (auxin) are recruited continuously, gradual accumulation of auxin‐responsive proteins may not be surprising in autophagy‐deficient backgrounds. Similar to our observations above (Fig 3D), these results again show that autophagy is also needed to modulate the intensity of new cellular programs when switched from auxin to cytokinin (Fig 3F, 3% of all responding proteins).

Importantly, our proteomic data ([Link], [Link]) also show that other proteostasis mechanisms may also function during temporary reprogramming. When plants are moved from ABA to flg22, the level of several proteins declines in the absence of autophagy, including ABF3 (Table EV1, cluster 1), an ABA‐inducible protein known to be degraded by the proteasome (Chen et al, 2013). Similarly, among proteins that decline in both WT and atg2‐2 when transferred from NAA to 6‐BA, we detected PIN2/EIR1 (Table EV2, cluster 1), also previously reported as a proteasomal target (Abas et al, 2006). Together, these results indicate that the proteasome, like autophagy, also participates in removing components from previous programs during temporary reprogramming.

Autophagy deficiencies lead to reduced phenotypic plasticity and increased heterogeneity

The above results indicate that cells lacking autophagic activity lose cellular homeostasis and accumulate signatures of different cellular programs and states. If so, autophagy deficiency may lead to increased heterogeneity during acclimatization to fluctuating environmental conditions. To assess this at an organismal level, we used a high‐throughput phenotyping chamber to compare the development of WT, atg2‐2, and atg5‐1 mutant plants grown in standard, stable conditions versus plants grown in highly variable conditions recorded for the Swedish spring of 2013 (Fig 4A). Data dispersion for dry weight was higher in atg plants than in WT, regardless of the conditions tested, albeit with more outliers for atg2 grown under variable conditions (Figs 4B and C, and EV4A and B). This indicates that the loss of cellular homeostasis in atg mutants translates into higher heterogeneity, and this increased heterogeneity might stem from their decreased ability to cope with daily variations that involve dynamic, temporary reprogramming events.

Figure 4. Autophagy deficiency leads to reduced phenotypic plasticity and increased heterogeneity.

1. Weather pattern recorded for the Swedish spring of 2013.
2. Dry weight of WT, atg5‐1, or atg2‐2 plants grown under stable (21/16°C 16/8‐h photoperiod) or following the Swedish Spring of 2013. Box plots: Centerlines show the medians; box limits indicate the 25^th and 75^th percentiles; and whiskers extend to the minimum and maximum. Significance calculated by Kruskal–Wallis or Feltz and Miller test for the equality of coefficients of variation as explained in Materials and Methods.Asterisks represent significant pairwise differences (****P < 0.0001).
3. Box plots displaying the heterogeneity of samples in (B). Box plots: Centerlines show the medians; box limits indicate the 25^th and 75^th percentiles; and whiskers extend to the minimum and maximum. Significance calculated by Kruskal–Wallis or Feltz and Miller test for the equality of coefficients of variation as explained in Materials and Methods. Asterisks represent significant pairwise differences (**P < 0.01, ****P < 0.0001).

Figure EV4. Autophagy deficiency leads to reduced phenotypic plasticity and increased heterogeneity.

A, B Representative picture of WT, atg2‐2, and atg5‐1 grown under stable conditions (A) or under Swedish spring of 2013 (B).

