Quantifying Microorganisms at Low Concentrations Using Digital Holographic Microscopy (DHM)

Podsumowanie

Digital holographic microscopy (DHM) is a volumetric technique that allows imaging samples 50-100X thicker than brightfield microscopy at comparable resolution, with focusing performed post-processing. Here DHM is used for identifying, counting, and tracking microorganisms at very low densities and compared with optical density measurements, plate count, and direct count.

Streszczenie

Accurately detecting and counting sparse bacterial samples has many applications in the food, beverage, and pharmaceutical processing industries, in medical diagnostics, and for life detection by robotic missions to other planets and moons of the solar system. Currently, sparse bacterial samples are counted by culture plating or epifluorescence microscopy. Culture plates require long incubation times (days to weeks), and epifluorescence microscopy requires extensive staining and concentration of the sample. Here, we demonstrate how to use off-axis digital holographic microscopy (DHM) to enumerate bacteria in very dilute cultures (100-10⁴ cells/mL). First, the construction of the custom DHM is discussed, along with detailed instructions on building a low-cost instrument. The principles of holography are discussed, and a statistical model is used to estimate how long videos should be to detect cells, based on the optical performance characteristics of the instrument and the concentration of the bacterial solution (Table 2). Video detection of cells at 10⁵, 10⁴, 10³, and 100 cells/mL is demonstrated in real time using un-reconstructed holograms. Reconstruction of amplitude and phase images is demonstrated using an open-source software package.

Wprowadzenie

Determination of accurate bacterial counts in very dilute samples is crucial in many applications: a few examples are water and food quality analysis^1,2,3; detection of pathogens in blood, cerebrospinal fluid, or sputum^4,5; production of pharmaceutical products, including sterile water⁶; and environmental community analysis in oligotrophic environments such as the open ocean and sediments^7,8,9. There is also increasing interest in detection of possible extant microbial life on the icy moons of Jupiter and Saturn, particularly Europa^10,11 and Enceladus^12,13,14, which are known to have subsurface liquid oceans. Because no mission since Viking in 1978 has attempted to find extant life on another planet, there has been limited development of technologies and instruments for bacterial identification and counting during space missions¹⁵.

Traditional methods of plate count find only culturable cells, which can represent a minority of species in environmental strains, sometimes <1%¹⁶. Plates require days or weeks of incubation for maximum success, depending upon the strain. Epifluorescence microscopy has largely replaced plate counts as the gold standard for rapid and accurate microbial enumeration. Nucleic-acid-labeling fluorescent dyes such as 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI), SYBR Green, or acridine orange that bind to nucleic acids are the typical dyes used^17,18,19, though many studies use fluorescent indicators of Gram sign^{20,21,22,23,24}. Using these methods without pre-concentration steps leads to limits of detection (LoDs) of ~10⁵ cells per mL. Improvements in LoD are possible using filtration. A liquid sample is vacuum-filtered onto a membrane, usually polycarbonate and ideally black to reduce background. Low-background dyes such as the DNA stains mentioned above may be applied directly to the filter²⁵. For accurate counting by eye, ~10⁵ cells are required per filter, which means that for samples more dilute than ~10⁵ cells per mL, significant sample volumes must be collected and filtered. Laser-scanning devices have been developed in order to systematically explore all regions of the filter and thus reduce the number of cells required for counting, pushing limits of detection down to ~10² cells per mL²⁶. However, these are not available in most laboratories, and require sophisticated hardware as well as software that permit expert confirmation that observed particles are bacteria and not debris.

For reference, adults with sepsis usually begin showing symptoms at <100 cells/mL of blood, and infants at <10 cells/mL. A blood draw from an adult takes 10 mL, and from an infant, 1 mL. PCR-based methods are inhibited by the presence of human and non-pathogenic flora DNA and by PCR-inhibiting components in the blood^27,28. Despite a variety of emerging techniques, cultures remain the gold standard for the diagnosis of bloodstream infections, especially in more rural areas or developing nations. For detection of life on other planets, thermodynamic calculations can estimate the energy budget for life and thus the expected possible biomass. 1 - 100 cells/mL are expected to be thermodynamically reasonable on Europa²⁹. It can be readily seen from these numbers that detection of very small numbers of cells in large amounts of aqueous solution is an important unsolved problem.

In this paper, we demonstrate detection of Serratia marcescens and Shewanella oneidensis (wild-type and non-motile mutant) at concentrations of 10⁵, 10⁴, 10³, and 100 cells/mL using an off-axis digital holographic microscope (DHM). The key advantage of DHM over traditional light microscopy is the simultaneous imaging of a thick sample volume at high resolution—in this implementation, the sample chamber was 0.8 mm thick. These sample chambers were constructed by the soft-lithography of polydimethylsiloxane (PDMS) from a precision-machined aluminum mold with a tolerance of ± 50 µm. This represents an approximately 100-fold improvement in depth of field over high-power light microscopy. DHM also provides quantitative phase information, allowing for measurements of optical path length (product of refractive index and thickness). DHM and similar techniques have been used for monitoring bacterial and yeast cell cycle and calculation of bacterial dry mass^30,31,32; scattering differences may even be used to differentiate bacterial strains³³.

