Functional diversity of isoprenoid lipids in Methylobacterium extorquens PA1

Hopanoids and carotenoids are two of the major isoprenoid‐derived lipid classes in prokaryotes that have been proposed to have similar membrane ordering properties as sterols. Methylobacterium extorquens contains hopanoids and carotenoids in their outer membrane, making them an ideal system to investigate the role of isoprenoid lipids in surface membrane function and cellular fitness. By genetically knocking out hpnE and crtB we disrupted the production of squalene and phytoene in M. extorquens PA1, which are the presumed precursors for hopanoids and carotenoids respectively. Deletion of hpnE revealed that carotenoid biosynthesis utilizes squalene as a precursor resulting in pigmentation with a C30 backbone, rather than the previously predicted canonical C40 phytoene‐derived pathway. Phylogenetic analysis suggested that M. extorquens may have acquired the C30 pathway through lateral gene transfer from Planctomycetes. Surprisingly, disruption of carotenoid synthesis did not generate any major growth or membrane biophysical phenotypes, but slightly increased sensitivity to oxidative stress. We further demonstrated that hopanoids but not carotenoids are essential for growth at higher temperatures, membrane permeability and tolerance of low divalent cation concentrations. These observations show that hopanoids and carotenoids serve diverse roles in the outer membrane of M. extorquens PA1.


| INTRODUC TI ON
Microorganisms can withstand a diversity of environmental stresses ranging from extreme temperatures to the immune defenses of multicellular organisms. The cell surface membrane serves as a first line of defense against environmental perturbations and the membrane's lipid composition is critical for stress resistance. On the one hand, the membrane must be robust enough to withstand chemical and physical challenges. On the other hand, the membrane must be fluid enough to support bioactivity. In eukaryotic organisms such as yeast, sterols play a crucial role in achieving a fluid yet mechanically robust cell surface membrane (Mouritsen & Zuckermann, 2004).
The absence of sterols from most prokaryotes suggests that alternate lipids may serve analogous roles in surface membranes. All three domains of life possess isoprenoid synthesis pathways derived from a common C 5 isoprene building block which give rise to a broad suite of diverse lipid classes including sterols, but also carotenoids and hopanoids, and the majority of archaeal lipids. Because of their structural similarities that are derived from a common C 5 isoprene building block, resulting in rigid and often semi-planar structures, isoprenoid-derived lipids may share certain biophysical features in membranes (Ourisson et al., 1987). However, the mechanism and exact influence of isoprenoid lipids on prokaryotic membrane properties and cellular fitness remains relatively unexplored.
There is increasing evidence pointing to the role of bacterial isoprenoid-derived lipids such as hopanoids and carotenoids in membrane stabilization in bacteria (Belin et al., 2018;Bramkamp & Lopez, 2015;Tookmanian et al., 2021). Hopanoids are predominately found in Gram-negative bacteria where they have been shown to order outer membrane lipids by interacting with lipid A in a similar manner to that exhibited by cholesterol and sphingolipids in eukaryotes (Saenz et al., 2012(Saenz et al., , 2015Silipo et al., 2014). In contrast, carotenoids (βcarotene and zeaxanthin) have been shown using molecular dynamics (MD) simulations to have a condensing effect similar to that of cholesterol on phospholipids (Mostofian et al., 2020). Physiologically, there is evidence that hopanoids are important for growth at higher temperatures (Belin et al., 2018;Doughty et al., 2011;Kulkarni et al., 2013a;Poralla et al., 1984;Schmidt et al., 1986), whereas carotenoids have been linked to cold acclimation in some bacteria (Chattopadhyay & Jagannadham, 2001;Fong et al., 2001;Seel et al., 2020). These contrasting phenotypes for temperature acclimation suggest that hopanoids and carotenoids may serve opposing roles in modulating membrane properties. Taken together, these observations suggest functional similarities between sterols and bacterial isoprenoid lipids.
However, the extent to which carotenoids and hopanoids have analogous or diverging biophysical properties and functions in biomembranes is not known and has not been systematically explored in a living model system. How do hopanoids and carotenoids contribute to the role of the outer membrane in adaptation to varying temperatures?
Methylobacterium extorquens is a Gram-negative bacterium with a well-characterized genome and a simple lipidome (Chwastek et al., 2020) that produces both hopanoids and carotenoids. This makes it an attractive model organism for studying the global phenotypes of disrupting the two pathways. In this study we have genetically disrupted the biosynthetic pathways of the two main isoprenoid lipid precursors; squalene (precursor for hopanoids) and phytoene (precursor for carotenoids), thus confirming the function of the gene hydroxysqualene oxidoreductase (hpnE) in M. extorquens PA1.
Additionally, we show that even though the genome of M. extorquens has the genes for the canonical C 40 -carotenoid biosynthetic pathway, the pigmentation has a C 30 -based backbone that is squalene derived.
We propose that the genes for the C 30 squalene derived pathway were acquired through lateral gene transfer (LGT). We demonstrate the importance of hopanoids, but not carotenoids for growth at high temperature, and low divalent cation concentration, as well as maintaining low membrane permeability. In contrast, the carotenoid pathway may play a role in protection against oxidative stress.

