lable at ScienceDirect

Tuberculosis 93 (2013) 690e698

Contents lists avai

Tuberculosis

journal homepage: http: / / int l .e lsevierhealth.com/journals / tube

Evolution and role of corded cell aggregation in Mycobacteriumtuberculosis cultures

Neus Caceres a, Cristina Vilaplana a, Clara Prats b, Elena Marzo a, Isaac Llopis b,Joaquim Valls b, Daniel Lopez b, Pere-Joan Cardona a,*

aUnitat de Tuberculosi Experimental (UTE), Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Universitat Autònoma deBarcelona, CIBERES, Crtra. de Can Ruti, Camí de les Escoles s/n, Edifici Escoles, 08916 Badalona, Spainb Escola Superior d’Agricultura de Barcelona, Departament de Física i Enginyeria Nuclear, Universitat Politècnica de Catalunya, C/Esteve Terradas, 8, 08860Castelldefels, Spain

a r t i c l e i n f o

Article history:Received 9 May 2013Received in revised form17 July 2013Accepted 5 August 2013

Keywords:Mycobacterium tuberculosisCordingCorded cell aggregationCulture growth characteristics

* Corresponding author.E-mail address: pjcardona@igtp.cat (P.-J. Cardona)

1472-9792/$ e see front matter � 2013 Published byhttp://dx.doi.org/10.1016/j.tube.2013.08.003

s u m m a r y

The aim of this study was to evaluate the evolution and role of corded cell aggregation in Mycobacteriumtuberculosis cultures according to growth time and conditions. Thus, in standard culture using aerated7H9 Middlebrook broth supplemented with 0.05% Tween 80, a dramatic CFU decrease was observed atthe end of the exponential phase. This phase was followed by a stable stationary phase that led todissociation between the optical density (O.D.) and CFU values, together with the formation of opaquecolonies in solid culture. Further analysis revealed that this was due to cording. Scanning electron mi-croscopy showed that cording led to the formation of very stable coiled structures and corded cell ag-gregations which proved impossible to disrupt by any of the physical means tested. Modulation ofcording with a high but non-toxic concentration of Tween 80 led to a slower growth rate, avoidance of asudden drop-off to the stationary phase, the formation of weaker cording structures and the absence ofopaque colonies, together with a lower survival at later time-points. An innovative automated imageanalysis technique has been devised to characterize the cording process. This analysis has led toimportant practical consequences for the elaboration of M. tuberculosis inocula and suggests theimportance of biofilm formation in survival of the bacilli in the extracellular milieu.

� 2013 Published by Elsevier Ltd.

1. Introduction

Stress, or lack of it, is known to be an important factor during thelife-cycle ofMycobacterium tuberculosis. The isolation of stationary-phase bacilli may therefore be important in experimental modelingin order to mimic natural infection and to elucidate the degree ofvirulence of bacilli under stressful conditions.

Bacteria can be classified into three stages, namely viable, dormantand dead, on the basis of their viability [1]. In M. tuberculosis, thedormant stage is mostly defined as a state of metabolic shut-downthat reflects an absence of activity [1] and has also been termedlatent, stationary, non-replicative persistent, and viable but not cul-turable, depending on the authors [2]. Although the nutrient starva-tion induced by the multiple-stress model results in expressionchanges in more genes than single-stress applications [3], thedormant stage can be reached by either starved or stressed bacilli. In

.

Elsevier Ltd.

this study, we decided to term the bacilli remaining in the metabolicshut-down stage as “stationary bacilli”.

CulturingM. tuberculosisH37Rv Pasteur strain in an aerated 7H9Middlebrook broth leads to a progressive bacillary aggregation/cording phenomenon that generates populations with totallydifferent physical conditions from the very onset of growth.

The present study was an attempt to characterize the cordingdynamics and to evaluate various methods for avoiding it. Ourstudy highlights how cording formation, which subsequently leadsto biofilm construction, appears to be an important process thatfavors the growth and survival ofM. tuberculosis in the extracellularmilieu.

2. Materials and methods

2.1. Bacteria and cultures

A Master Lot was prepared by growing M. tuberculosis H37RvPasteur strain in a 250-mL Pyrex bottles in a shaking incubator at37 �C and 120 rpm in Middlebrook 7H9 broth (Becton Dickinson)

Table 1Characteristics of the protocols used to disrupt the cords present in the H37Rv liquidcultures.

Protocol A Protocol B Protocol C Protocol D Protocol E

1 Vortex withglass beads1 min

Vortex withglass beads1 min

Vortex withglass beads1 min

Vortex withglass beads1 min

Vortex withglass beads1 min

2 Sonication2 min

Sonication2 min

Sonication4 min

Sonication4 min

Sonication4 min

3 Resting1 min

Resting1 min

Resting1 min

Resting1 min

Resting1 min

4 Sonication1 min

Sonication2 min

Sonication2 min

Sonication4 min

Sonication4 min

5 e e e e Centrifugation10 min300 g

N. Caceres et al. / Tuberculosis 93 (2013) 690e698 691

supplemented with 0.2% glycerol, 0.5% albumin-dextrose catalase(Becton Dickinson) and 0.05% Tween 80, the bottle cap being lefthalf open to allow unlimited O2 availability. Bacteria were left togrow to mid-log phase then stored at �70 �C in 3-mL aliquots,which were used when setting up a new culture to be followed-up.These cultures to be followed-up were performed in triplicate(0.05% Tween 80) or duplicate (0.5% Tween 80) using the sameconditions as for the Master Lot (except for the percentage ofTween 80). These cultures were followed-up until up to day 77without adding new medium. The absorbance, the bacillary load,the cords analysis and presence of opaque colonies were deter-mined at different time points (Table 2) The absorbance ofM. tuberculosis 7H9 Middlebrook cultures was determined bymeasuring duplicate samples at 595 nm using a Genesys 20 spec-trophotometer (Thermo Scientific, Madrid, Spain).