Dedifferentiation is severely impaired in autophagy‐deficient cells

To substantiate the importance of autophagy in preserving phenotypic plasticity, we examined iPSC formation and organ regeneration—well‐known examples of somatic cell reprogramming—in autophagy‐deficient plants. Plant PSCs can be induced in vitro by adjusting the ratios of auxin and cytokinin to trigger dedifferentiation and proliferation of an unorganized mass of pluripotent cells (callus) (Sugimoto et al, 2010; Sang et al, 2018). When we placed WT, atg2‐2, and atg5‐1 root explants in callus‐inducing medium (CIM) to induce PSC and callus formation (Valvekens et al, 1988), WT root explants dedifferentiated into PSC and formed visible calli, but atg2 and atg5 root explants did not (Fig 5A). When we then transferred these to shoot‐inducing medium (SIM) (Valvekens et al, 1988) to trigger de novo organogenesis, 80% of WT explants formed shoots after 21 days but atg2‐2 (5%) and atg5‐1 (20%) were severely defective in shoot formation (Fig 5A and B). These results are consistent with data from zebrafish in which autophagy deficiency impairs the reprogramming that is necessary for muscle regeneration (Saera‐Vila et al, 2016) and also reduces self renewal and proliferative ability of hematopoietic stem cells (Ho et al, 2017).

Figure 5. Dedifferentiation is severely impaired in autophagy‐deficient cells.

1. Representative images of root explants from WT, atg2‐2, and atg5‐1 in CIM (6 days) or CIM + SIM (6 + 21 days). Scale bar: 2 cm
2. Quantification (%) of explants presenting shoots after 21 days after incubation on SIM. Results were obtained from three independent experiments with at least 20 calli per condition. Bars marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.

Unlike mammals and flowering plants, bryophytes can naturally form PSCs without the need for exogenous hormone treatment or overexpression of transcription factors. In particular, the moss Physcomitrella patens is able to dedifferentiate somatic cells into chloronema stem cells to repair damaged tissue upon wounding (Ishikawa et al, 2011; Kofuji & Hasebe, 2014; Li et al, 2017). Using this system, we compared reprogramming efficiency in WT, atg5, and atg7 lines upon wounding. As it can be seen in Fig EV5A–D, autophagy‐deficient moss lines are slower and less efficient than WT. These results indicate that, like iPSC formation in Arabidopsis (Fig 5) and zebra fish muscle regeneration (Saera‐Vila et al, 2016), autophagy deficiency in P. patens severely impairs iPSC formation and tissue regeneration. This suggests autophagy‐mediated dedifferentiation and organogenesis are evolutionary conserved across kingdoms.

Figure EV5. Autophagy is necessary for wound‐induced dedifferentiation and tissue repair in Physcomitrella patens .

1. Representative image of gametophore leaves from WT (Gransden) and atg5, undergoing dedifferentiation and cell protrusion 144 h after wounding.
2. Number of gametophore leaves displaying cell protrusions after wounding for the genotypes given. Results are given as mean ± standard deviation of the mean of at least 30 individuals from three independent experiments. Points marked with an asterisk (*) are statistically significant (P < 0.05) according to the t‐test.
3. Representative image of gametophore leaves from WT (Reuter) and atg7, undergoing dedifferentiation and cell protrusion 144 h after wounding.
4. Number of gametophore leaves displaying cell protrusions after wounding for the genotypes given. Results are given as mean ± standard deviation of the mean of at least 30 individuals from three independent experiments. Points marked with an asterisk (*) are statistically significant (P < 0.05) according to a t‐test.

Autophagy‐deficient cells lose control over de novo organogenesis upon prolonged cultivation in pluripotent cell inducing media

Since atg2‐2 and atg5‐1 mutants made small amounts of callus, we wondered whether prolonging CIM treatment could help them in callus formation. We therefore extended the time explants were kept on CIM from 6 to 21 days, while maintaining the subsequent SIM step to 21 days (Weigel & Glazebrook, 2009). As expected, at 21 days on CIM, atg2‐2, and atg5‐1 mutants produced more calli, but still less than WT (Fig 6A and B). Interestingly, when then moved to SIM, atg mutants quickly started to catch up in callus mass (Fig 6A and C) and produced significantly more callus tissue and shoots (P < 0.05) than WT. This is consistent with our proteomic data; atg mutants cannot modulate cellular programs and thus end up with exaggerated calli and shoot formation. This is analogous to comparing acceleration down a developmental “slope” with (WT) or without (atg mutants) brakes. These results are in agreement with those of Ho et al (2017) who demonstrated that hematopoietic cells with impaired autophagy lost their stemmnes and accelerated differentiation. While Ho and colleagues concluded that the altered metabolic rate of atg‐deficient cells led to enhanced differentiation, our data also support a model in which gradual accumulation of conflicting programs, which are not “cleaned” from the cell, could reduce stem cell control.