| Characterizing the function of the genes hpnE and crtB in M. extorquens PA1
To compare the effects of hopanoids and carotenoids on growth, adaptation, and outer membrane mechanics, we first aimed to create strains deficient in either hopanoid or carotenoid synthesis. Hopanoids are derived from squalene (Kannenberg & Poralla, 1999) and carotenoids were previously predicted to be derived from phytoene in M.
extorquens (Dien et al., 2003). Therefore, we first disrupted the biosynthesis of the isoprenoid precursors squalene and phytoene by deleting the putative squalene synthase gene hydroxysqualene oxidoreductase (hpnE) and the phytoene synthase gene (crtB). We next aimed to evaluate the effect of knocking out hpnE and crtB on hopanoid and carotenoid biosynthesis. First, we ran total lipid extracts of the mutant strains on thin layer chromatography (TLC) and, as expected, diplopterol (hopanoids) were not present in the ∆hpnE strain ( Figure S1). LC-MS analysis also revealed that the ∆hpnE strain no longer produced detectable amounts of diplopterol (hopanoids; Table 1), and that it accumulated hydroxysqualene which is the precursor of squalene biosynthesis (Table 1 and Figure S2; Pan et al., 2015). Next, we measured the absorbance spectra of lipids extracted from the strains WT, ∆crtB and ∆hpnE as a readout for carotenoid pigmentation. Surprisingly, the ∆crtB mutant strain showed no loss in pigmentation compared to the WT (Figure 1a), whereas the ∆hpnE mutant strain was non-pigmented ( Figure 1b). These unexpected observations led us to investigate whether carotenoid biosynthesis in M. extorquens was derived from squalene (Furubayashi et al., 2014) rather than phytoene.  (Goodwin, 1980;Umeno et al., 2002). However, recently a squalenederived C 30 carotenoid pathway was discovered and has been identified in a few species, including Planctomycetes, which also produce hopanoids (Santana-Molina et al., 2020;Steiger et al., 2012;Takaichi et al., 1997;Taylor, 1984). Hence, we analyzed the distribution of that these genes were transferred together, that is, in the same DNA fragment/locus. Therefore, unlike CrtI-CrtD or HpnCDE enzymes which indicated an ancestral feature of Proteobacteria, the C 30 carotenoid pathway in some Alphaproteobacteria orders suggest that they originated later by LGT from Planctomycetes.

| Carotenoids are derived from the C 30 pathway in M. extorquens PA1
In order to confirm that carotenoid biosynthesis uses squalene as extorquens all have a C 30 backbone, and no C 40 backbone-based carotenoids were detected confirming its squalene origin ( Figure S3).
Our observations demonstrated that the deletion of genes in the proposed squalene-derived C 30 carotenoid pathway (crtN, crtP) produced non-pigmented mutant strains, where the phenotype was eliminated by gene complementation on an inducible plasmid ( Figure S4a,b). Whereas knocking out genes in the C 40 carotenoid biosynthesis pathway had no effect on pigmentation, LC-MS analysis confirmed the presence of a C 30 backbone of the carotenoid pigment extracted from the WT, ∆shc and ∆crtB strains ( Figure S3).
These results suggest that C 40 biosynthetic pathway was not used to synthesize carotenoids, at optimal growth conditions in M. extorquens PA1.
When characterizing phenotypes of mutant strains, it is import- at the tested conditions ( Figure S3). However, we observed an accumulation of squalene in the ∆shc mutant strain as detected by LC-MS and TLC (Table 1 and Figure S5), which could generate artifactual   (Saenz et al., 2012). We previously showed that temperature has one of the largest effects on lipidomic remodeling and growth rate, relative to other experimental parameters such as detergent and salt concentrations in M. extorquens (Chwastek et al., 2020). Here, we observed that interrupting hopanoid biosynthesis (∆hpnE, ∆shc) caused a growth impairment especially at temperatures higher than the optimum (30℃) (Figure 4a).
Moreover, growth of the ∆shc mutant strain was impaired even more than for the ∆hpnE mutant strain, which could potentially be linked

