International Journal of Molecular Sciences Article Pathological ATX3 Expression Induces Cell Perturbations in E. coli as Revealed by Biochemical and Biophysical Investigations Diletta Ami 1,2,† , Barbara Sciandrone 1,† , Paolo Mereghetti 3 , Jacopo Falvo 1 , Tiziano Catelani 1 , Cristina Visentin 4 , Paolo Tortora 1 , Salvador Ventura 5 , Antonino Natalello 1,2, * and Maria Elena Regonesi 1,2, * 1 Department of Biotechnologies and Biosciences, University of Milano-Bicocca, 20126 Milan, Italy; diletta.ami@unimib.it (D.A.); barbara.sciandrone@unimib.it (B.S.); j.falvo@campus.unimib.it (J.F.); tiziano.catelani@unimib.it (T.C.); paolo.tortora@unimib.it (P.T.) 2 Milan Center of Neuroscience (NeuroMI), 20126 Milan, Italy 3 Bioinformatics Consultant, 15061 Arquata Scrivia, Italy; paolo.mereghetti@gmail.com 4 Department of Biosciences, Università degli Studi di Milano, 20133 Milan, Italy; Cristina.Visentin@unimi.it 5 Institut de Biotecnologia i de Biomedicina and Departament de Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, 08193 Barcelona, Spain; salvador.ventura@uab.cat * Correspondence: antonino.natalello@unimib.it (A.N.); mariaelena.regonesi@unimib.it (M.E.R.) † These authors contributed equally to this work. Abstract: Amyloid aggregation of human ataxin-3 (ATX3) is responsible for spinocerebellar ataxia type 3, which belongs to the class of polyglutamine neurodegenerative disorders. It is widely accepted that the formation of toxic oligomeric species is primarily involved in the onset of the disease. For this reason, to understand the mechanisms underlying toxicity, we expressed both a physiological (ATX3-  Q24) and a pathological ATX3 variant (ATX3-Q55) in a simplified cellular model, Escherichia coli. It has  been observed that ATX3-Q55 expression induces a higher reduction of the cell growth compared to Citation: Ami, D.; Sciandrone, B.; ATX3-Q24, due to the bacteriostatic effect of the toxic oligomeric species. Furthermore, the Fourier Mereghetti, P.; Falvo, J.; Catelani, T.; transform infrared microspectroscopy investigation, supported by multivariate analysis, made it Visentin, C.; Tortora, P.; Ventura, S.; possible to monitor protein aggregation and the induced cell perturbations in intact cells. In particular, Natalello, A.; Regonesi, M.E. it has been found that the toxic oligomeric species associated with the expression of ATX3-Q55 are Pathological ATX3 Expression responsible for the main spectral changes, ascribable mainly to the cell envelope modifications. Induces Cell Perturbations in E. coli as A structural alteration of the membrane detected through electron microscopy analysis in the strain Revealed by Biochemical and expressing the pathological form supports the spectroscopic results. Biophysical Investigations. Int. J. Mol. Sci. 2021, 22, 943. https://doi.org/ 10.3390/ijms22020943 Keywords: ataxin-3 expression; amyloids; Escherichia coli; FTIR microspectroscopy; multivariate analysis; oligomer toxicity; protein aggregation Received: 22 December 2020 Accepted: 16 January 2021 Published: 19 January 2021 1. Introduction Publisher’s Note: MDPI stays neutral Protein misfolding and subsequent aggregation in amyloid deposits are associated with regard to jurisdictional claims in with more than 20 neurodegenerative diseases, commonly referred to as “protein mis- published maps and institutional affil- folding diseases” or “conformational diseases” [1]. Examples of these pathologies are iations. Alzheimer’s, Parkinson’s, prion, and polyglutamine (poly-Q) diseases. Distinct amyloido- genic polypeptides are implicated in the progression of these disorders. Protein aggregation into highly ordered β-sheet-rich supramolecular structures is a common feature of many polypeptide chains and a hallmark of the protein misfolding diseases [2,3], despite reduced Copyright: © 2021 by the authors. sequence homology between amyloid-forming polypeptides. Licensee MDPI, Basel, Switzerland. The aggregation process from a native protein state into amyloid deposits impli- This article is an open access article cates several steps. A general scheme provides a misfolded protein that assembles into distributed under the terms and low molecular weight oligomers, followed by the formation of larger aggregates, such conditions of the Creative Commons as protofibrils and mature fibrils [4]. A different pathway to amyloid formation called Attribution (CC BY) license (https:// “first-aggregation-and-then-misfolding” has also been identified, in which aggregation creativecommons.org/licenses/by/ of globular proteins starts from conformational states close to native ones, with no need 4.0/). Int. J. Mol. Sci. 2021, 22, 943. https://doi.org/10.3390/ijms22020943 https://www.mdpi.com/journal/ijms Int. J. Mol. Sci. 2021, 22, 943 2 of 21 for transitions across the major energy barrier for unfolding [5]. One of the proteins that displays this behavior is the Josephin domain (JD) of ataxin-3 (ATX3) [6,7], the protein responsible for spinocerebellar ataxia of type 3 (SCA3), one of the ten known poly-Q disorders [8,9]. ATX3 consists of the N-terminal globular JD followed by an unstructured C-terminal region containing its poly-Q that triggers SCA3 when it exceeds a critical length of 50–55 consecutive residues [10,11]. This implies that the aggregation pathway consists of two steps: the first, only requiring the JD, gives rise to SDS-soluble protofibrils, and the second, solely accessible to variants carrying expanded poly-Qs, results in the formation of mature, SDS-insoluble fibrils [12–14]. Ruggeri and coauthors demonstrated the confor- mational transition of JD from native spheroidal oligomers, through misfolded oligomers, to the protofibrils [7]. In particular, the spheroidal oligomers contain JD with native-like secondary structure, while the conformational transition towards β-enriched structures occurs only later. This underlines the intermediate nature of the oligomers, some sharing the structural properties of native monomers (native-like), the others being quite similar to fibrils (fibril-like) [7]. The interest in oligomer polymorphism arises from the primary role of these species in cellular toxicity [15–17]. A leading theory on the molecular basis of amyloid toxicity argues that unstable early aggregates interact with cell membranes either non-specifically and/or at specific interaction sites, thus disturbing membrane ordered structure and/or impairing the activity of specific membrane proteins involved in signal transduction or ion transport, resulting in permeabilization or signaling deregulation [17–19]. Moreover, in SCA3, the species formed along the ATX3 self-assembly pathway seem to affect different cellular systems, including protein degradation pathways, gene transcription regulation, mitochon- dria and constituents implicated in the calcium homoeostasis [20–23]. It is challenging to discriminate if these effects are related to a toxic gain of function of the aggregated species or to a loss of the pleiotropic functions of ATX3. In this complex scenery, the employment of a simple transgenic system might be useful not only for better understand- ing the pathophysiological basis of the human disease, but also for the development of therapeutic approaches to treat this condition. Escherichia coli has been proved to be a simplified, but effective model system for the study of protein aggregation within a cellular environment [24,25]. An increasing number of published works support a similar relative aggregation propensity for amyloid proteins and protein mutants in humans and, when overexpressed, in bacteria. These results open the possibility of using E. coli for in-cell studies on protein aggregation and for high-throughput/ultrahigh-throughput functional screening of amyloid aggregation inhibitors [26,27]. Interestingly, different aggregated species of a chimeric protein containing a poly-Q tract have been detected during the heterologous expression in E. coli, allowing a better characterization of the complex ag- gregation process occurring in the cellular environment [25]. In a previous work, we used E. coli to reproduce the intracellular two-step aggregation pathway of the expanded ATX3 [28]. Noteworthy, in this model system the proteotoxicity of an expanded ATX3 carrying 55 Gln was found to correlate with the presence of soluble species and not with the amount of insoluble protein aggregates [28]. In this work, we took advantage of Fourier transform infrared (FTIR) microspec- troscopy, supported by multivariate analysis, in combination with other biochemical and microscopical analyses, to study the mechanisms at the basis of ATX3 proteotoxicity. In particular, we investigated in situ the spectral changes that occur in intact E. coli cells expressing two ATX3 variants, the physiological Q24 and the pathological Q55, at two times of induction, i.e., when different protein aggregated species were formed. FTIR mi- crospectroscopy is a non-invasive and label-free vibrational approach that gives a chemical fingerprint of the investigated sample, providing information on the content and structure of its main biomolecules (proteins, lipids, carbohydrates, and nucleic acids) [29]. Due to their complexity, the FTIR spectra of biological systems are the result of the overlapping absorptions of their main biomolecules. For this reason, to obtain the significant and non-redundant information contained in the spectra, the use of an appropriate multivariate Int. J. Mol. Sci. 2021, 22, 943 3 of 21 analysis is necessary [30]. Here, we processed the FTIR data of intact E. coli cells expressing ATX3-Q24 and ATX3-Q55 by Partial Least Squares-Discriminant Analysis (PLS-DA), that enabled not only to evaluate the statistical significance of the observed spectral differences, but also to identify the spectral components responsible for the discrimination among the different analyzed classes, thus providing molecular-level information on the system under investigation. Our biochemical and biophysical investigations allowed correlating the protein ex- pression and aggregation with the induction of the cell perturbations. In particular, we showed that toxic oligomeric species of the pathological expanded ATX3 are responsible for the main spectral changes, ascribable mainly to the cell envelope modifications, in accordance with the severe membrane alterations observed through microscopic analysis. 2. Results 2.1. Toxic Effect of the Pathogenic Ataxin-3 Variant on E. coli Growth In our previous work [28], we demonstrated that the wild type ATX3 protein carrying 24 consecutive glutamines (ATX3-Q24) and the pathological ATX3 variant with 55 residues (ATX3-Q55) reproduce in E. coli cell the aggregation pathway observed in vitro and display different toxicity level; the pathogenic form, indeed, exhibits higher toxicity with respect to the wild type. To investigate the mechanisms at the basis of this difference, we first compared the growth curves of E. coli Rosetta-Gami strains expressing the ATX3-Q24 and the ATX3-Q55 variants, lacking any tag or fusion partner [28]. As a control, we employed the β-lactoglobulin (BLG), a non-amyloidogenic protein that aggregates into inclusion bodies (IBs) when overexpressed in E. coli [31]. Growth profiles were obtained by inducing protein expression after 1 h from the inoculum with 0.4 mM IPTG and curves were followed up to 8 h