The bacillary load of samples from the liquid cultures wasdetermined by culturing samples on Middlebrook 7H11 agar plates(Becton Dickinson, Madrid, Spain) at 37 �C for 21 days. Visiblecolony forming units (CFU) were counted and the bacillary loadexpressed as log CFU/mL.

2.2. Calculation of growth parameters

The maximum growth rate of an in vitro culture is usually iden-tified from the maximum slope achieved by the growth curve duringthe exponential phase in a semi-logarithmic representation. This wasdetermined from a semi-logarithmic representation of the bacillaryload (log CFU/mL) vs. time (days) by the linear regression betweenthe two consecutive time-points that frame the maximum growth(y ¼ ax þ b2, where a is the maximum growth rate parameter).

The lag-phase parameter of an in vitro culture can be determinedmathematically as the intersection between the initial cell concen-tration and the prolongation of the exponential growth-phase line ina semi-logarithmic representation of the growth curve [4]. Thisintersection occurswhen the horizontal line resulting from the initialCFU concentration (y¼ b1) and the line resulting from themaximumgrowth in the exponential phase (y¼ axþ b2) share the same y value,which is the initial CFU concentration. Thus, the lag-phase is deter-mined using the expression b1 ¼ ax þ b2, where b1 is the CFU/mL atthe initial time point, a is the slope (i.e., the maximum growth rate),b2 is the Y-intercept when x¼ 0, and x is the lag-phase. The lag-phaseis (b1 � b2)/a.

2.3. ZiehleNeelsen staining and viability assay of M. tuberculosis

Methanol-fixed slides containing two 20-mL drops of pure 7H9Middlebrook H37Rv culture were stained with phenolated fuchsin,discolored with acidealcohol solution and counterstained withmethylene blue.

Slides containing two 5-mL samples from exponential- andstationary-phase H37Rv cultures were stained, according to themanufacturer’s instructions, using the BacLight� Bacterial ViabilityKit (Invitrogen, Carlsbad, CA) and then observed under a ZeissAxioskop epifluorescence microscope using Axiovision Rel. 4.8software (Carl Zeiss, Madrid, Spain).

2.4. Scanning electron microscopy

Samples from liquid cultures containing 0.05% or 0.5% Tween 80were extracted at the log and stat phase for examination by scan-ning electron microscopy. Briefly, the samples were deposited intriplicate on 0.22-mm Nucleopore track-etch membranes (What-man, United Kingdom) and processed after air drying. They werefixed with osmium and glutaraldehyde vapor impregnation for48 h in a laminar flow hood. The samples were then mounted on

adhesive carbon films and coated with gold before being furtherexamined under a scanning electron microscope at an acceleratingvoltage of 10e15 kV.

2.5. Image analysis of bacillary aggregates

At each time-point (Table 2), two 20-mL drops from the 7H9Middlebrook cultures were fixed on a glass slide and stained usingthe Ziehle9Neelsen procedure. Between 8 and 10 pictures weretaken using an Eclipse 50i microscope (Nikon, Japan) equippedwith a DS-Fi 1 camera (Nikon, Japan) at 100� using the Nis Ele-ments imaging software (Nikon, Japan). Single cells and aggregateswere detected by image analysis using the MATLAB software(MATLAB, vs. 7.9.0.529; The MathWorks�). The original coloredimage was first converted into a gray-scale image, and then athreshold was fixed to convert the image into a black and whiteone. The black isolated regions were defined as “spots”, whichcould be any particle detected (from single bacilli to large cords).After determination of the area of each aggregate, the area distri-bution was calculated automatically for each time-point consid-ering all pictures taken, in other words the frequency of spots in 15area intervals, logarithmically distributed between 1 and 106 mm2.

2.6. Methods to avoid/disaggregate the cording

The different protocols applied are detailed in Table 1. A Ban-delin Sonorex sonicator (Bandelin Electronic, Germany) was usedfor the sonication and sonication þ centrifugation protocols.

2.7. Evaluating the differences between different stress conditions

A protocol involving heat-shock stress was used in order toobtain stressed bacilli, and the samples obtained compared to freshand frozen cultures. A total of 200 mL of M. tuberculosis H37RvPasteur in 7H9 medium frozen at �70 �C was inoculated in a tubecontaining 8-mL of previously heated PBS and the sample main-tained in a water bath at 50 �C for 30 min. Heat-shock stress wasstopped by placing the tube in ice for 1 min.

Heat-stressed and non-heat-stressed samples were inoculatedin tubes containing Middlebrook 7H9 medium (in duplicate) at afinal concentration of approximately 104 CFU/mL in order toanalyze the length of the lag phase. The tubes were incubated in ashaking incubator at 37 �C and 120 rpm and the caps were left halfopen to allow unlimited O2 availability. The viability of the cultureswere analyzed by determining the CFU. RNA expression of theconstitutive genes, those related to bacillary growth and thoserelated to the stress response were analyzed under the followingconditions: exponential phase frozen culture (log frozen),

Table 2Time points analyzed from the H37Rv growth curves obtained in the 7H9 Mid-dlebrook cultures containing 0.05 or 0.5% Tween 80. The presence of data of CFU/absorvance, cording analysis and presence of opaque colonies for each time pointhas been represented as aþ.