| Disruption of isoprenoid synthesis influences lipid packing, membrane permeability, and sensitivity to divalent cation concentration
The high temperature growth impairment observed for mutant strains that cannot produce hopanoids implicated a membraneinduced defect. Therefore, we next asked how hopanoid and carotenoid synthesis influenced outer membrane properties. Such an approach is complicated by the fact that membranes have complex lipidomes that can compensate to varying degrees for the loss of individual lipid species. While some lipids are crucial for viability (e.g., sterols in mammalian cells), others may have little or no effect at all when they are removed from the system. In this sense, the uniqueness of a lipids' property or function is implicated by how much its absence affects the membrane. In this regard, rather than asking what particular biophysical effect different isoprenoids have on membranes, we instead asked how much does the absence of a particular isoprenoid pathway impact the ability for the membrane to maintain its physical properties, using the WT as a reference.
To investigate the membrane properties of different hopanoid knockout strains, we used the lipophilic dye Di-4 ANEPPDHQ (Di-4) which reports on lipid packing density. The emission spectrum of Di-4 is sensitive to changes in lipid packing density and can be semiquantitatively measured through the calculation of the general polarization (GP) index, whereby higher GP values indicate more densely packed lipids (Amaro et al., 2017). Lipid packing is correlated with a number of key membrane properties including viscosity and bending rigidity, thereby providing a robust and sensitive readout of variations in the physical state (Ma et al., 2018;Steinkühler et al., 2019). It has been previously shown that Di-4 selectively labels the surface membrane, F I G U R E 4 Growth phenotypes of disrupted isoprenoid biosynthesis for physical and chemical stress conditions. Growth rate comparison with WT at different temperatures of (a) Hopanoid-deficient strains, (b) C 30 carotenoid-deficient strains. Growth of Methylobacterium extorquens PA1 mutant strains under various stress conditions assessed by disk diffusion assay with Ampicillin, Penicillin, Chloramphenicol, Tetracycline, Vancomycin, and H 2 O 2 of (c) Hopanoid-deficient strains, (d) C 30 Carotenoid-deficient strains. Error bars represent standard deviation (n = 3). *p < .05, **p < .01, and ***indicates weak inhibition