after induction. According to our previous work [28], the induced strains showed slower growth rates in comparison with the non-induced (NI) control strains (Figure 1A,B). The curves of the strains expressing ATX3-Q24 and BLG were quite similar, thus showing that the effect on cell growth exerted by wild type ATX3 is merely due to a metabolic burden effect caused by heterologous protein over-expression [32,33]. In contrast, the ATX3-Q55 growth curve was constantly slower than the physiological variant, up to two-fold lower at the end of the induction (t = 9 h, Figure 1B). In order to correlate the ATX3-Q55 toxicity to a decrease in cell viability, we analyzed the Colony Formation Unit (CFU) at 5 h and 8 h after induction. The results showed that after 5 h there were no differences in CFU count between the two induced strains and with respect to the relative NI strains. On the other hand, at 8 h after IPTG addition, both induced strains displayed a 10-fold lower CFU count than the NI strains, but there were no significant differences between them (Figure 1C). Taken together, these results demonstrate that, although the expression of the two proteins have the same bactericidal effect on cell viability, the poly-Q expansion gives to the ATX3-Q55 pathological variant an additional bacteriostatic effect through mechanisms that will be further investigated. Int. J. Mol. Sci. 2021, 22, 943 4 of 21 Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 4 of 21 Figure 1. Figure 1. Effects Effects of of ataxin-3 ataxin-3 (ATX3) (ATX3) variants variants expression expression on on cell cell growth growth and and viability. viability. (A) (A) Growth Growth curves curves of of E. E. coli coli Rosetta Rosetta strains expressing ATX3-Q24 (black dots), ATX3-Q55 (light grey dots) and β-lactoglobulin (BLG) (grey squares). Cultures strains expressing ATX3-Q24 (black dots), ATX3-Q55 (light grey dots) and β-lactoglobulin (BLG) (grey squares). Cultures were inoculated at OD600 = 0.1 (t0), induced after 1 h with 0.4 mM of IPTG (t1, black arrow) and growth profiles were were inoculated at OD600 = 0.1 (t0), induced after 1 h with 0.4 mM of IPTG (t1, black arrow) and growth profiles were monitored at 37 °C for 8 h. As control, the same cultures (ATX3-Q24, black stars; ATX3-Q55, light grey stars; BLG, not monitored at 37 ◦ C were filled grey squares) for 8 h. As control, inoculated the same without IPTGcultures addiction(ATX3-Q24, blackNI). (Not Induced, stars; ATX3-Q55, Error light grey bars represent stars;deviations standard BLG, not filled grey squares) were inoculated without IPTG addiction (Not Induced, NI). Error bars represent and are obtained from nine independent experiments. (B) Quantification of optical density increases. Bars represent the standard deviations and fold are obtained increases of E.from nine independent coli Rosetta stains growth experiments. expressing(B) Quantification ATX3-Q24 of optical (black bars), density ATX3-Q55 increases. (light Barsand grey bars) represent the BLG (grey bars)increases fold normalizedof E.on colit0Rosetta of eachstains strain. Error expressing growth bars represent error standard ATX3-Q24 and ATX3-Q55 (black bars), are derived from (light nine grey independent bars) and BLG experiments. (grey Statistical on bars) normalized analysis was made t0 of each strain.between BLG Error bars and ATX3-Q24 represent strains and error standard anddid are not showfrom derived significance. ** p-value nine independent < 0.01, * p-value experiments. < 0.05. (C) Statistical Colony analysis wasFormation Unit (CFU) made between BLG andcount of ATX3-Q24 ATX3-Q24 strainsand andATX3-Q55 did not showI andsignificance. NI strains. At 5 and 8 ** p-value h after IPTG induction, 1 mL of cell cultures were collected, normalized at OD 600 = 0.5 and 1:10 serial dilutions were < 0.01, * p-value < 0.05. (C) Colony Formation Unit (CFU) count of ATX3-Q24 and ATX3-Q55 I and NI strains. At 5 and performed. Moreover, 100 µL of 10−6 and 10−5 dilutions were seeded on single LB agar plate supplemented with ampicillin 8 h after IPTG induction, 1 mL of cell cultures were collected, normalized at OD600 = 0.5 and 1:10 serial dilutions were 100 µg/mL and glucose 0.1%. After drying, colonies were let to growth over-night at 37 °C and then counted. Data were performed. Moreover, 100 µL of 10−6 and 10−5 dilutions were seeded on single LB agar plate supplemented with ampicillin represented as CFU/mL with standard deviations of three independent experiments. 100 µg/mL and glucose 0.1%. After drying, colonies were let to growth over-night at 37 ◦ C and then counted. Data were represented as CFU/mL with standard Takendeviations together,ofthese threeresults independent experiments. demonstrate that, although the expression of the two proteins have the same bactericidal effect on cell viability, the poly-Q expansion gives to the ATX3-Q55Toxicity 2.2. ATX3-Q55 pathological variantto an Is Not Correlated additional a Higher bacteriostatic Protein Expression effectof the or Increase through mechanisms Soluble that will be further investigated. Fraction To assess if the higher toxicity displayed by the ATX3 pathogenic form was due to a 2.2. ATX3-Q55 different proteinToxicity is not we expression, Correlated to a the quantified Higher Protein amount Expression of total ATX3 or Increase protein in of the ATX3-Q24 Soluble Fraction and ATX3-Q55 expressing strains by western blot. Equal amount of cells at different growth timesTo(0,assess 1, 2, 5,ifand the8higher h aftertoxicity displayed induction) by the ATX3 were collected pathogenic and processed form was to obtain totaldue to a protein different extracts andprotein samplesexpression, we quantified were