The instrument we use is custom-built specifically for use with microorganisms, as previously published^34,35, and its design and construction are demonstrated and discussed. Aqueous solutions are continuously supplied to a 0.25 µL volume sample chamber via syringe pump; the flow rate is determined by camera frame rate in order to ensure imaging of the entire sample volume. A statistical calculation predicts the number of sample volumes that must be imaged in order to detect a significant number of cells at a given concentration.

For cell-detection applications, reconstruction of the holograms into amplitude and phase images was not required; analysis was performed on the raw hologram. This saves significant computational resources and disk space: a 500 Mb hologram video will be 1 - 2 Tb when reconstructed. However, we do discuss reconstruction through depth of the sample to confirm that the holograms represent the desired species. An important feature of DHM is its ability to monitor both intensity and phase of the images. Organisms that are nearly transparent in intensity (such as most biological cells) appear clearly in phase. As it is a label-free technique, no dyes are used. This is an advantage for possible space flight applications, since dyes may not survive the conditions of a mission and—more importantly—cannot be assumed to work with extraterrestrial organisms, which may not use DNA or RNA for encoding. It is also an advantage for work in extreme environments such as the Arctic and Antarctic, where dyes may be difficult to bring to the remote location and may degrade upon storage. Reconstruction of images into phase and amplitude is performed using an open-source software package that we have made available on GitHub (SHAMPOO) or using ImageJ.

Protokół

1. Growth and Enumeration of Bacteria

NOTE: This is applicable to almost any bacterial strain grown in the appropriate medium³⁶. In our example, we use three strains: Serratia marcescens as a common, easy identifiable lab strain; and a smaller, highly motile environmental strain, Shewanella oneidensis MR-1. To compare detection of motile vs. non-motile cells, a non-motile Shewanella mutant, Δ FlgM, is also used for comparison³⁷. All strains are grown in lysogeny broth (LB).

1. Prepare sterile LB medium (per liter of distilled water: 10 g bacto tryptone, 5 g bacto yeast, 10 g NaCl; filter or autoclave). Prepare standard 100 mm diameter LB-agar plates (same recipe as medium plus 15 g agar per liter; autoclave (121 °C, 20 min) to sterilize and pour when cool enough to handle). Store medium and plates in refrigerator.
2. The day before the experiment seed 5 - 6 mL of "Master" culture medium from a fresh bacterial colony, using correct sterile and biosafety technique. Incubate on a shaker (120 rpm) at 30 °C for ~12 h. Incubation time will be dependent on the strain and growth rate. In our experiments, the Shewanella strains are incubated for 12 hours; however, Serratia is harvested after 8 hours, to ensure the culture will be exhibiting mid-logarithmic growth.
3. The day of the experiment, take a spectrophotometric reading (OD₆₀₀) of the bacterial Master culture, which is expected to be in the range of 0.6 to 0.7. It should be in the logarithmic growth phase (as determined previously). If not, subculture 100 µL into 5 mL of medium and allow cell growth to the mid-log phase.
4. Take a sample of the Master culture and count the cells directly using a Petroff-Hausser counting chamber. This will enumerate both live and dead cells.
   1. Extract a 10 µL sample of the undiluted culture with a micropipette and transfer into the chamber.
   2. Image under a high-dry objective microscope (40X or 63X, NA 0.7 - 0.8) using phase contrast.
   3. Count the bacteria in at least 20 squares and average.
      NOTE: Count only those bacteria which are entirely within a square and only those crossing over the top and left boundaries (or bottom and right, if you prefer). If squares are separated by several lines, choose one as a boundary.
   4. Calculate concentration as the average of 20 squares x dilution factor. Note that direct counts do not work well for concentrations <10⁷/mL.
5. Make serial dilutions of the culture for counting of colony-forming units (CFU). This will enumerate live cells only.
   1. Make a serial dilution of each of the selected bacterial samples with sterile 0.9% saline solution. Transfer 20 µL of the bacterial solution and dilute it with 180 µL saline. Repeat until the lowest concentration is ~10³ cells/mL.
   2. Take 100 µL from at least two dilutions-suggested are 10³ and 10⁴/mL — and plate on an appropriate solid media plate. Spread with a sterile spreader. Perform at least 3 replicates of each dilution.
   3. Incubate at an appropriate temperature for your bacteria overnight or until colonies grow.
   4. Count colonies and calculate colony forming units (CFU) according to (# of colonies x dilution factor)/volume plated = CFU/mL. Average CFU over the replicates.

2. Preparation of Highly Dilute Samples for DHM

1. Make serial dilutions of the Master culture for DHM and post-DHM counting of colony-forming units on LB media plates (See 1.4.1 - 1.4.4). Perform this double-blind so that the person doing the recordings is unaware of the concentration in each measured sample.
   1. Dilute the bacteria into 10 - 15 mL of a minimal medium that will encourage motility (as appropriate) but inhibit cell division, so that the concentration of cells does not change appreciably during the experiment. This may be 0.9% saline or a more specific motility medium. For example, if using Escherichia coli, motility medium must contain EDTA³⁸. For these examples, we use 0.9% saline.