(d)
most likely due to its bulky polar headgroup which prevents flipping to the inner leaflet (Lorent et al., 2020). We therefore used Di-4 to measure in vivo outer membrane lipid packing of the WT and isoprenoiddeficient mutant strains ∆crtN, ∆crtP, ∆shc, and ∆hpnE ( Figure 5).
Our first observation was that the GP values obtained for WT in vivo were comparable to the GP values reported by Sáenz et al.
for the in vitro measurement of purified outer membranes (Saenz et al., 2015). This confirmed that under normal conditions, Di-4 is selectively reporting outer membrane lipid packing. ∆crtN, and ∆crtP mutant strains had slightly increased lipid packing compared to the WT strain ( Figure 5). Since precursors for carotenoid synthesis were not detectable in ∆crtN, and ∆crtP mutant strains ( Figure S3), this suggested that increased lipid packing was due to the absence of carotenoids, and not the accumulation of precursors. In contrast we observed dramatically lower Di-4 GP values in ∆shc, and ∆hpnE mutant strains. On the one hand, this could indicate that the absence of hopanoids resulted in lower lipid packing, which we previously demonstrated in purified outer membranes from WT and ∆shc mutant strains (Saenz et al., 2015). However, GP values that we observed in vivo for ∆shc and ∆hpnE mutant strains were considerably lower than observed for ∆shc purified outer membranes (Saenz et al., 2015), raising the possibility that Di-4 was not selectively reporting outer membrane lipid packing in the hopanoid-deficient mutants.
Since hopanoids are crucial for modulating lipid order and membrane permeability Mangiarotti et al., 2019;Poralla et al., 1980;Saenz et al., 2012Saenz et al., , 2015Schmerk et al., 2011), it is possible that deletion of hopanoids dramatically impaired the integrity of the outer membrane, allowing normally impermeable molecules like the GP value of the WT and carotenoid-deficient strains is similar to purified outer membranes (Saenz et al., 2015), and closer to the values of liposomes composed of SM (Sphingomyelin) and cholesterol, which are in a liquid ordered state ( Figure S6). Thus, the large negative shift in GP values for the hopanoid-deficient mutants indicates that Di-4 is reporting the less ordered phospholipid inner membrane rather than exclusively the more ordered outer leaflet of the outer membrane. Taken together, these observations point toward dramatically increased outer membrane permeability in the hopanoid-deficient mutants.
To measure changes in membrane permeability we used an assay that determined relative changes in fluorescein diacetate (FDA) diffusivity . Fluorescein diacetate is non-fluorescent, and rapidly diffuses through the membrane where it is hydrolysed by . We therefore examined whether supplementing the growth medium with Ca 2+ could rescue growth and membrane permeability of the hopanoid-deficient strains. First, we evaluated the effect of increasing Ca 2+ concentration on WT, ∆shc and ∆hpnE. We observed a negligible effect for WT, but saw a rescue of both ∆shc and ∆hpnE growth with an optimal concentration around 1.67 mM Ca 2+ ( Figure S8). We next examined growth with amended (1.67 mM) and unamended Ca 2+ with varying temperature and observed a rescue of growth of both ∆shc and ∆hpnE mutant strains at higher temperature (34℃; Figure 7). Finally, we measured relative changes in membrane permeability by FDA hydrolysis (Figure 6), and observed that permeability was rescued Furthermore, this suggested that the hopanoid pathway provides a means for maintaining outer membrane stability for Gram-negative bacteria that occupy environments with low or fluctuating divalent cation concentrations. of crtB showed no phenotype in pigmentation, whilst deletion of hpnE yielded non-pigmented mutant strains that also lacked hopanoids. These unexpected results revealed that carotenoid biosynthesis was derived from squalene rather than phytoene through a non-canonical pathway that has recently been shown to produce C 30 carotenoids (Furubayashi et al., 2014).  l C a C -2 l C a C + 2 9 1 0 M M L p absence of any significant phenotype associated with disrupted carotenoid synthesis was surprising, since it has been shown in model membranes that carotenoids could share some of the lipid ordering properties of sterols (Gabrielska & Gruszecki, 1996;Kostecka-Gugała et al., 2003;Mostofian et al., 2020;Socaciu et al., 2000;Subczynski et al., 1992). We did observe a small increase in lipid packing in the carotenoid-deficient mutants, opposite to what is predicted from measurements in model membranes but similar to observations in Pantoea sp. (Kumar et al., 2019). In the case of Pantoea sp., increased membrane rigidity (qualitatively comparable to increased lipid packing) was attributed to a compensatory increase in the abundance of saturated acyl chains, rather than the depletion of carotenoids (Kumar et al., 2019). However, it has been shown that, in some bacteria, carotenoid production is increased at lower temperatures (Fong et al., 2001;Seel et al., 2020), pointing toward a role in fluidizing membranes. Carotenoids have a plethora of diverse structures that could alter their physicochemical effects on the membrane (Milon et al., 1986;Wisniewska et al., 2006), which could account for diverse biophysical properties. Alternatively, carotenoids may serve in a different capacity unrelated to the physical properties of the membrane in M. extorquens. For example, carotenoids play an important role in light scavenging in photosynthetic organisms (Polívka & Frank, 2010), and protecting cells from oxidative stress (Kim et al., 2019;Kumar et al., 2019). Indeed, a role for carotenoids in oxidative stress is suggested by our observation of increased sensitivity of ∆crtN and ∆crtP to hydrogen peroxide. Taken together our observations indicate that the carotenoid pathway is not essential for temperature adaptation or membrane permeability in M. extorquens, but may play a role in protection against oxidative stress.