analyzed by SDS-PAGE the amount (Figure of S1)total ATX3 protein and western in ATX3- blot (Figure 2A). Q24 results The and ATX3-Q55 showed aexpressing strainsexpression different ATX3 by western blot. level inEqual the twoamount strains;ofin cells at different particular, the ATX3-Q24 growth times variant (0, 1,seems to be8 expressed 2, 5, and at a higher h after induction) werelevel and forms collected andmore aggregate processed than to obtain the ATX3-Q55 total protein. protein extracts andHowever, samples the were presence analyzed ofbyproteolytic SDS-PAGE fragments (Figure S1)andand aggregates western in both blot strains (Figure 2A). did not The allowshowed results an accurate quantification. a different For this level ATX3 expression reason, we two in the performed strains; dot-blot in analysis particular, the to quantify the ATX3-Q24 ATX3 variant protein seems level. to be The same expressed at asamples higher used level in andwestern forms blot analysis were applied directly on a nitrocellulose membrane more aggregate than the ATX3-Q55 protein. However, the presence of proteolytic in a single spot and ATX3 signals were then quantified by immunofluorescence (Figure 2B,C). The data fragments and aggregates in both strains did not allow an accurate quantification. For this were obtained normalizing reason, with respectdot-blot we performed to the PNPase analysis content (a cytosolic to quantify proteinprotein the ATX3 constantly expressed level. The same at 37 ◦ C and samples onein used ofwestern the mostblot expressed analysisproteins in E. coli were applied [34,35]) directly on and to the t0 signal a nitrocellulose of each membrane in a single spot and ATX3 signals were then quantified by immunofluorescence (Figure 2B,C). The data were obtained normalizing with respect to the PNPase content (a cytosolic protein constantly expressed at 37 °C and one of the most expressed proteins in E. coli Int. J. Mol. Sci. 2021, 22, 943 5 of 21 Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 5 of 21 [34,35]) and to the t0 signal of each strain. The results confirmed the highest ATX3-Q24 strain. The results expression, with anconfirmed the increase of highest about ATX3-Q24 1.5-fold at 8 h expression, withafter (p-value < 0.01) an increase of about induction, with 1.5-fold at 8 h (p-value < 0.01) after respect to the ATX3-Q55 (Figure 2C). induction, with respect to the ATX3-Q55 (Figure 2C). Figure Figure 2. 2. Analysis Analysis of ATX3 protein of ATX3 protein expression expressionatatdifferent differentgrowth growthtimes. times.(A) (A)SDS-PAGE SDS-PAGE(12%) (12%) and and western western blot blot analysis analysis of of protein total extracts (TE) obtained from E. coli Rosetta strains expressing ATX3-Q24 and ATX3-Q55. protein total extracts (TE) obtained from E. coli Rosetta strains expressing ATX3-Q24 and ATX3-Q55. Cells were collected at Cells were collected at different times after induction (t = 0, 1, 2, 5, and 8 h) and processed as detailed in Materials and Methods. different times after induction (t = 0, 1, 2, 5, and 8 h) and processed as detailed in Materials and Methods. Moreover, 20 µL Moreover, 20 µL of TE samples were boiled for 10 min and separated in SDS-PAGE, blotted onto Odyssey nitrocellulose of TE samples were boiled for 10 min and separated in SDS-PAGE, blotted onto Odyssey nitrocellulose membrane (LI-COR). membrane (LI-COR). ATX3 Z46 rabbit polyclonal antibodies [36] and IRDye fluorescent rabbit secondary antibody (LI- COR) Z46 ATX3 was rabbit used forpolyclonal antibodies [36] Black the immunodetection. and IRDye fluorescent bars indicate the rabbit secondary migration antibody of protein (LI-COR) markers. Arrows wasrepresent used for the the immunodetection. migrations Black bars of ATX3-Q24 indicate theThe and ATX3-Q55. migration relatedofgel protein stained markers. ArrowsGel with EZ-Blue represent stainingthe migrations solution of ATX3-Q24 (Sigma) is reportedand in ATX3-Q55. Figure TheDot S1. (B) related blot gel andstained with EZ-Blue (C) densitometric Gel staining analysis of TE.5solution µL of TE(Sigma) is reported samples in Figure were diluted S1. (B) in 150 µL Dot finalblot and (C) volume of Sample Buffer. densitometric Then, ofwere analysis TE.5 µLboiled of TE10samples min andwerevacuum diluted in filtered 150 µL in 48-well final volumedot-blot of Sample apparatus through Buffer. Then, Odyssey were boiled 10 nitrocellulose min and vacuum membrane filtered in(LI-COR). ATX3 variants 48-well dot-blot apparatus and PNPase through protein Odyssey were revealed nitrocellulose using anti-ATX3 membrane (LI-COR). Z46 ATX3and anti- variants PNPase and PNPaserabbit polyclonal protein antibodies were revealed respectively using anti-ATX3[37]. Z46Dots were quantified and anti-PNPase rabbitbypolyclonal densitometry and normalized antibodies respectively using [37]. PNPase as a loading control and t0 signal of each strain as blank. Lines represent ATX3-Q24 (black Dots were quantified by densitometry and normalized using PNPase as a loading control and t0 signal of each strain as dots) and ATX3-Q55 (light grey dots). Error bars represent standard error and derive from at least three independent experiments. ** p-value < blank. Lines represent ATX3-Q24 (black dots) and ATX3-Q55 (light grey dots). Error bars represent standard error and 0.01. derive from at least three independent experiments. ** p-value < 0.01. Finally, Finally, weweanalyzed analyzedthe soluble the solubleATX3 ATX3fraction (Figure fraction 3) to 3) (Figure obtain information to obtain about information the content about of native the content and and of native oligomeric oligomericspecies, which species, whichareareSDS-soluble SDS-solubleand