Sampling obtention

Timepoint(days)

0.05% Tween 80 0.5% Tween 80

CFU/Abs

Cordanalysis

Presence ofopaquecolonies

CFU/Abs

Cordanalysis

Presence ofopaquecolonies

0 þ þ1 þ þ2 þ þ3 þ5 þ þ6 þ7 þ þ þ þ þ þ8 þ9 þ þ11 þ12 þ13 þ14 þ18 þ þ þ þ þ þ20 þ23 þ28 þ þ þ þ þ þ30 þ35 þ þ þ þ þ þ37 þ þ þ42 þ þ þ þ þ45 þ þ þ þ þ þ48 þ52 þ þ þ þ þ þ56 þ þ þ þ þ61 þ64 þ69 þ þ þ72 þ þ þ þ þ þ77 þ þ þ

N. Caceres et al. / Tuberculosis 93 (2013) 690e698692

exponential phase growing culture (log fresh), stationary phasefrozen culture (stat frozen) and heat-shock stress applied to a freshexponentially growing culture (log fresh þ heat-shock). A 1.5-mLsample of 7H9 Middlebrook culture was snap-frozen in liquid ni-trogen and stored at �70 �C until required.

RNA extraction and cDNA synthesis. RNA extraction was per-formed using TRIzol (Gibco BRL, Gran Island, NY, USA) in combi-nation with FastPrep products (Qbiogene Inc., Illkirch, France).Briefly, Lysing Matrix B tubes (Qbiogene Inc.) were filled with anappropriate volume of TRIzol and sample, then homogenized in aFastPrep cell fragmenter for two cycles of 45 and 20 s each atmaximum speed, chilling on ice between cycles. RNA extractionwas then performed following the manufacturer’s recommenda-tions. The total RNA concentration was determined by spectro-photometry. Total RNA was subjected to a DNAse treatment withDNA-free kit (Ambion, Woodward Austin, TX, USA). Subsequently,5 mg of the resulting RNA was reverse-transcribed using a Super-script RT kit (Gibco BRL) and Random hexamer (Gibco BRL),following the manufacturer’s recommendations, to obtain cDNA.

mRNA quantification. Real-time PCR was carried out in glass cap-illaries to afinalvolumeof 10mL in thepresenceof 1mL of 10� reactionbuffer (Taq polymerase, dNTPs, MgCl2, SYBRGreen, Roche Bio-chemicals) and 1 mL of cDNA (or water as negative control, whichwasalways included). MgCl2 (to a final concentration of 2e5 mM) andprimers (to a final concentration of 0.5 mM)were also added. A singlepeak was obtained for each PCR product by melting curve analysis.

Primer design and normalization to a housekeeping gene. Allprimers were designed by the STOPLATENT consortium and aredescribed in Supplementary Table 1. 16S DNA was analyzed forevery target sample in order to normalize for DNA synthesis effi-ciencies and RNA input amounts. A ratio was obtained according tothe 16S DNA expression of every sample.

2.8. Statistical analysis

The regression, one-way Anova and Student’s t-test analyseswere performed using the GraphPad Prism Software (v4.03.354; LaJolla, California, USA); differences of p< 0.05 were considered to bestatistically significant.

3. Results

3.1. Characterization of the H37Rv growth curve

The parameters absorbance, CFU visible on solid medium whensub-culturing the liquid culture, proportion of opaque colonies, andproportion of cords and free bacilli in the liquid growth mediumwere analyzed at different time points in order to characterize thegrowth curve (Table 2).

Absorbance and CFU-counting revealed a pre-exponentialperiod (I: lag phase) which lasted 1.3 days, an exponential growthphase (II: log phase) until day 8, with a maximum absorbancegrowth rate of 0.713 per day, followed by a stationary phase (III) inwhich absorbance and CFU detection first diminished drasticallyand then kept a constant bacillary load, characterized by a slightincrease and decrease cycle, until day 77 (Figure 1A, above).

The same absorbance measurement corresponded to a higherCFU count in the logarithmic phase than in the stationary phase. Aregression analysis of the bacillary load and absorbance values wassignificant (r2 ¼ 0.5253, p< 0.0001), showing three areas accordingto the points that defined the regression line: 1) corresponded tothe lag-phase points and included low CFU and absorbance values;2) corresponded to the exponential phase values and included thehighest CFU and absorbance values; and 3) corresponded to thestationary phase values and included intermediate values, although

those detected during the initial stationary phase overlapped theexponential phase area (Figure 1A, below).

Whereas a single type of colony morphology, known as thetranslucent type, appeared throughout phases I and II, another kindof colony morphology, known as the opaque type, was observedthroughout the stationary phase (Figure 2A). Further analysisinvolved determining the proportion of opaque colonies at severaltime-points belonging to the stationary phase (Table 2). The opaquecolonies represented between 5.3% and 36.7% of the total numberof CFUs detected (Figure 2B).

3.2. Effects of high Tween 80 concentration

The Tween 80 concentration in 7H9 Middlebrook medium wasincreased 10-fold, and this change resulted in different growthdynamics, as shown in Figure 1, panel B. Thus, this increase madebacillary growth during the exponential phase more difficult,extending the lag phase (I) of H37Rv (6.3 days) when compared tothe standard culture (1.3 days) and the time needed to reach themaximum CFU peak (12 vs. 8 days). Both the maximum growthrate (0.322 vs. 0.713 per day) and the maximum CFU detected (8.2vs. 9 log CFU/mL) were also lower. Conversely, the maximumabsorbance detected was 1.1 in both cases. However, the greatestdifferences between the two cultures were found for the sta-tionary phase (III) as the sudden decrease in CFU after the end ofthe exponential growth period was not observed. Indeed, thestationary phase for the 0.5% Tween 80 culture was characterizedby several fluctuations in CFU and absorbance measurements,with a global tendency of both values to decrease to the final