| D ISCUSS I ON
Having established that carotenoids do not account for the growth and membrane phenotypes resulting from disrupted squalene synthesis we next examined the contribution of the hopanoid pathway. Hopanoids have been shown to modulate bacterial membrane properties in a manner analogous to eukaryotic sterols (Saenz et al., 2012(Saenz et al., , 2015. In M. extorquens disruption of hopanoid synthesis by deleting the gene shc resulted in a large growth deterioration at higher temperatures as well as increased membrane permeability comparable to what we observed for deleting the gene hpnE. It has previously been shown in other organisms that hopanoids are associated with sensitivity to high temperatures (Belin et al., 2018;Doughty et al., 2011;Kulkarni et al., 2013b;Poralla et al., 1984;Schmidt et al., 1986), and MD simulations also suggest that hopanoids could reinforce membranes at higher temperatures (Caron et al., 2014). The ordering effect of hopanoids, like sterols, has also been shown to reduce membrane permeability Mangiarotti et al., 2019;Poralla et al., 1980;Schmerk et al., 2011). We further demonstrated that the growth deficiency and increased membrane permeability could be partially rescued by increasing the Ca 2+ concentration of the media. Since Ca 2+ and other divalent cations stabilize the outer membrane by ordering lipopolysaccharide, this result has two important implications. First it provides evidence that the phenotypes associated with disrupted hopanoid synthesis are linked to a reduction in lipopolysaccharide packing density from the loss of hopanoids. Second, the sensitivity of hopanoid-deficient mutants to stress at lower Ca 2+ concentrations suggested that hopanoids may be especially important for Gram-negative bacteria that occupy environments with low divalent cation concentrations.
Methylobacterium extorquens is on its way toward becoming a well-characterized and robust model system for studying the role of lipid structure in membrane function and organismal fitness. It was recently shown that M. extorquens has the simplest lipidome so far observed in any organism (Chwastek et al., 2020), making it an ideal system for exploring the principles of lipidome adaptation.
While the phospholipidome is relatively well-explored by comparison, the role of isoprenoid lipids is still relatively undefined. Our observations show that carotenoids and hopanoids serve divergent roles in the outer membrane. By revealing that carotenoids are squalene-derived and identifying genes in the carotenoid pathway, this study now provides a new tool to explore the property-function relationship of carotenoids and their relationship with hopanoids in Methylobacterium.

| Media, growth conditions
Methylobacterium strains were grown at 30℃ in minimal medium described by Delaney et al. (2013) referred to as hypho medium, with 9.9 mM disodium succinate (Sigma Aldrich, W327700) as the carbon source at 160 rpm shaking (ISF1-X Kuhner shaker). Escherichia coli strains were grown at 37℃ in LB medium (Carl Roth, X968).
Triparental conjugation was performed on Nutrient broth medium (Carl Roth, X929.1). All solid media plates were prepared with 1.5% Agar-Agar (Carl Roth, 1,347). Antibiotics for selection were at the

| Evolutionary analyses for C 30 and C 40 carotenoid pathway
We performed protein searches of CrtI (P54980), CrtD (Q01671), CrtN (O07855), and CrtP (Q2FV57) against NCBI database using phmmer (Finn et al., 2011) and with e-value threshold of 1e-5. We combined all the sequences obtained and using GTDB taxonomy (Parks et al., 2018), we removed redundant sequences by taxonomic orders (from 90 up to 50% of identity threshold for the less and more represented groups respectively). We then aligned this set of nonredundant sequences using MAFFT (Katoh & Standley, 2013) and performed a fast phylogenetic tree using FastTree (Price et al., 2009) to exclude spurious sequences. Once we obtained the final set of sequences, we re-aligned with MAFFT and removed those enriched gap positions using trimAl (Capella-Gutiérrez et al., 2009). For the final phylogenetic reconstruction, we used IQ-TREE (Nguyen et al., 2015). We obtained branch supports with the ultrafast bootstrap (Hoang et al., 2017) and the evolutionary models were automatically selected using ModelFinder (Kalyaanamoorthy et al., 2017) implemented in IQ-TREE and chosen according to BIC criterion.
For the phylogenetic profile, the distribution of HpnCDE, Sqs and CrtB enzymes were obtained from data previously generated (Santana-Molina et al., 2020). The distribution of FAD-dependent desaturases was obtained from the phylogenetic reconstruction performed in this study. The taxonomic tree was obtained from GTDB repository (https://gtdb.ecoge nomic.org/) pruning those sequences of interest. Phylogenetic trees were visualized and annotated in iTOL (Letunic & Bork, 2019).