andmigrate migrate as as monomeric monomeric protein in SDS-PAGE [28,38,39]. The quantification of western blot signals protein in SDS-PAGE [28,38,39]. The quantification of western blot signals of of monomeric band did monomeric band did not notdisplay displaysignificant significantdifferences differences between between ATX3-Q24 ATX3-Q24 andand ATX3- ATX3-Q55 Q55 soluble soluble fractions, fractions, at any at any timetime of induction. of induction. Overall, Overall, the data the data demonstrate demonstrate thatthat ATX3- ATX3-Q55 Q55 toxicity is not caused by a higher level of expression or amount of soluble fraction but toxicity is not caused by a higher level of expression or amount of soluble fraction but argue argue the the hypothesis hypothesis of of an an intrinsic intrinsic toxicity toxicity of of ATX3-Q55 ATX3-Q55 oligomeric species. oligomeric species. 2.3. FTIR Analysis of E. coli Cell Modifications Induced by the Expression of ATX3-Q24 and ATX3-Q55 To explore in situ if the expression of the two ATX3 variants differently perturbed the host cells, we applied FTIR microspectroscopy to E. coli intact cells, at 5 and 8 h from protein induction. As a control, we also analyzed NI cells. To investigate the response of the main cell biomolecules, the FTIR analysis was performed in different spectral ranges. We analyzed the second derivatives of the FTIR spectra, which allow resolving the different overlapping components in the complex absorption bands. Moreover, to evaluate the reproducibility of the results and to identify the most significant spectral variations that enabled us to discriminate between induced and NI cells, we supported the FTIR data with a multivariate analysis approach, namely PLS-DA. Furthermore, to investigate if the Int. J. Mol. Sci. 2021, 22, 943 6 of 21 expression of the two protein variants differently perturbed the host cells, the Euclidean distance between the PLS-DA scores was calculated between the induced and NI cells for Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 6 of 21 each ATX3 variant. Figure 3. Analysis of soluble protein fraction of ATX3 variants. (A) Western blot of soluble fractions Figure 3. Analysis of soluble protein fraction of ATX3 variants. (A) Western blot of soluble of E. coli Rosetta strains expressing ATX3-Q24 and ATX3-Q55. Soluble fractions were collected at fractions of E. coli Rosetta strains expressing ATX3-Q24 and ATX3-Q55. Soluble fractions were differentattimes collected aftertimes different induction (t = 0, 1, 2, after induction (t =5,0,and 1, 2,8 5,h)and and8 processed as detailed h) and processed in Material as detailed in and Methods. Material andBlack bars indicate Methods. the indicate Black bars migration theofmigration protein markers. Arrows of protein represent markers. Arrows therepresent migrations theof ATX3-Q24,ofATX3-Q55, migrations ATX3-Q24,and PNPase. and ATX3-Q55, (B) Western PNPase.blot signals were (B) Western quantified blot signals werebyquantified densitometry by and densitometry and normalized normalized respect to PNPaserespect signalstoandPNPase signals t1 signal andstrain. of each t1 signal of each Lines strain.ATX3-Q24 represent Lines (black represent dots) andATX3-Q24 ATX3-Q55(black (light dots) and ATX3-Q55 grey dots) trends. Error(light barsgrey dots) trends. represent standardError bars error and represent derive from at standard error least three and deriveexperiments. independent from at least three independent experiments. 2.3. FTIR For Analysis of E. coli sake of clarity, inCell the Modifications Induced following figures we by the Expression reported of ATX3-Q24 the second derivativesand of the ATX3-Q55 representative FTIR absorption spectra identified by the multivariate analysis, as described in the To Materials explore inand situMethods section. of if the expression Thetheaverage two ATX3 absorption variantsand second perturbed differently derivative spectra the with the host cells, we standard deviation applied FTIR from the three to microspectroscopy independent experiments E. coli intact are 8shown cells, at 5 and h fromin Figure S2. In Table S1 we reported—for all of the analyzed spectral ranges—the protein induction. As a control, we also analyzed NI cells. To investigate the response of assignment of the the relevant main wavenumbers cell biomolecules, theidentified by PLS-DA. FTIR analysis was performed in different spectral ranges. We analyzed the second derivatives of the FTIR spectra, which allow resolving the 2.3.1. FTIR Study of Protein Secondary Structure Modifications: Analysis of Amide I Band different overlapping components in the complex absorption bands. Moreover, to In Figure evaluate 4, we report the the reproducibility of second derivative the results and to analysis identify of E. most the coli ATX3-Q24 significantand ATX3- spectral Q55 induced variations that cells and of NI cells at 5 h of induction, in the Amide I band, enabled us to discriminate between induced and NI cells, we supportedbetween 1700 − 1 and 1600 cm , arising mainly from the C=O stretching vibration of the peptide bond that the FTIR data with a multivariate analysis approach, namely PLS-DA. Furthermore, to provides information on the secondary structures [40] of the whole cell proteins. investigate if the expression of the two protein variants differently perturbed the host In particular, the NI cell spectra were characterized by two main bands at ~ 1658 cm−1 , cells, the Euclidean distance between the PLS-DA scores−1was calculated between the due to α-helix and random coil structures, and at ~1639 cm , mainly assigned to native β- induced and NI cells for each ATX3 variant. sheets [40]. After induction, in