Figure 6. Autophagy‐deficient cells lose control over de novo organogenesis and senesce earlier upon prolonged cultivation in pluripotent cell inducing media.

• A
  Representative images of root explants from WT, atg2‐2, and atg5‐1 in CIM (21 days) or CIM + SIM (21 + 21 days or 21 + 35 days). Scale bar: 2 cm.
• B, C
  Fresh weights of calli after incubation on CIM (B) or CIM + SIM (21 + 21) (C). Box plots: Centerlines show the medians; box limits indicate the 25^th and 75^th percentiles; and whiskers extend to the minimum and maximum. Results were obtained from three independent experiments with at least 20 calli per condition, asterisks mark statistical significance to WT according to the t‐test (**P < 0.001; ***P < 0.0001).

As autophagy helps to fight aging (Kaushik & Cuervo, 2015; Fernández et al, 2018; Leidal et al, 2018) and preserve mammalian stem cell function (Vilchez et al, 2014; Ho et al, 2017), we wanted to address this in the context of plant stem cells. For this, we prolonged calli culture on SIM to 5 weeks and observed that atg2‐2 and atg5‐1 mutant calli displayed premature senescence and death (Fig 6). Taken together, our results demonstrate that autophagy mediates cellular reprogramming necessary for iPSC formation, modulates subsequent de novo organogenesis, and assists in maintaining longevity of plant iPSC masses.

Conclusion

In summary, our results at the cellular and organismal level show that autophagy is rapidly activated upon perception of diverse stimuli to maintain cellular competence by mediating cellular clean‐up during various reprogramming events in plants. Defects in autophagy led to increased heterogeneity and chaotic cellular decisions, presumably because signatures of previous programs are not removed efficiently and interfere with the execution of new programs. Because accumulation of these programs happens gradually during development, the longer cells experience this proteostasis deficiency, the higher becomes the probability for conflicting cellular decisions. Thus, the numerous, apparently opposite conclusions on autophagic functions in different developmental systems (Liu et al, 2005; Hofius et al, 2009; Mortensen et al, 2011; Ho et al, 2017) may be partially explained by stochasticity emerging from random accumulation of proteins in autophagy‐deficient backgrounds over time. Similarly, difficulties with clearing cellular programs may also explain some of the reported discrepancies on the role of autophagy in reprogramming of somatic cells into PSCs (Wang et al, 2013; Wu et al, 2015). Given enough time, autophagy deficiencies generate cellular environments in which pluripotent cells proliferate and organs are formed de novo without control. Once pluripotency is achieved, autophagy functions as a brake to keep subsequent tissue/organ regeneration steps at “cruise control”. Altogether, our results suggest an evolutionarily conserved function of autophagy in preserving cellular homeostasis during somatic cell reprogramming. As reprogramming is central to environmental acclimation and tissue regeneration in all organisms, our findings have exciting implications for fundamental and applied biology in plants and animals.

Materials and Methods

Experimental model and subject details

Arabidopsis plants were grown in 9 × 9‐cm pots in growth chambers at 21°C and 70% relative humidity and with an 8‐h photoperiod. The intensity of the light was set at 140 μE/m²/s. The following Arabidopsis lines were used in this study: Columbia (Col‐0), atg2‐2 (Wang et al, 2011), atg5‐1 (Thompson et al, 2005), GFP‐ATG8a (Svenning et al, 2011), and YFP‐mCherry‐NBR1 (Svenning et al, 2011). Arabidopsis callus was grown in 9‐cm Petri dishes at 21°C 16/8‐h day/night photoperiod. For immunoblot and proteomic experiments, seedlings grown on solid MS medium (0.44% w/v agar, 1% w/v sucrose, pH 5.7) were kept under 16 h of light (140 μE/m²/s) at 21°C, after seed surface sterilization with 1.3% v/v bleach followed by 70% ethanol. Seedlings were then moved from solid to liquid MS media and let to acclimate for 2 days before experiments were performed.

