Introduction

This is an interim analysis of the flow cytometry (FCM) time kill curve (TKC) assay development reporting on the 6th attempt at serial quantification of BCG in liquid cultures after introduction of antimicrobials, using FCM and colony counts. This is the first attempt with any sensible results.

The main objectives here were to:

This is to progress the overall aim of developing a meaningful and reproducible FCM based in vitro time kill assay for mycobacteria.

Methods

BCG Culture set up

Two 500ml culture flasks containing 84ml of 7H9 broth and Tween80 0.5% v/v were inoculated with two different stored isolates of the same strain of M. bovis BCG (replicate “A” and “B”) to a starting concentration of ~10^3 CFU/ml. These were sealed and incubated at 35oC with 150rpm agitation.

After 6 days incubation, both replicate A and B were divide into seven 50ml falcon tubes each containing 12ml of the original broth culture. Antimicrobials (Rifampicin, R; or Isoniazid, H) were added to 6/7 falcon tubes to give 3 different final concentrations of each. These concentrations represented multiples of the Minimum Inhibitory Concentration (MIC), previously determined using the microtiter plate method. The concentrations were:

  1. R at 0.1 x MIC
  2. R at 1 x MIC
  3. R at 10 x MIC
  4. H at 0.1 x MIC
  5. H at 1 x MIC
  6. H at 10 x MIC
  7. Nil antimicrobial control

Serial sampling for quantification

Each condition/replicate was sampled at day 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 14 and 17 after initial set up. (These are time points -5, -4, -3, -2, -1, 0, +1, +2, +3, +4, +6, +8, +11 days from addition of antimicrobials.)
Samples for CFU counting were 500uL of neat broth liquid, washed once (spun at 3000G for 2 minutes in micro-centrifuge tube, 400uL supernatant removed, resuspended with 400uL of PBS with 0.5% v/v Tween80).

Samples for FCM were diluted 10-fold in PBS with 0.5% v/v Tween80, or 100-fold if broth was visibly cloudy. An aliquot of each sample for FCM was heat killed by placing in 60oC water bath for 10 minutes. Both live and dead samples for FCM underwent mixing by repeated reverse pipetting, 60 seconds of vigorous bench-top vortexing, and 15 seconds of sonication in a water bath, in order to disperse clumps.

Staining procedures for FCM

The following fluorescent dyes were used for FCM counts.

  • SYBR-Gold (SG) : Nucleic acid stain; >1000-fold green fluorescence enhancement upon binding. Membrane penetration is unclear. Peak excitation 495nm (Accuri C6 Laser 1); Peak emission 537nm (AccuriC6 filter FL1). No previous publications of use in mycobacterial flow cytometry. Has not previously been used as a live-dead probe for any bacteria.
  • Calcein-AM (CA) : Cell permeant non-fluorescent esterase substrate becomes fluorescent green and cell membrane impermeant when metabolised. Green fluorescence therefore suggests (1) cell has esterase activity, and (2) membrane is intact enough to prevent leakage of the metabolite. Peak excitation 495nm (Laser 1); Peak emission 515nm (FL1). Calcein-violet used in Hendon-Dunn 2016 as vitality probe.
  • SYTOX-red (SR) : Membrane impermeant nucleic acid stain, 500-fold red fluorescence enhancement upon binding. Red fluorescence suggests incompetent membrane. Peak excitation 640nm (Laser 2); peak emission 658nm (FL3/FL4). SYTOX-green used in Hendon-Dunn 2016 as dead cell probe.
  • Fluorescene diacetate (FDA) : Cell permeant non-fluorescent non-specific esterase substrate becomes fluorescent green and cell membrane impermeant when metabolised. Green fluorescence therefore suggests (1) cell has esterase activity, and (2) membrane is intact enough to prevent leakage of the metabolite. Peak excitation ~500nm (L1); peak emission ~515nm (FL1). FDA has been used as fluorescent live-dead probe for mycobacteria, particularly in experimental drug sensitivity assays (DeCoster 2005; Reis 2004; Norden 1995; Kirk 1998; Bownds 1996).