the ATX3-Q24 cell spectrum these components were found For sake of clarity, in the following figures we reported the second derivatives of the to decrease in intensity, and a shoulder appeared at ~1630 cm−1 that can be assigned to representative FTIR absorption spectra identified by the multivariate analysis, as protein aggregates [29,41]. Similar features were observed for the ATX3-Q55 cell spectrum, described in the Materials and Methods section. The average absorption and second where, however, the native β-sheet component displayed a minor decrease compared derivative spectra with the standard deviation from the three independent experiments to ATX3-Q24, and the intermolecular β-sheet shoulder was less intense (Figure 4A). In are shown in Figure S2. In Table S1 we reported—for all of the analyzed spectral ranges— addition, a reduction of the α-helix and random coil absorption was detected. Noteworthy, the assignment of the relevant wavenumbers identified by PLS-DA. besides the component around 1660–1654 cm−1 , assigned to α-helix and random coil structures (Figure 4B), the PLS-DA (Table S1) identified three distinct components of 2.3.1. FTIR Study of Protein Secondary Structure Modifications: Analysis of Amide I the β-sheet structures responsible for the discrimination between the analyzed classes: Band ~1642–1640 cm−1 , due to native β-sheets, ~1630 cm−1 tentatively assigned to native-like In Figure 4, we report the second derivative analysis of E. coli ATX3-Q24 and ATX3- Q55 induced cells and of NI cells at 5 h of induction, in the Amide I band, between 1700 and 1600 cm−1, arising mainly from the C=O stretching vibration of the peptide bond that provides information on the secondary structures [40] of the whole cell proteins. Int. J. Mol. Sci. 2021, 22, 943 7 of 21 β-sheets and β-sheets in protein aggregates, and ~1625 cm−1 due to intermolecular β- sheets [29,40,41]. Therefore, the shoulder at ~1630 cm−1 in the second derivative spectra is likely due to the overlapping absorption of two components, ~1630 cm−1 and ~1625 cm−1 , as suggested by PLS-DA. Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 7 of 21 FigureFigure 4. FTIR analysis of protein secondary structuremodifications modifications investigated ininthethe1700–1600 cm cm spectral range.range. −1 4. FTIR analysis of protein secondary structure investigated 1700–1600 −1 spectral The second derivatives of representative FTIR spectra - identified by PLS-DA—of E. coli cells expressing ATX3-Q24 and The second ATX3-Q55derivatives of representative at 5 (A) and FTIR spectra 8 (D) hours of induction, and of - identified NI cells areby PLS-DA—of reported. In (B) (5E.h)coli andcells (E) expressing ATX3-Q24 (8 h), the loading plots and ATX3-Q55 at 5 (A)the representing and 8 (D) components spectral hours of induction, and for more relevant of NIthe cells are reported. discrimination In (B) between the (5 h) and classes are(E) (8 h), the displayed. loading In (C) (5 h) plots and (F) (8 representing theh),spectral the distribution componentsof distances between Ifor more relevant andtheNIdiscrimination is represented asbetween box plots,the where the horizontal classes line within are displayed. In (C) (5 h) and (F)the (8 box is the h), the median. The distribution box ends show of distances the Ifirst between and(Q1)NI isand third quartiles represented as box(Q3), the where plots, upper whisker is the minimum the horizontal line within the between the absolute maximum and Q3 + 1.5* interquartile range (IQR), and the lower whisker is computed as the box is the median. The box ends show the first (Q1) and third quartiles (Q3), the upper whisker is the minimum between maximum value between the absolute minimum and Q1 − 1.5*IQR. Here, IQR is the interquartile range computed as Q3– the absolute Q1. Valuesmaximum and Q3(outliers) beyond whiskers + 1.5* interquartile rangeThe are shown as dots. (IQR), andare distances thecomputed lower whisker is computed as following: as the the Euclidean maximum distance value between between the thePLS-DA absolute minimum scores and Q1for was computed − each 1.5*IQR. Here, IQR combination is the interquartile of samples within the two range groups computed I and NI, as Q3–Q1.anValues obtaining beyondensemble whiskers of (outliers) distances for areATX3-Q24 shown asand ATX3-Q55 dots. independently. The distances are computed The horizontal bar above as following: the the two boxdistance Euclidean plots shows the the between significance of the difference between the ATX3-Q24 and ATX3-Q55 distance distribution. The significance is computed PLS-DA scores was computed for each combination of samples within the two groups I and NI, obtaining an ensemble of using an unpaired two-sample T-test. **: p-value < 0.01. The reported spectroscopic analyses were performed on spectra distances fromfor ATX3-Q24 three and experiments, independent ATX3-Q55 independently. where 5–7 spectra The horizontal were collected for bareach above the two condition inbox eachplots shows the significance experiment. of the difference between the ATX3-Q24 and ATX3-Q55 distance distribution. The significance is computed using an unpaired two-sample T-test. **: p-value < 0.01. The In particular, thereported spectroscopic NI cell spectra analyses were were characterized byperformed two main bandson spectra fromcm at ~ 1658 three −1, independent experiments, where due5–7to spectra α-helix were and random collected coil forstructures, each conditionand atin~1639 cm , mainly assigned to native β- −1 each experiment. sheets [40]. After induction, in the ATX3-Q24 cell spectrum these components were found to At 8 h of in decrease protein induction intensity, (Figure 4D), and