| Disk diffusion stress assays
The protocol was adapted after the Kirby Bauer method (Hudzicki, 2009)

| Carotenoid extraction for absorbance scan
Bligh and Dyer extraction (Bligh & Dyer, 1959) was used to extract carotenoids for cells grown to late exponential. 10 ml of cells of different Methylobacterium mutants were collected at 5,000 rcf, for 10 min, washed once with 1× D-PBS. Wet weight of cell pellet was weighed. Cells were resuspended in water to 200 µl (taking weight into account), adding 250 µl chloroform (Carl Roth, Y015), and 500 µl of methanol (VWR chemicals,20,903.368

| Isolation and saponification of carotenoids
Carotenoids extraction was adapted from Kim and Lee (2012).
Briefly, 10 mg wet weight of each sample was extracted using 1 ml methanol containing 6% KOH and incubated for at least 14 hr at 4℃ in the dark. Supernatant was collected after centrifugation (1,500×g, 5 min) and reduced in a speedvac concentrator (Savant SPD111V; Thermo Fisher Scientific, Massachusetts, USA). EtOAc and saturated NaCl were added in equal volumes while thoroughly mixing after each addition. Upper organic phase was collected after centrifugation (10,000×g at 4℃ for 5 min), washed twice with distilled water, and completely dried.  where I is the intensity measured at the specified channels.

| Liposome preparation and in vitro DI-4 spectroscopy
Liposomes were prepared from two mixtures of lipids: DOPC to achieve a liquid disordered membrane and SM:

| Strains, construction of plasmids, generation of mutants, and gene complementation
M. extorquens PA1 with cellulose synthase deletion was used in this study and referred to hereafter as WT (Chubiz et al., 2013;Delaney et al., 2013), ∆shc was already available (Saenz et al., 2015). Genes for carotenoid biosynthesis were identified based on M. extorquens gene annotations for phytoene desaturase and phytoene synthases, and BLASTp was used to reconstruct the carotenoid biosynthesis pathway as shown.
Mutants were constructed by unmarked allelic exchange as described (Hmelo et al., 2015;Marx, 2008), for each gene primers were designed to include 500 bp upstream and 500 bp downstream overhangs of the gene. The produced PCR product was then used as a template for the construction of two plasmids one to delete the gene and one for the inducible expression of the gene in the knockout strain as explained in (Table 2). For gene deletion: plasmid pCM433 (Marx, 2008) was linearized via restriction digestion using enzymes NotI-HF, and SacI-HF (NEB, R3189, R3156 respectively), overhangs upstream and downstream of the gene of interest were amplified (primers sequences available in Table S1) and purified then cloned into linearized pCM433 using In-Fusion HD Cloning plus kit (Takara), primer design was done using primer design tool (Takara).
For inducible expression of the gene: plasmid pLC291 was linearized GP = I 540nm − I 670nm I 540nm + I 670nm GP = I 540nm − I 670nm I 540nm + I 670nm using EcoRI-Hf, and KpnI-HF restriction enzymes (NEB R3101, R3142 respectively), gene was then PCR amplified and purified then cloned into plasmid pLC291 (Chubiz et al., 2013) using In-Fusion HD Cloning plus kit (Takara). All PCR products and linearized vector were purified (Macherey and Nagel, Nucleospin PCR clean-up Gel extraction).
Deletion vectors were introduced into WT via triparental conjugation. WT cells were mated with E. coli pRK2073 helper cells, and E. coli Stellar cells that carry the deletion/expression vector, and the mating was done using a ratio of 5:1:1 Acceptor-strain : helperstrain : donor-strain. The conjugation was done on NB-Agar plates at 30℃, overnight, then the cells were plated on hypho media-agar plates with Tmp and Tc. The clones were then grown for 9 hr in liquid media, then plated on 10% sucrose plates for selection of mutants.
Colony PCR was then performed on clones from the sucrose plates, using Primers (Table S1) for gene template. PCR products of the correct size for gene deletion were then sequenced to confirm deletion of genes.

ACK N OWLED G M ENTS
The authors thank members of the Sáenz group especially Grzegorz Chwastek; André Nadler, and Michael Schlierf for discussions; André Nadler for manuscript comments; Lisa-Maria Müller and Lisa Junghans for technical assistance. This work was supported by the B CUBE,

TU Dresden, a German Federal Ministry of Education and Research
BMBF grant (to J.S., project 03Z22EN12), and a VW Foundation "Life" grant (to J.S., project 93090). The Authors declare no conflict of interest. Open Access funding enabled and organized by Projekt DEAL.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.