For each sample, six wells of a 96-well plate were filled with 200uL of the ‘live’ dilution, and one well with 200uL of the heat killed preparation of the sample. An 8th well was filled with 200uL sterile 7H9 media which had undergone incubation, PBS/Tween80 dilution, pipetting, vortexing, and sonication as per the culture samples, providing ‘negative control’ wells. In addition, for each plate, 7 wells were filled with 100uL of the live preparation and 100uL of the heat killed preparation of the nil antibiotic control culture (providing 50:50 ratio live/dead ‘positive control’ wells).

Working stock concentrations were prepared in DMSO for each dye such that 10uL of working stock in 200uL of sample gave the specified final concentrations. Incubation was at room temperature in the dark for 60 minutes.

Dye / sample permuatations
One well per sample, plus one negative control well, and one positive control well, were each stained with the following dye combinations:

  1. CA
  2. FDA
  3. SR + SG
  4. SR + CA
  5. SR + FDA
  6. CA + FDA
  7. No dye

The well containing 200uL of heat killed preparations were stained with:

  1. SR + SG

The purpose of this SR/SG stained heat killed preparation was to provide a reliable total cell count, because heat killed cells should be highly permeant to SG and SR (maximising sensitivity), and the use of two fluorescent dyes with excellent signal to noise characteristics allows dual thresholding on red and green fluorescence signals (maximising specificity). This novel approach is an extension of ‘expert opinion’ recommendations for FCM of bacteria, and was found to be a necessary because of the difficulty of reliably staining mycobacteria for absolute counts by more conventional methods.

Example plate set up (for one replicate):

Flow cytometry parameters and gating

All FCM was carried out on a BD Accuri C6 with data aquired and analysed on BD CFlow-plus software v1.0.264.15. 96-well plates were aggitated before an at intervals during collection to keep cells in suspension. A minimum of 20uL of each sample was analysed on Medium (35 ??L/min, 16-??m core) or Fast (66 ??L/min, 22-??m core) flow rate and core size. All data aquisation was on logarithmic scale.

Thresholding On the Accuri C6 both the trigger channel and the digital cut off for recording a signal as an event are set by a single threshold value; up to two thresholds can be set on FL signals or light scatter.

Events from the heat-killed SG|SR stained wells and one negative control well were collected with dual fluorescence thresholds of FL1(green) >1000 and FL4(red) >1000. These cut offs were previously derived empirically as maximising counts of heat killed BCG with low coefficient of variation between replicates and minimal ‘false positives’ from SG|SR stained sterile broth. Thresholding was checked manually by real time observation of event collection on histograms and scatter plots to ensure discrete populations of events were not bisected by the threshold values.

Events from live cell preparations (all other dye permutations, ie CA, FDA, SR|SG, SR|CA, SR|FDA, CA|FDA, No dye) were collected twice under two alternative dual thresholding strategies: (1) FSC >6000 plus SSC >2000; and (2) SSC >2000 plus FL1>1000. In addition events from live cell preparations dual-stained with SR (ie SR|SG, SR|CA, SR|FDA) were collected with SSC>2000 plus FL4>1000 thresholds. This was to provide additional replicates for absolute counts inthese wells, and check for systematic differences between the alternate thresholding strategies.

Definition of cell populations and calculation of absolute counts The positive control wells (1:1 ratio mix of live and heat killed preparations of nil antimicrobial control culture) under each staining permutation were used to set gates on SSC versus FL1 or FL4 scatter plots which were then applied to the live preparation wells to define SG+, SR+, FDA+ and CA+ populations. Note that SG-, SR-, FDA- and CA- populations were not assessed because of the difficulty of reliably distinguishing BCG cells from debris or other noise on the basis of light scatter parameters alone.