a shoulder the spectral appeared at ~1630features cm−1 thatofcan thebeinduced assigned cells to were protein similar aggregates to those at the [29,41]. early time Similar features As of induction. were observed reported in thefor loading the ATX3-Q55 cell plot of Figure 4E, thespectrum, where, however, PLS-DA identified as relevant the native β-sheet component to the discrimination of thedisplayed same main a minor decreasefound components forcompared to ATX3-Q24, 5 h of induction. All ofand thisthe intermolecular considered, β-sheet the band at shoulder around 1639 was less cm−intense 1 (Figure (Figure 4A,D), of 4A). In addition, a reduction of the α-helix and random coil higher intensity in cells expressing ATX3-Q55, could reflect−1the presence not only of native absorption was detected. Noteworthy, besides the component around 1660–1654 cm , assigned to α-helix and β-sheets, but also of native-like β-sheets of toxic species that, as we will illustrate in the random coil structures (Figure 4B), the PLS-DA (Table S1) identified three distinct following, could be responsible for the more important cell perturbation observed for the components of the β-sheet structures responsible for the discrimination between the expanded analyzedvariant classes:compared ~1642–1640to cmcells −1, due expressing ATX3-Q24. to native β-sheets, ~1630 cm−1 tentatively assigned The Euclidean distances between the to native-like β-sheets and β-sheets in protein aggregates, PLS-DA scores of and I and~1625 NI cellscm−1weredue found to to be intermolecular significantly different β-sheets in the ATX3-Q55 [29,40,41]. andthe Therefore, ATX3-Q24 shoulderexpression at ~1630 cm for both −1 in thethe induction second times. In particular, derivative a higher spectra is likely due todistance was observed the overlapping absorption for ofATX-Q24 compared two components, to ATX- ~1630 Q55 cm(Figure 4C,F) −1 and ~1625 cmin agreement −1, as suggested with the higher ATX3-Q24 expression detected by the by PLS-DA. biochemicalAt 8 hanalyses of proteinreported inductionin (Figure 4D), the spectral the previous paragraph features (Figureof the2). induced cells were similar to those at the early time of induction. As reported in the loading plot of Figure 4E, the PLS-DA identified as relevant to the discrimination of the same main components found for 5 h of induction. All of this considered, the band at around 1639 cm−1 (Figure 4A,D), of higher intensity in cells expressing ATX3-Q55, could reflect the presence not only of native β-sheets, but also of native-like β-sheets of toxic species that, as we will Int. J. Mol. Sci. 2021, 22, 943 8 of 21 Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 8 of 21 illustrate in the following, could be responsible for the more important cell perturbation 2.3.2. FTIR Study of Lipid observed Modifications: for the expanded variantAnalysis ofcells compared to theexpressing 3050–2800 cm−1 Spectral Range ATX3-Q24. The Euclidean After the investigation distances of the between in differences the the PLS-DA scores whole of Iprotein cell and NI cells were foundstructure secondary to be significantly different in the ATX3-Q55 and ATX3-Q24 expression for both the changes upon protein induction, we wondered whether the ATX3 expression could induce induction times. In particular, a higher distance was observed for ATX-Q24 compared to modifications inATX-Q55 the membrane lipids (Figure 4C,F) and, in with in agreement particular, the higher if ATX3-Q24 these changes could expression be different detected by the biochemical in cells expressing the two protein analyses variants. reported in the previous paragraph (Figure 2). To this aim,2.3.2. in Figure 5A, we reported the second derivative spectra of ATX3-Q24 and FTIR Study of Lipid Modifications: Analysis of the 3050–2800 cm−1 Spectral Range ATX3-Q55 cells at 5After h oftheinduction, investigationand ofdifferences of the NI cells,in in thethe whole 3050–2800 cm−1 spectral cell protein secondary structure range, dominated by the absorption changes of CHinduction, upon protein n groups weofwondered lipid hydrocarbon whether the ATX3 chains. In particular, expression could the induce spectra of NI cells aremodifications characterized in the membrane by four main lipids and, in particular, bands: ~2959ifcm −1changes these (CH3 could be asymmetric different−in1 cells expressing the two protein variants. stretching), ~2921 cm (CH2 Figure To this aim, in antisymmetric stretching), ~2874 cm−1 (CH3 symmetric 5A, we reported the second derivative spectra of ATX3-Q24 stretching), and and ~2852 − cm cells1 (CH ATX3-Q55 at25symmetric h of induction,stretching) and of NI cells, [42]. In 3050–2800 in the the induced cell spectra, cm−1 spectral range, dominated particularly in ATX3-Q55, webyobserved the absorption of CHn groups an increase of the of lipid hydrocarbon intensity of the chains. CH2Inbands particular, the spectra of NI cells are characterized by four main bands: ~2959 cm−1 (CH3 at ~2921 cm−1 and 2852 cm −1 compared to NI cells. As displayed asymmetric stretching), ~2921 cm−1 (CH2 antisymmetric stretching), ~2874 cm−1 (CH3 in the loading plot (Figure 5B), the ~2854 cm −1 component was−1identified by the PLS-DA as the most relevant symmetric stretching), and ~2852 cm (CH2 symmetric stretching) [42]. In the induced cell to discriminate spectra, between the different particularly sample in ATX3-Q55, classes.anThis we observed result increase of theindicates intensity ofan theimportant CH2 bands at ~2921 cm−1 and 2852 cm−1 compared to NI cells. As displayed in the loading plot variation of the chemico-physical properties of cell lipids in the ATX3-Q55 cells, where the (Figure 5B), the ~2854 cm−1 component was identified by the PLS-DA as the most relevant CH2 band intensity was significantly greater to discriminate between the different sample compared classes.to ATX3-Q24 This cells result indicates an (Figure important 5A,C). In particular, the lipid of variation content of cells expressing the chemico-physical properties ofthe cell expanded ATX3 variant lipids in the ATX3-Q55 could be cells, where characterized bythelonger CH2 band intensity was chains hydrocarbon significantly thatgreater would compared lead toto aATX3-Q24 variation cells of(Figure membrane 5A,C). In particular, the lipid content of cells expressing the expanded ATX3 variant could fluidity and permeability be characterized[43].byInlonger addition, this result hydrocarbon chainscould that wouldalso reflect lead to the acetylation a variation of of lipid A, a component of cell membrane lipopolysaccharide (LPS), as response membrane fluidity and permeability [43]. In addition, this result could also reflect the to a stress acetylation that is counteracted of lipid A, acell for promoting component survival of [44]. cell membrane lipopolysaccharide (LPS), as response to a stress that is counteracted for promoting cell survival [44]. Figure 5. FTIR analysis of lipid modifications investigated in the 3050–2800 cm−1 CHn stretching range. The second Figureof5. derivatives FTIR analysis representative of lipid FTIR spectra, modifications identified by PLS-DA, ofinvestigated in theATX3-Q24 E. coli cells expressing 3050–2800 cm−1 CH and ATX3-Q55 5 stretching at n range. The second derivatives of representative FTIR spectra, identified by PLS-DA, of E. coli cells (A) and 8 (D) hours of induction, and of NI cells are reported. In (B) (5 h) and (E) (8 h), the loading plots representing the spectral components more relevant for the discrimination between the classes are displayed. In (C) (5 h) and (F) (8 h), the expressing ATX3-Q24 and ATX3-Q55 at 5 (A) and 8 (D) hours of induction, and of NI cells are distribution of distances between I and NI is represented as box plots, as in Figure 4. The reported spectroscopic analyses reported. In (B) (5 h) and (E) (8 h), the loading plots representing the spectral components more relevant for the discrimination between the classes are displayed. In (C) (5 h) and (F) (8 h), the distribution of distances between I and NI is represented as box plots, as in Figure 4. The reported spectroscopic analyses were performed on spectra from three independent experiments, where 5–7 spectra were collected for each condition in each experiment. **: p-value < 0.01. At 8 h of protein induction (Figure 5D–F), a quite similar spectral behavior was de- tected, even if the differences between ATX3-Q55 and NI cells are less important compared to the early time of induction. In this spectral range, the Euclidean distances between the PLS-DA scores of I and NI cells were found to be significantly higher in the ATX3-Q55 expressing cells compared to ATX3-Q24 (Figure 5C,F), indicating more important lipid modifications induced by the expanded variant. At 8 h of protein induction (Figure 5D–F), a quite similar spectral behavior was detected, even if the differences between ATX3-Q55 and NI cells are less important compared to the early time of induction. In this spectral range, the Euclidean distances Int. J. Mol. Sci. 2021, 22, 943 between the PLS-DA scores of I and NI cells were found to be significantly higher in 9 ofthe 21 ATX3-Q55 expressing cells compared to ATX3-Q24 (Figure 5C,F), indicating more important lipid modifications induced by the expanded variant. 2.3.3. FTIR Study of Lipid and Peptidoglycan Modifications: Analysis of the 1500–1200 1500–1200 cm− cm−1 Spectral 1 Spectral Range Range In Figure Figure 6A, 6A, the thesecond secondderivative derivativespectra spectraofofATX3-Q24 ATX3-Q24 and ATX3-Q55 and ATX3-Q55 cells at 5ath5ofh cells induction, −1 of induction, and of NI cells, are reported in the spectral range between 1500 andcm and of NI cells, are reported in the spectral range between 1500 and 1200 1200, mainly due todue cm−1, mainly theto overlapping absorption the overlapping of lipids, absorption peptidoglycans, of lipids, and phosphates, peptidoglycans, thus and phosphates, giving information on possible modifications of the cell envelope properties [45,46]. thus giving information on possible modifications of the cell envelope properties [45,46]. Figure 6. Figure 6. FTIR FTIR analysis analysis ofof the the 1500–1200 1500–1200 cm cm−−11 spectral spectral range. range. The The second second derivatives derivatives ofof representative FTIR spectra, representative FTIR spectra, identified by identified by PLS-DA, PLS-DA, ofof E. E. coli colicells cellsexpressing expressingATX3-Q24 ATX3-Q24and andATX3-Q55 ATX3-Q55atat5 5(A) (A)and and8 (D) hours 8 (D) of of hours induction, and induction, of NI and of cells are reported. In (B) (5 h) and (E) (8 h), the loading plots representing the spectral components more relevant for the NI cells are reported. In (B) (5 h) and (E) (8 h), the loading plots representing the spectral components more relevant for discrimination between the classes are displayed. In (C) (5 h) and (F) (8 h), the distribution of distances between I and NI the discrimination between the classes are displayed. In (C) (5 h) and (F) (8 h), the distribution of distances between I and is represented as box plots, as in Figure 4. The reported spectroscopic analyses were performed on spectra from three NI is represented independent as box plots, experiments, whereas in 5–7Figure 4. were spectra The reported collectedspectroscopic analyses for each condition wereexperiment. in each performed *: onp-value spectra