For callus experiments, roots were excised and placed in CIM (1× Gamborg's B5 salts with vitamins, 20 g glucose, 0.5 g/l MES, 8 g/l agar, 0.5 mg/l 2,4‐D, and 0.05 mg/l kinetin). The pH was adjusted to 5.7 using 1.0 M KOH. After 6 days, calli were moved to SIM (1× Gamborg's B5 salt mixture with vitamins, 20 g/l glucose, 0.25 g/l MES, 8 g/l agar, 5 mg/l 6‐(γ,γ‐dimethylallylamino)purine (2‐IP), and 0.15 mg/l indole‐3‐acetic acid [IAA]). The pH was adjusted to 5.7 using 1.0 M KOH. For long‐term CIM incubation (21 days), SIM recipe was adjusted as follows: 1× MS salts with vitamins, 30 g/l sucrose, 7.5 g/l agar, 0.1 mg/l IAA, and 1.0 mg/l BAP. pH was set to 5.8 with 1M NaOH.

To investigate the rate of reprogramming initiation in Physcomitrella, the top part of 4‐week‐old gametophores were isolated and the tips were dissected with a surgical knife and placed on a new plate overlaid with cellophane. The gametophore tips were checked for reprogramming activation every 12 h using a Leica MZ16 F Fluorescence Stereomicroscope. For imaging, gametophore tips were dissected as described above and the individual tips were placed in small dots of 2% methyl cellulose 15cP (Sigma) in an empty Petri dish, overlaid with cellophane, and cooled BCD‐AT media on top (Bressendorff et al, 2016). Pictures were taken every 24 h using a Sony α6000 camera mounted on a Leica MZ16 F Fluorescence Stereomicroscope.

Plant chemical treatments

For confocal microscopy, seedlings were set on liquid half‐strength MS media and the elongation and meristematic zone of the roots were visualized 7 days after germination. Samples were treated with half‐strength MS media containing one of the following: Flg22 (1 μM), a peptide from bacterial flagellum that elicits immune responses; PEP1 (1 μM), a small danger peptide produced upon wounding; ATP (100 μM), perceived as a DAMP; ABA (1 μM), a hormone associated with abiotic stress responses; ACC (10 μM), a precursor of ethylene involved in development and senescence; BL (10 nM), a steroid hormone involved in growth; NAA (1 μM), an auxin analogue involved in growth modulation; and 6‐BA (1 μM), a synthetic cytokinin regulating development. For state switch experiments, plants were pre‐treated with MS + ABA (1 μM) for 16 h, then media was removed, and plants were gently washed with fresh MS before treating the plants with MS + ABA (1 μM) as control, MS + Flg22 (1 μM), or just half‐strength MS for 30 min. The same setup was used for the NAA to 6‐BA state switch.

For TSPO Western blot and ABA/Flg22 proteomic experiments, plants were pre‐treated with half‐strength MS + ABA (1 μM) for 16 h, then media was removed, and the plants were gently washed with fresh MS before treatment with half‐strength MS + ABA (1 μM) as control, MS + Flg22 (1 μM), or just half‐strength MS for 3 h. For TSPO Western blot in ABA/NAA, samples were treated as above but instead of flg22, samples were treated with MS + NAA (1 μM) for 3 h. For NAA/6‐BA proteomics and catalase Western blots, plant were treated as above but pre‐treatment was MS + NAA (1 μM) for 16 h and samples were changed to MS + 6‐BA (1 μM) or MS for 3 h. Samples were then collected and flash‐frozen for further analyses.