Figure 1. M. tuberculosis bacillary load and absorbance values from the 0.05% Tween 80 (A) and 0.5% Tween 80 (B) supplemented Middlebrook 7H9 culture over time. Top: Growthcurve according to CFU (black line) and absorbance (dashed line) showing the different growth phases (divided by vertical dashed lines): lag phase (phase I), exponential phase(phase II) and stationary phase (phase III). Bottom: Linear regression analysis of CFU vs. absorbance values. The corresponding growth phase from each point is indicated by thecolors blue (lag phase), red (log phase) and green (stationary phase). Note that whereas CFU and OD decrease suddenly and then reach a plateau in the 0.05% Tween 80 culture, bothvalues tend to decrease more slowly but continuously in the 0.5% Tween 80 culture. (For interpretation of the references to color in this figure legend, the reader is referred to theweb version of this article.)

N. Caceres et al. / Tuberculosis 93 (2013) 690e698 693

value of 5.5 log CFU/mL at day 72. A regression analysis of thebacillary load and absorbance values proved significant(r2 ¼ 0.4588, p < 0.0001), showing two groups: a) correspondingto the lag phase values, with low CFU and absorbance, and b)values of both exponential and stationary phases, with highvariability. No opaque colonies were observed in none growthphases. However, and despite the differences encountered whencompared with the standard culture, the attempt to avoid cordingby increasing the Tween 80 concentration failed as cords werepresent throughout all growth phases.

Figure 2. Presence of opaque and translucent M. tuberculosis colonies on 7H11 Middlebroasterisk) and translucent (white asterisk) colonies. Panel B: Evolution of CFU (black line) wphase. (For interpretation of the references to color in this figure legend, the reader is refe

3.3. Cording analysis

Several sonication and sonication-plus-centrifugation protocols(described in Table 1) were applied to the H37Rv liquid cultures.These techniques did not affect the bacillary load andwere found topartially disrupt the cords during the exponential phase but not inthe stationary phase (data non shown). Although slightly betterresults were achieved when combining centrifugation with soni-cation, in general, we can consider these attempts to be disap-pointing and were discarded.

ok solid medium during the stationary phase. Panel A: Morphology of opaque (redith regard to the percentage of opaque colonies (gray line) detected in the stationaryrred to the web version of this article.)

Figure 3. Disposition of dead and live bacilli in the cords depending on the growth phase of a 0.05% Tween 80 culture: log (panel A) and stat (panel B). Samples were stained usingthe BacLight� Bacterial Viability Kit, with green-stained particles corresponding to live bacilli and red-stained particles to dead bacilli. The orientation of the long axis of the bacilliis parallel to the long axis of the cord. There is a predominance of red-stained (dead) bacilli on the stationary phase. Scale bars: 100 mm. Original magnification: 600�. (Forinterpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

N. Caceres et al. / Tuberculosis 93 (2013) 690e698694

ZiehleNeelsen-stained samples obtained from the culturesshowed that aggregated bacilli, which resembled cords, were pre-sent throughout all the growth phases. Staining of exponential- andstationary-phase cultures using the green and red fluorescentcompounds contained in the BacLight� Bacterial Viability Kitconfirmed the presence of cords in both growth phases (Figure 3). Apredominance of green-stained (viable) bacilli was observed in theexponential phase (Figure 3, panel A), whereas a predominance ofred (dead) bacilli was observed in the stationary phase (Figure 3,panel B).

Scanning electron microscopy (SEM) analysis showed that theculture produced a corded cell aggregation consisting of aggregatedmycobacteria and an extracellular material with different charac-teristics depending on the growth state (Figure 4A, C, E and G). Inthe logarithmic phase, the low Tween culture presented compactcords and free bacilli, whereas in the stationary phase, the cordswere bigger, more compact and even round. No free bacilli weredetected.

These results suggest that the sudden decrease in bacillary loadand absorbance detected at the beginning of the stationary phase isprobably due to a change in the cording pattern leading to theformation of dense aggregates.

The SEM also showed that the culture with high-percentage ofTween 80 presented aggregation but notwith a cordmorphology inthe logarithmic phase. In the stationary phase, and despite thepresence of cords, the corded cell aggregation observed was moreextended and flat compared to that for the standard culture(Figure 4B, D, F, H).

3.4. Area distribution patterns and dynamics of bacillary spots

Samples of the fresh cultures were obtained at several time-points (Table 2) and stained with ZiehleNeelsen. The images ofthese samples were submitted to an automated spot analysis,recording the area for both single and aggregated bacilli and thearea frequency distribution. The results of the area frequency dis-tribution obtained are detailed in Figure 5. The results showed twocharacteristic patterns: during the exponential phase, the fre-quency distribution presented a log normal distribution in whichthe predominant peaks were those representing the area intervals1e2.5 and 2.5e6.3 mm2 (the first and second bars displayed), whichcomprised single bacilli and aggregates containing a few bacilli. Thesecond pattern appeared gradually from day 18 onwards and wascharacterized by a bimodal distribution containing a second peak,which represented a most frequent area interval of 251e631 mm2

(bar 7) and thus the predominance of large cords. The area intervalsof each bar are detailed in the Figure.

A comparison of the average area distribution for the timepoints included in the stationary phase (from day 18 on) with thatfor the exponential phase distribution clearly shows the above-mentioned bimodal and log normal distributions. Indeed, the re-sults show that the exponential phase presents a majority of singlebacilli and few bacillary aggregates, whereas the stationary phase ischaracterized by a marked increase in the size of the cords.