3. Recording DHM Videos

1. Using a sterile syringe, pull in about 10 mL of the dilution of interest.
2. Connect the syringe to the DHM sample chamber using sterile fittings and tubing.
   NOTE: A custom-made microfluidic chamber was constructed in order to allow the consistent flow of sample through the optical path of the instrument. See the Results section for more details. Prior to the experiment, all components of the sample chamber should be sterilized by autoclaving.
3. Flow the sample from the syringe through the sample chamber continuously using a syringe pump.
   NOTE: Appropriate flow rates vary depending on fluidic channel dimensions, desired throughput as well as data acquisition limitations.
4. As the sample is flowed through the sample chamber, acquire holograms consecutively at an appropriate frame rate. A time-stamp file will be created which logs the time of each image capture to be used later during data analysis (protocol section 4). Obtain all holograms at ambient temperature (23° C). Record holograms by executing a pre-written C++ executable file provided by the camera manufacturer.
   NOTE: Appropriate frame rates vary depending on the flow rate chosen. For this experiment, it is recommended to use a frame rate such that a bacterium would be imaged >2 times as it traveled across the instrument's field of view.
5. Allow sufficient time for the entire 10 mL of sample to be flowed through the DHM.
6. To ensure that bacteria are not growing or dying during the experiments, inoculate enriched media plates with 100 µL of the spent media post-DHM image capture. Spread with a sterile spreader and incubate at appropriate temp for 24 hours; count colonies present.

4. Calculation of Cell Density and Limits of Detection

1. With the acquired hologram videos, analyze the data for presence of bacteria.
   1. Calculate the median pixel value for a time series of holograms then subtract this median value from each respective pixel to eliminate stationary artifacts in the hologram. This may be done in ImageJ by performing the following steps:
      1. Convert to 32-bit (Image-Type-32 bit)
      2. Calculate the median (Image-Stack-Zproject-Median)
      3. Subtract the median from the rest of the stack (Process-ImageCalculator-Subtract)
   2. Count the number of visible Airy rings and in-focus cells manually. In ImageJ, this is done by pointing at the objects with the Point tool.
   3. Each series of holograms will be accompanied by a time-stamp file (recorded at time of hologram acquisition). Use the time-stamp file to calculate the total amount of sample pumped at the time the image was captured.
2. Calculate the cell density by dividing the total number of cells detected by the total volume of sample imaged. Average 5 - 10 frames for accurate statistics.

5. Image Reconstruction to Amplitude and Phase

1. Choose representative videos with 10 - 200 cells per frame. Load raw (not median-subtracted) holograms into the reconstruction software (examples: ImageJ Numerical Propagation³⁹; SHAMPOO [GitHub]).
   1. Using ImageJ, go to OD-Numerical Propagation-Numerical Diffraction.
   2. At the prompt, draw a Fourier mask to choose the real or virtual image and eliminate any sinusoidal noise features.
   3. Reconstruct amplitude and phase at appropriate z-steps by entering the reconstruction distance.
2. Median subtract the amplitude data to reduce noise as in 4.1.1.
3. Plot data as a 3D cube or projection. Go to Plugins—Volume Viewer.

Wyniki

The results should indicate the ability to detect living and dead bacteria at very low levels by DHM. The number of bacteria counted should be consistent with the results obtained using the Petroff-Hauser counting chamber and plate counts. Standard statistical methods provide information about the accuracy of the different detection methods at various bacterial concentrations.

Figure 1 shows the Pet...

Dyskusje

Numerical reconstruction of holograms: For the numerical reconstruction of holograms, the angular spectrum method (ASM) is used. This involves the convolution of the hologram with the Green's Function for the DHM. The complex wavefront of the image at a particular focal plane can be calculated by employing the Fourier Convolution Theorem as follows:

(1)

Where 

Materiały

Name	Company	Catalog Number	Comments
Bacto Yeast	BD Biosciences	212750
Bacto Tryptone	BD Biosciences	211705
Sodium chloride	Sigma-Aldrich	7710	Many options for purchase
Bacto Agar	BD Biosciences	214010
10 cm Petri dishes	VWR	10053-704
15 mL culture tubes	Falcon (Corning Life Sciences)	352002	Loose-capped
Petroff-Hauser chamber	Electron Microscopy Sciences	3920
10 mL syringes	BD Biosciences	309604	Luer-Lok tip not necessary
Male Luer to 1/16” barbed fitting	McMaster-Carr	51525K291
Autoclavable 1/16” ID PVC tubing for flow	Nalgene	8000-0004	Sold by length, purchase accordingly
Syringe pump	Harvard Apparatus	PHD 2000
Sample Chamber	Custom	n/a	See Materials Section
Holographic Microscope	Custom	n/a	See Wallace et al.
Open-source software	Custom	https://github.com/bmorris3/shampoo