Confocal and light microscopy

All images were taken using a LSM700 Zeiss confocal microscope. All Arabidopsis root images were taken with a 63× water objective. The confocal images were analyzed with Zen2012 (Zeiss) and ImageJ software.

Protein extraction

Protein was extracted as described previously (Rodriguez et al, 2018). In brief, a buffer containing 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 10% (v:v) glycerol (Applichem), 10 mM DTT (Applichem), 10 mM EDTA (Sigma), 0.5% (v:v), PVPP (Sigma), protease inhibitor cocktail (Roche), and 0.1% (v:v) Triton X‐100 (Sigma) was used to extract proteins. Afterward, 3× SDS with 50 mM DTT was added to the samples and this was followed by 20‐min centrifugation at 4°C and 13,000 g. Supernatant was then collected and heated at 95°C for 5 min before loading samples for SDS–PAGE.

SDS–PAGE and immunoblotting

Protein samples were separated on 12% SDS–PAGE gels, electroblotted to nitrocellulose membrane (GE Healthcare), then blocked (1 h in 5% [w:v] BSA [Merck] or 5% [w:v] milk in TBS [50 mM Tris–HCl, pH 7.5 150 mM NaCl, 0.1% Tween‐20 [Sigma]), and incubated 2 h to overnight with primary antibodies: anti‐NBR1 (AS14 2805, Agrisera; 1:5,000), anti‐TSPO (Guillaumot et al, 2009), anti‐GFP (TP401 AMSBio; 1:1,000), or anti‐CAT (AS09501, Agrisera; 1:2,000). Membranes were incubated in secondary anti‐rabbit HRP conjugate (Promega; 1:5,000) for 1 h. Chemiluminescent substrate (homemade or ECL Plus; Pierce) was applied before exposure to camera detection. Western blot band quantification was performed with FIJI (Schindelin et al, 2012) and normalized to loading controls.

Quantitative proteomics

Frozen plant material (500 mg) was lysed in lysis buffer (4% SDS, 100 mM DTT, 100 mM Tris–HCl, pH 7.5), and supernatant was collected after centrifugation at 20,000 g for 15 min. Samples (8 μg/ml concentration) were used for mass spectrometry measurements. FASP and desalting steps were performed as previously described (Käll et al, 2007). These samples are then labeled with TMT according to the manufacturer's instructions (Thermo Fisher). Labeled samples were separated into fractions using an SCX system (Thermo Fisher) and analyzed in LC‐MS/MS (Roitinger et al, 2015). SCX was performed using an Ultimate System (Thermo Fisher) at a flow rate of 35 μl/min and a TSKgel column (ToSOH; 5‐μm particles, 1 mm i.d. × 300 mm). The flow‐through was collected as a single fraction, along with the gradient fractions, which were collected every minute. In total, 130 fractions were collected and stored at −80°C.

For data analysis, raw files were processed in Proteome Discoverer (version 1.4.1.14, Thermo Fisher Scientific, Bremen, Germany). MS Amanda (Dorfer et al, 2014) (version 1.4.14.8240) was used to perform a database search against the TAIR10 database supplemented with common contaminants. Oxidation of methionine was set as dynamic modification and carbamidomethylation of cysteine as well TMT at lysine, and peptide N‐termini were defined as fixed modifications. Trypsin was defined as the proteolytic enzyme allowing for up to two missed cleavages. Mass tolerance was set to 5 ppm for precursors and 0.03 Da for fragment masses. Reporter ion intensities were extracted in Proteome Discoverer using the most confident centroid within an integration boundary of 10 ppm. Identified spectra were FDR filtered to 0.5% on PSM level using Percolator. Peptides shorter than seven amino acids were removed from the results. Identified peptides were grouped to proteins applying strict maximum parsimony. Quantification of proteins is based on unique peptides only. Quantified proteins were exported and further processed in the R environment (version 3.4.3). Proteins were ranked by their similarity to an expected regulation pattern according to Pearson correlation. Furthermore, proteins regulated more than 1.5‐fold were subdivided into clusters using k‐means clustering.