The second peak in the bimodal distribution for the stationaryphase, which corresponds to the seventh bar and covers the areainterval 251e631 mm2, was analyzed in greater detail to distinguishthe size distribution within this interval. The predominant cordspresent inside this interval measured between 350 and 450 mm2.Cords could sometimes aggregate to form larger round structures(Figure 4, panel E). Such aggregation dynamics might be due to therotation process used to constantly aerate the cultures. Indeed, ifthe rotation is stopped, the cords no longer coil up and tend to builda strong net, or corded cell aggregation.

Despite the differences in phenotypic characteristics betweenthe cultures with high and low percentage of Tween 80, the dis-tribution patterns of the bacillary aggregates were similar in bothcases. Thus, according to the area distribution analysis, 0.5% Tween80 cultures showed the same log normal distribution in the expo-nential phase and bimodal distribution in the stationary phase(Figure 5), thus meaning that the exponential phase presents amajority of single and few aggregated bacilli, and the stationaryphase shows an increase of cords, at both Tween 80 concentrations.

3.5. Differences between samples obtained from fresh, frozen andheat-shock stressed cultures

Whereas the viability from the frozen exponential-phase H37Rvstock was 47.5%, only 15.4% of these bacilli were viable after heat-shock stress. As no differences were found in the statistical anal-ysis between the two conditions for any of the time points tested,the lag phase was considered to be the same for both heat-shockstressed and non-heat-shock stressed bacilli.

RNA analysis showed lower 16S RNA quantification in the logfresh culture than under the other conditions, although this dif-ference was not statistically significant. The mean RNA expressionfor all genes was higher for the log fresh culture than for the othersamples, with this difference being statistically significant for genesAg85 and ESAT-6 (Supplementary Figure 1).

4. Discussion

Bacilli contained in aerosols are supposed to face stressfulconditions in the external milieu such as a decrease in

Figure 4. SEM images of liquid cultures according to their Tween 80 content and the moment the samples were obtained. All pictures on the left side correspond to cultures with0.05% of Tween 80, with those on the right corresponding to cultures with 0.5% Tween 80. Panels A and B correspond to samples obtained at day 7 (log phase), 200�. Panels C and Dare the samples observed with higher magnification (5000�, bars are 2 mm). Panels EeH correspond to samples obtained at day 37 (stat phase), E and F with a 200� magnification,and G, H: 3000�. SEM parameters: 10 kV; spot size: 12; working distance: 5 mm.

N. Caceres et al. / Tuberculosis 93 (2013) 690e698 695

temperature and humidity and the presence of oxygen radicals,ozone and UV radiation, amongst others, prior to entering thehost [5,6]. Once bacillary growth has been stopped in the gran-uloma as a result of the immune response, stressed bacilli surviveinside foamy macrophages, which might allow the bacilli toendogenously re-infect healthy parenchyma by draining themtowards the upper airways [7]. In light of this, the numeroushostile environments encountered by the bacilli during theinfection process may be sufficient to force them to enter thestationary phase in order to survive, thus meaning that the bacilliwhich reach the lungs at the precise moment of infection are in astationary phase (non-replicating, with a high proportion of deadbacilli), something suggested by Garton et al. when demonstrated

the presence of Persister-Like Bacilli in the Sputum of tuberculouspatients [8].

In 1918 Buchanan described the life-cycle of bacteria to involvean initial lag phase, followed by an exponential phase, a stationaryphase, and the death phase, in which the number of viable cellsdecreases [9]. The decrease in bacillary load and absorbance justafter reaching the end of the exponential phase observed in ourstudy was found to occur because of progressive cording andcorded cell aggregation.

Cording was first described by Koch in 1882, who referred tothese structures as “densely bunched and braided groups” [10], andwas one of the first virulence factors described in tuberculosis [11].This process is characterized by the formation of tight bundles of

Figure 5. Characteristic area frequency distribution patterns for cords determined using the measurements obtained during automated image analysis. Each column represents thefrequency of spots for each of the 15 area intervals, logarithmically distributed between 1 and 106 mm2. The graphs represent the results of the cultures containing 0.05% (left) or0.5% of Tween 80 (right). The upper graphs show the area frequency distribution from the exponential phase (day 7) and the lower graphs show the average measurements from thetime points belonging to the stationary phase (from day 18 ahead). Note the more symmetric bimodal distribution shown by the culture containing a low concentration of Tween 80when compared to that with a higher concentration.

N. Caceres et al. / Tuberculosis 93 (2013) 690e698696

bacilli in which the orientation of the long axis of each cell is par-allel to the long axis of the cord [12], and has been related to cell-wall formation, especially mycolic acid cyclopropanation [12e15].

An ability to cord has also been related to biofilm formation, atleast in Mycobacterium marinum [16]. Formation of a biofilm is thetraditional way to obtain BCG [17] in non-shacked cultures. Simi-larly, it has been related to the induction of tolerance or resistanceagainst antibiotics [18,19] and is considered to explain how bacillipersist in necrotic tissue [18]. We have observed in previous studieshow corded cell aggregation formation biofilm-like can follow thecording phenomenon and we consider this aggregation might beessential for the bacilli to grow under extracellular conditions,something only happening once liquefaction takes place duringevolution of the granuloma towards the cavity [20].

Despite the fact that hydrolysis of Tween 80 might release freefatty acid which the bacteria could use, addition of a 10-fold higherTween 80 concentration to the 7H9 Middlebrook culture was usedin an attempt to avoid cording due to its dispersant ability. In-creasing the Tween 80 percentage prolonged the lag phase andslowed down the exponential growth. Moreover, although it wasunable to avoid cording, it showed a different cording pattern thatwas less compact during the exponential phase and a moreextended corded cell aggregation phenotype during the stat phase,thus supporting the dispersant effects of Tween 80 onmycobacteria

[11]. No other Tween 80 concentrations or other detergents wereassayed.