For the heat killed, SR|SG stained samples, thresholded on FL1 and FL4 as described above, FSC~SSC scatter plots were inspected across all replicates and a gate encompasing a discrete population found on these plots was created. Plots without a discrete population were excluded as technical failures. This straegy defined a region of events which showed bright staining with SR and SG (implying presence of a large conecntrated quantity of nucleic acid), plus a definite SSC and FSC distribution suggestive of a homogenous population. They were therefore defined as BCG cells and used for a total cell counts in each sample.

Next, the SG+, SR+, FDA+, CA+ and total cell count population gates were applied to the events from negative control wells (the broth only preparations). Resulting counts were defined as a false negative count for each staining permutation on each plate.

All the above counts were corrected for dilution factor (10^-1 or 10^-2) and volume processed to give absolute count per ml of origibal culture. \(AbsoluteCount/ml = GateCount * (1000/volume(uL)) / DilutionFactor\). For each staining permutation, sample absolute count per ml was adjusted by subtracting the false positive count per ml.

Finally, the total cell counts derived from the heat killed SR|SG samples were used as a denominator to calculate proportions SG+, SR+, FDA+ and CA+ respectively.

CFU counting

Non-selective 7H10 tri-segmented plates were prepared for inoculation for CFU counting by standard methods. Breifly, after washing, 6 serial 10-fold diultions of samples were prepared and plated at 50uL per segment with disposable loop spreading. The plates were incubated lid side down at 37oC/5%CO2 for 21 days. Colonies were counted and their location marked with permanent marker on the outside of the plate at 14 days, and rechecked for further colony appearance at 21 days. Number of colonies was adjusted for dilution factor and inoculum volume to give absolute counts per ml.

Because of the resource implications, CFU counting was not carried out for all samples at all time points, and plates were not made in replicate.

Results

General points / unexpected limitations

  • Samples containing 1x MIC concentration of R or H became cloudy with growth at roughly the same time as antimicrobial negative controls and 0.1x MIC samples. This was obviously unexpected and is presumably due to either (1) technical error in earlier microtite dilution experiements used to measure MIC, (2) technical error when preparing R and H working stocks, or (3) a difference in MIC under different conditions (the microtite dilution plate SOP and the larger broth cultures used in this experiment) - e.g. an innoculum effect.
  • SG, FDA and CA staining was successful, but SR staining did not separate events into distinct + and - populations as anticipated from previous experiments. This was true under all antimicrobial conditions and became progressively worse over time points. In the last 2 time points live and heat killed controls were not clearly separated by red fluorescence after SR staining. One possibility is that handling of the dye stock was not optimal causing loss of fluorescence over time. SR+ populations were therefore not analysed.
  • Because of the adjustment by subtracting negative control from all sample counts in a few cases there were ‘negative’ counts - these are excluded from the analysis.

Serial absolute counts

In this section, parallel growth curves for CFU count per ml, CA+ event counts per ml, FDA event counts per ml, SG event counts per ml, and total cell counts per ml as defined by the heat killed SG+/SR+ population are plotted. (abbreviated to CFU, CA, FDA, SG and HK_SGSRcount in the plot legends).

The BCG growth curve in the absence of antibiotics can be observed for both replicate A and replicate B by combining day 1 to 6 results (from the large culture flasks) and the day 7 to 17 results from the nil antimicrobial control samples.

Throughout exponential growth, CFU count, CA+ count, FDA+count, and total FCM (HK_SGSR) count are seen to closely approximate each other, suggesting they are measuring the same cells, and these cells are metabolically active and capable of colony formation. SG+ counts also rise exponentially, but with lower numbers at all time points and perhaps a slower rate of growth (ie a lower y intercept and slope). Overall, these results are in keeping with SG only staining a minority population with membrane dysfunction (ie SG is likely a cell impermeant stain for BCG); this membrane incompetent sub-population seems to be produced continually during exponential growth so could represent membrane dysfunction in immature cells, or a constant injury/death rate even during exponential growth. Unfortunately plateau growth has not been observed.