Plant propagation and high‐throughput phenotyping

Seeds were stratified at 4°C in the dark for 4 days in the phytotron. The substrate (Einheitserde, ED63) was sieved (6 mm) and every single pot was filled with the same amount of substrate, 70–72 g) by using a scale to facilitate a homogenous packing density. The prepared pots were all covered with blue mats to enable a robust performance of the high‐throughput image analysis algorithm (Junker et al, 2015). Seedlings (96) for Col‐0 (WT), atg5‐1, and atg2‐2 genotypes were propagated. The individual plants were arranged randomly by shelf. For randomization, an in‐house developed R‐based randomization tool was used.

The HT plant phenotyping phytotron was set to controlled conditions at 21°C during day with a night drop to 16°C at night, 60% rel. humidity, and ca. 160 μmol light with a very balanced light spectrum. For the control experiment, these conditions were set for the entire duration of the experiment.

For the stress experiment, the environmental conditions were changed at 21 days after sowing to a dynamic simulation of the Swedish spring by using the hourly recorded data of the Swedish field site (Ullstorp) during 25.4. – 31.5. 2013.

The dry weight was scored by harvesting a random sample of 36 replicates per genotype (plants cut off at the base just above the soil).

This particular phytotron allows high‐throughput plant phenotyping with the integrated and automated x, y, z sensor‐to‐plant RGB imaging system, delivering images of 1,260 plants in one go. Images were taken twice a day during standard conditions, in the morning 1 h after the lights went on and in the late afternoon, 1 h before lights went off. During the Swedish conditions, the number of pictures per day was increased to five in order to score possibly quick changes in the phenotypes. In total, 102,059 pictures were taken for the experiment with the simulation of Swedish spring.

For the data processing, quality control, and preliminary analysis, we have used PHENOComp, an R package developed by the VBCF BioComp facility. The R programming environment (R Core Team, 2018) was used for performing the statistical analysis and generating the visualizations for the Figs. Data clean‐up and transformations were done using the tidyr package (https://cran.r-project.org/package=tidyr). Datasets were checked for normality and heteroscedasticity to select the appropriate statistical tests. Statistical significance for differences in mean values for dry weight and seed weight was determined using the Kruskal–Wallis non‐parametric test. Statistical significance for differences in the variance was determined using Levene's test from the car package (https://cran.rproject.org/package=car). In both cases, pairwise comparisons were calculated using Dunn's test as implemented in the PMCRM package (https://cran.r-project.org/package=PMCMR) and corrected for multiple testing using the Holm method. The post hoc test for pairwise comparisons of the variances between groups was calculated using the absolute values of the residuals (deviation from the median of the group), as recommended in Boos and Brownie (2005). We tested for difference in the coefficient of variation using the asymptotic Feltz and Miller test for the equality of coefficients of variation (Feltz & Miller, 1996), as implemented in the package cvequality (https://cran.rproject.org/package=cvequality). Plots were produced using the packages ggplot2 (https://CRAN.R-project.org/package=ggplot2) and ggpubr (https://CRAN.Rproject.org/package=ggpubr.

Author contributions

Conceptualization: MP and ER; Investigation: ER, JC, JA, VK, FK, JO, JC, CL, MS, SK, JJ, ZZ and ENE; Formal analysis: ER, JC, JA, JO, GD, KM, PSD, YD; Resources: SS, SL; Writing: MP, YD and ER Visualization: ER, JC, YD; Supervision: MP, YD, ER; Funding Acquisition MP, YD. All authors read the manuscript and agreed with the findings reported.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

The EMBO Journal (2020) 39: e103315

Data availability

“The mass spectrometry data from this publication have been deposited to the PRIDE archive (https://www.ebi.ac.uk/pride/archive/) and assigned the identifier [PXD016575]”.