As discussed previously [16], certain growth conditions for theculture used in this study, such as the absence of oleic acid in theADC supplementation and the shaking, may enhance the degreeof cording. The cording ability of mycobacteria has been reportedin numerous studies and has been linked to virulence capacity[21e23], especially in a study from our group concerning thevirulence determination of clinical strains in which automatedimage analysis of cords is related to fitness [24]. The cordingphenomenon has also been observed in non-pathogenic myco-bacteria [12,16], especially in Mycobacterium smegmatis, wherecording was observed in both the exponential and stationaryphases: M. smegmatis bacilli tended to stick together in clumpsthat increased in size and became more compact towards the endof the growth and during the stationary phase [25]. This obser-vation agrees with the evolution of cord dynamics reported in thepresent study, although Smeulders et al. used the term “clump”instead of cord to refer to the aggregated bacilli. Clumps differfrom cords in that no organization can be seen in the bacilli foundin clumps, whereas the cryo-scanning electron micrographs pre-sented in Smeulders’ study showed a parallel disposition of thebacilli to the long axis of the cord. Gonzalez-y-Merchand et al.have recently studied and characterized this phenomenon also in

N. Caceres et al. / Tuberculosis 93 (2013) 690e698 697

M. smegmatis cultures, suggesting the presence of two cell pop-ulations depending on clumping [26]. Cording of the H37Rv strainhas been observed previously in Sauton, Younmans, Dubos and7H9 Middlebrook culture media, all of which contain 0.05%Tween 80 [10].

It is commonly known that mycobacteria in vitro cultures showfluctuations and poor correlation between the CFU counts and ODover growth period, and this is considered to be due to the presenceof cords. However, in our knowledge, we are the first to quantifythis process in M. tuberculosis. The repeated area frequency distri-bution pattern of the stationary phase, which consists of twomaximum peaks in a bimodal distribution, shows that the cordaggregation and disintegration phenomenon does not follow arandom behavior and therefore that there are characteristic sizesfor compact aggregates. Cording dynamics might thus offer anadvantage for survival of the bacilli.

We also consider the results of the oscillating CFU values inboth Tween 80 cultures related to the continuous aggregationand disintegration of bacilli during the stationary phase. Indeed,the fact that the difference in proportion between smaller ag-gregates (left-sided distribution) and larger ones (right-sideddistribution) is higher in the 0.5% Tween 80 culture than in the0.05% Tween 80 culture supports this hypothesis as a high Tween80 concentration should have a greater disintegrating effect.However, further work would be required to confirm this hy-pothesis statistically.

In conclusion, the observed fluctuation of CFU in the stationaryphase may be due to the formation and disintegration of cords intobigger structures, and the error in the regression line between theCFU and absorbance could be influenced by the distribution patternof the bacilli into cords.

Changes in the cording compaction and pattern of area distri-bution between the exponential and stationary phases could occuras a result of an increase in cell wall hydrophobicity in stationary-phase bacilli [25], as has been reported for other bacteria such asthe S14 strain of Vibrio sp., which form cellular aggregates orclumps after prolonged starvation [27].

Similarly, cording may offer survival advantages to the bacillias the predominance of dead bacilli found in the stationary phasecould offer the surrounding viable bacilli a source of useful nu-trients. This hypothesis has previously been made by Smeulderset al., who suggested that the stationary bacilli contained in cordscould use the cellular components released by dying neighbors[25]. The fact that the speed of growth (according to the timeneeded to reach the maximum CFU peak and the maximumgrowth rate) was slower at high Tween 80 concentration than atlow Tween 80 concentration may support this statement as thedispersing effect of Tween 80 produces more branched cords inthe exponential phase, thus making replication of the bacillimore difficult. A study by Shleeva et al. found that filteringM. tuberculosis through 1.5-mm pores prevented it from formingcolonies on solid agar even though microscopy revealed thepresence of large numbers of bacilli [28], which should favor theviability of bacilli if aggregated. Our results show that the pres-ence of single bacilli is related to the high CFU detected on solidMiddlebrook 7H11, as there seems to be an abundance of singlebacilli in the exponential phase, and few single bacilli when theCFU decreases dramatically to around 5.5 log CFU value, in ZiehleNeelsen-stained droplets from the liquid culture. This observa-tion is in contrast to that reported by Shleeva et al., although thebacilli filtered through the 1.5-mm pores were in the stationaryphase.

No correlation was found between the CFU values in the sta-tionary phase and the proportion of opaque colonies in the 0.05%Tween 80 culture in this study, although further work is required

to confirm this finding. The presence of opaque colonies found onthe stationary phase only may be related to bacillary load or to thearea frequency distribution of cords, as the presence of thesecolonies coincides with the dramatic CFU decrease and emer-gence of the bimodal distribution of cord area frequency in the0.05% Tween 80 culture. Thus, induction of opaque colonies seemsto start after the end of this constant decrease. Although theintrinsic mechanism for induction of opaque colonies has notbeen determined, it seems likely that compact and large cordstructures are responsible for formation of the opaque colonies.Indeed, it can be speculated that the aggregates are based on deadbacilli in which some live bacilli are immersed (according toFigure 3). Once in a solid medium, this might lead to asynchronicgrowth due to the presence of bacilli at different stages in theaggregates, thus producing these opaque colonies. However, noopaque colonies could be seen in any of the growth phases in 0.5%Tween 80 culture, where the CFU fluctuation and bimodal dis-tribution of cord area were also present. This absence of opaquecolonies coincides with flat corded cell aggregation formation. Wepropose that this could be related to similar observations reportedby Matsuyama, who found that Serratia mutants which producedless surfactant built colonies with thick branches (opaque col-onies), whereas those producing the normal amount tended tobuild more homogeneous (translucent) colonies [29]. FollowingMasuyama observations, the detergent properties of Tween 80could explain the absence of opaque colonies in those cultureswhere its content was increased.