Similar results are seen for both R and H at 0.1 and 1 x MIC after antmicrobials were added on day 6:

R at 10xMIC and H at 10xMIC were the only conditions in which there was a fall in CFU; from these samples we can therefore look at how the different dyes ‘behaved’ under confirmed antimicrobial killing conditions. In these time-kill curves, the relationships between dyes are very different to no antibiotic control samples, and there are also differences between the antimicrobials.

Compared to controls, R10 and H10 show a relative rise in SG+ events, with SG+ cells becoming the dominant population over time exposed to antimicrobials. Total cell population (defined by the heat killed SG+SR+ count) stops rising but never declines, implying that antibiotic killed cells do not disintegrate over this time period. Finally, the tight relationship between CFU, CA+, FDA+, and total cell counts seen in control growth curves, is seen to diverge under antimicrobial time-kill.

Further contrasts are observed by comparing H10 to R10. Under H10, the FDA+ count falls early and rapidly, shadowing the same fall in CFU, but diverging from the CA+ count. Under R10, CA+ count falls with a similar slope to H10, but by contrast, FDA+ count does not fall rapidly in R10. Consequently FDA+ counts are actually higher than CA+ counts at later time-points in R10, a reversal of the relationship seen in H10. Oddly, CFU count is fairly flat in R10, which complicates interpretation of contrasts with H10. As mentioned above, SG+ events become dominant under both R10 and H10.

…

Above plots emphasise difference between CFU and FCM count methods. More familiar way of representing time kill curves (emphasising antimicrobial condition) is shown below:

Serial proportions

Here the amount of cells seen to be SG+, FDA+ and CA+ as a are examined as a proportion of the total FCM cell count defined by heat killed samples as discussed above.

Firstly, these serial proportion plots show more within sample/condition variability from time point to time point. I’m not sure if this represents genuine biological variability, or is an artifact of how proportions have been calculated (ie by compounding errors on two counts - this might be testable with more data using Poisson statistics…).

Despite this high variability some patterns are clear. There is an upward trend in proportion CA+ and in proportion FDA+ during early exponential growth (day 1 to 7, NilAbx.CTRL). The lower antimicrobial concentration conditions again look similar to the control samples, but R10 and H10 are very different, both show decline in FDA+ and CA+ proportions, with (roughly reciprocal) replacement by SG+ proportion. The FDA+ proportion falls more rapidly and profoundly in H10 than R10. A striking phenomenon of a peak in CA+ proportion immediately after addition of H at 10xMIC is seen (in fact surpassing 100%… more events were CA+ than there were events in the supposed total count based on heat killed preparation: the total count must be an underestimate). This was also very visible on the original FCM scatter plots: mean fluorescence of FDA+ events rises markedly in H10 samples at 1 day post addition antimicrobial (data not shown).

Discussion

Next steps

I would like to repeat but:

I am slightly worried about slippage though.

At minimum need to validate this system in MTB (eg non-virulent mutant) and check how dyes stand up to fixation procedures before making a plan for doing in vitro time-kill on KDH study MTB stored isolates.

Would like to do fluorescent microscopy on samples in parallel with FCM. This would be important for publications (lab scientists seem to love a picture), might give extra clues by relating staining to morphology, could check clumps are dispersed, and could be an important fall back if cant get flow to work on direct patient samples (ie blood).

Regarding collaboration with Warner group, need to think through how these findings can be adapted into a MIC/MBC screening tool for their novel compound testing on M smegmatis, preferably at a songle time point. I suspect a CA+count/SG+count ratio, along with an absolute total count, might work well, but need to do the full time kill curves for Smegmatis, cf. different mesures at different time points against gold standard to see.