Different attempts to avoid clumpling have been previouslypublished [26,30]. In our hands, several protocols were tested toavoid cording, but none of them succeeded. The techniques in-volving sonication and sonication-plus-centrifugation succeeded toa large extent when the procedure was applied to exponential-phase bacilli but failed when applied to stationary-phase bacilli.

We tried to obtain stressed bacilli by applying heat-shockstress. Even if a high mortality was detected, the application ofheat shock did not extend the lag phase when compared to thenon-heat-stressed bacilli. We propose the effect of heat-shockmay be eclipsed by the fact of the samples were previouslyfrozen, especially if considered the results obtained in the RNAanalysis. Significant differences in the RNA expression of thegenes analyzed in this study (structural genes and those impli-cated in replication and the stress response) were only foundbetween fresh log (non-stressed) samples and all the othersamples evaluated. The frozen samples showed similar profile ofRNA expression of the genes as the samples obtained from freshculture in stationary phase and from the heat-shock stressedcultures. Because of these results, we suggest the freezing processmight represent a sufficient stress to obtain a stationary-likephenotype in the bacilli.

According to our results, when the inoculum is prepared from afrozen stock the bacilli might be stationary irrespective of the growthphase in which they were frozen, therefore we can conclude thatexperimentalmodeling including infectionwith frozenM. tuberculosismight well mimic the natural conditions under which the host isinfected. We would also like to stress that this work opens up a newperspective regarding the role of the cording phenomenon and itsdynamic characteristics over time, which might provide a survivaladvantage for the bacilli. The cording dynamics of M. tuberculosisshould therefore be studied in greater depth in order to better un-derstand the survival mechanisms of this microorganism.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.tube.2013.08.003.

N. Caceres et al. / Tuberculosis 93 (2013) 690e698698

Funding: This work was supported by the Spanish Ministry ofScience and Innovation (CGL2010-20160 and BIO2005-07949-C02-02; CIBER Enfermedades Respiratorias, programa CRP-TB), theSpanish Ministry of Health (FIS PI080785; National Plan IþDþI FISCM06/00123) and the European Community’s 7th FrameworkProgramme (FP7/2007-2013: STOPLATENT-TB project under grantagreement 200999). The funders had no involvement in the studydesign, in the collection, analysis and interpretation of data; in thewriting of the manuscript; or in the decision to submit the manu-script for publication.

Competing interests: The authors declare that there is noconflict of interest.

Ethical approval: Not required.

References

[1] Lehtinen J. Improvements in the assessment of bacterial viability and killing.Annales Universitatis Turkuensis AI 372 (ed). Turku: University of Turku;September 2007 posting date.

[2] Gomez JE, McKinney JD. M. tuberculosis persistence, latency, and drug toler-ance. Tuberculosis (Edinb) 2004;84:29e44 doi:S1472979203000866 [pii].

[3] Deb C, Lee CM, Dubey VS, Daniel J, Abomoelak B, Sirakova TD, Pawar S, Rogers L,Kolattukudy PE. A novel in vitro multiple-stress dormancy model for Mycobac-terium tuberculosis generates a lipid-loaded, drug-tolerant, dormant pathogen.PLoS One 2009;4:e6077. http://dx.doi.org/10.1371/journal.pone.0006077.

[4] Prats C, Giro A, Ferrer J, Lopez D, Vives-Rego J. Analysis and IbM simulation of thestages in bacterial lag phase: basis for an updated definition. J Theor Biol2008;252:56e68 doi: S0022-5193(08)00033-7 [pii] 10.1016/j.jtbi.2008.01.019.

[5] Cardona PJ. A dynamic reinfection hypothesis of latent tuberculosis infection.Infection 2009;37:80e6. http://dx.doi.org/10.1007/s15010-008-8087-y.

[6] Cardona PJ. Revisiting the natural history of tuberculosis. The inclusion ofconstant reinfection, host tolerance, and damage-response frameworks leadsto a better understanding of latent infection and its evolution towards activedisease. Arch Immunol Ther Exp 2010;58:7e14. http://dx.doi.org/10.1007/s00005-009-0062-5.

[7] Cáceres N, Tapia G, Ojanguren I, Altare F, Gil O, Pinto S, Vilaplana C, Cardona P-J. Evolution of foamy macrophages in the pulmonary granulomas of experi-mental tuberculosis models. Tuberculosis 2009;89:175e82.

[8] Garton NJ, Waddell SJ, Sherratt AL, Lee SM, Smith RJ, Senner C, Hinds J,Rajakumar K, Adegbola RA, Besra GS, Butcher PD, Barer MR. Cytological andtranscript analyses reveal fat and lazy persister-like bacilli in tuberculous sputum.PLoSMed 2008;5:e75. http://dx.doi.org/10.1371/journal.pmed.0050075.

[9] Buchanan RE. Life phases in a bacterial culture. J Infect Dis 1918;23:17.[10] Gao Q, Kripke K, Arinc Z, Voskuil M, Small P. Comparative expression studies

of a complex phenotype: cord formation in Mycobacterium tuberculosis.Tuberculosis (Edinb) 2004;84:188e96. http://dx.doi.org/10.1016/j.tube.2003.12.013. S1472979204000071 [pii].

[11] Middlebrook G, Dubos RJ, Pierce C. Virulence and morphological character-istics of mammalian tubercle bacilli. J Exp Med 1947;86:175e84.

[12] Julian E, Roldan M, Sanchez-Chardi A, Astola O, Agusti G, Luquin M. Micro-scopic cords, a virulence-related characteristic of Mycobacterium tubercu-losis, are also present in nonpathogenic mycobacteria. J Bacteriol 2010;192:1751e60 doi:JB.01485-09 [pii] 10.1128/JB.01485-09.

[13] Glickman MS, Cox JS, Jacobs Jr WR. A novel mycolic acid cyclopropane syn-thetase is required for cording, persistence, and virulence of Mycobacteriumtuberculosis. Mol Cell 2000;5:717e27 doi: S1097-2765(00)80250-6 [pii].

[14] Bloch H, Sorkin E, Erlenmeyer H. A toxic lipid component of the tuberclebacillus (cord factor). I. Isolation from petroleum ether extracts of youngbacterial cultures. Am Rev Tuberc 1953;67:629e43.

[15] Sartain MJ, Dick DL, Rithner CD, Crick DC, Belisle JT. Lipidomic analyses ofMycobacterium tuberculosis based on accurate mass measurements and thenovel “Mtb LipidDB”. J Lipid Res 2011;52:861e72. http://dx.doi.org/10.1194/jlr.M010363.

[16] Hall-Stoodley L, Brun OS, Polshyna G, Barker LP. Mycobacterium marinumbiofilm formation reveals cording morphology. FEMS Microbiol Lett2006;257:43e9 doi: FML143 [pii] 10.1111/j.1574-6968.2006.00143.x.

[17] Dietrich G, Mollenkopf HJ, Weber H, Knapp B, Diehl KD, Hess J, Blackkolb F,Bröker M, Kaufmann SH, Hundt E. Cultivation of Mycobacterium bovis BCG inbioreactors. J Biotechnol 2002;96:259e70 doi: S0168165602000469 [pii].

[18] Lenaerts AJ, Hoff D, Aly S, Ehlers S, Andries K, Cantarero L, Orme IM,Basaraba RJ. Location of persisting mycobacteria in a Guinea pig model oftuberculosis revealed by r207910. Antimicrob Agents Chemother 2007;51:3338e45.

[19] Ojha AK, Baughn AD, Sambandan D, Hsu T, Trivelli X, Guerardel Y, Alahari A,Kremer L, Jacobs WR, Hatfull GF. Growth of Mycobacterium tuberculosisbiofilms containing free mycolic acids and harbouring drug-tolerant bacteria.Mol Microbiol 2008;69:164e74 doi: MMI6274 [pii] 10.1111/j.1365-2958.2008.06274.x.

[20] Grosset J. Mycobacterium tuberculosis in the extracellular compartment:an underestimated adversary. Antimicrob Agents Chemother 2003;47:833e6.

[21] Perez E, Samper S, Bordas Y, Guilhot C, Gicquel B, Martin C. An essential rolefor phoP in Mycobacterium tuberculosis virulence. Mol Microbiol 2001;41:179e87 doi: mmi2500 [pii].

[22] Byrd TF, Green GM, Fowlston SE, Lyons CR. Differential growth characteristicsand streptomycin susceptibility of virulent and avirulent Mycobacteriumtuberculosis strains in a novel fibroblast-mycobacterium microcolony assay.Infect Immun 1998;66:5132e9.

[23] Bhatt A, Fujiwara N, Bhatt K, Gurcha SS, Kremer L, Chen B, Chan J, Porcelli SA,Kobayashi K, Besra GS, Jacobs Jr WR. Deletion of kasB in Mycobacteriumtuberculosis causes loss of acid-fastness and subclinical latent tuberculosis inimmunocompetent mice. Proc Natl Acad Sci U S A 2007;104:5157e62 doi:0608654104 [pii] 10.1073/pnas.0608654104.

[24] CaceresN, Llopis I,Marzo E, Prats C, VilaplanaC, deViedmaDG, Samper S, LopezD,Cardona PJ. Lowdose aerosolfitness at the innatephase ofmurine infection betterpredicts virulence amongst clinical strains of Mycobacterium tuberculosis. PLoSOne 2012;7:e29010. http://dx.doi.org/10.1371/journal.pone.0029010.

[25] Smeulders MJ, Keer J, Speight RA, Williams HD. Adaptation of Mycobacteriumsmegmatis to stationary phase. J Bacteriol 1999;181:270e83.

[26] Gonzalez YMJA, Zaragoza-Contreras R, Guadarrama-Medina R, Helguera-Repetto AC, Rivera-Gutierrez S, Cerna-Cortes JF, Santos-Argumedo L, Cox RA.Evaluation of the cell growth of mycobacteria using Mycobacterium smeg-matis mc2 155 as a representative species. J Microbiol 2012;50:419e25.http://dx.doi.org/10.1007/s12275-012-1556-0.

[27] Siegele DA, Kolter R. Life after log. J Bacteriol 1992;174:345e8.[28] Shleeva MO, Bagramyan K, Telkov MV, Mukamolova GV, Young M, Kell DB,

Kaprelyants AS. Formation and resuscitation of “non-culturable” cells ofRhodococcus rhodochrous and Mycobacterium tuberculosis in prolongedstationary phase. Microbiology 2002;148:1581e91.

[29] Matsuyama T, Sogawa M, Nakagawa Y. Fractal spreading growth of Serratiamarcescens which produces surface active exolipids. FEMS Microbiol Lett1989;52:243e6.

[30] Anton V, Rouge P, Daffe M. Identification of the sugars involved in myco-bacterial cell aggregation. FEMS Microbiol Lett 1996;144:167e70.