Picking Systems You Can Count On

Colony counting and colony picking are widely recognized as boring, repetitive tasks. In 1963 Joshua Lederberg noted, "The most tedious operation in bacterial genetics is counting-colonies."1 So it is not surprising that researchers have a strong desire to automate the process. The first commercially available automated colony-counting system was produced in the 1970's, according to Michael Zervoudis, sales and marketing manager at Gainesville, Va.-based BioLogics. Artek Systems, BioLogics' pred

May 13, 2002
Jeremy Peirce
Colony counting and colony picking are widely recognized as boring, repetitive tasks. In 1963 Joshua Lederberg noted, "The most tedious operation in bacterial genetics is counting-colonies."1 So it is not surprising that researchers have a strong desire to automate the process.

The first commercially available automated colony-counting system was produced in the 1970's, according to Michael Zervoudis, sales and marketing manager at Gainesville, Va.-based BioLogics. Artek Systems, BioLogics' predecessor, developed the Cytotally plaque-counting system under contract with Fisher Scientific around 1969, Zervoudis explains. "About five years later, they discovered that the system would also work extremely well for colony counting." This modified instrument was marketed as the Artek Counter. In the following three decades, automatic colony-counting systems became increasingly popular, powerful, and versatile. Automated colony counter manufacturers now estimate the time saved with an automated instrument versus manual colony counting between 75% and 90%.

What's a Colony Anyway?

Trivial as it seems, creating an automated colony counter requires one to actually define what constitutes a colony. Optically speaking, a colony is just a spot on an image. Bacterial colonies tend to have fairly uniform shapes and sizes, and most available colony-counting hardware is designed to count them. Systems capable of mammalian colony counting also exist, but because mammalian colonies are less round, less regular, and otherwise quite morphologically distinct from bacterial colonies, an algorithm that works for counting one type of colony may not always succeed when counting another.

Automated colony counters include three basic components: a controlled lighting system (illumination), an imaging system, and a mechanism for counting colonies using the resulting image. Each of these components affects the system's accuracy and should be considered when making a purchasing decision. Optional components exist too, such as the BioStack automated plate handler for Karben, Germany-based BioSys' BioCount system.

The first critical consideration is illumination. Lighting should be even at all points on the plate to minimize position effects on colony detection. This is especially important when analyzing spiral plates or plates with zones of inhibition, as the information derived from such experiments depends on colony position.

Antibiotic inhibition zones are circular, colony-free regions surrounding antibiotic disks placed on the plate's surface. Scientists use spiral plates to estimate an organism's concentration using a single plate, instead of creating multiple dilutions. Decreasing amounts of bacterial sample of unknown concentration are plated in a spiral pattern on an agar dish. The region containing colonies spaced sufficiently far apart to be counted is analyzed—and since the volume of sample deposited on that region is known, the number of colonies can be used to estimate the sample concentration.

Transmission illumination uses a light source below the plate and is most useful when the plate is nearly transparent with opaque, relatively high-contrast colonies. Reflective illumination uses a light source above the colonies and is effective with opaque media such as blood agar. Dark field illumination uses light sources below the plate at an angle that would, if the plate and colonies were entirely transparent, miss the system's optics entirely, giving a dark field-of-view. This type of illumination is especially useful for imaging small colonies, because their light scattering is more visible against the dark background than it would be with transmitted or reflected light.

Most systems include all three light sources. The Countermat Flash from IUL Instruments of Cincinnati, Ohio, is an exception. Product manager Miquel Lleonart explains, "We have developed a kind of optical system that is always reflected light but can have two different backgrounds, and it catches pretty much any type of colonies on any type of media."

Most automated colony-counter cameras have lenses with a relatively long, fixed focal length. This arrangement simplifies camera operation, produces relatively distortion-free images, and encourages reproducibility. Images are usually captured with a charge-coupled device (CCD), essentially a light-sensitive semiconductor that generates an output voltage proportional to the amount of light absorbed. In monochrome instruments, this voltage is translated in a straightforward manner to compute a given pixel's gray level, but color devices can also use CCDs. One-chip color CCD cameras rely on a single chip to read red, green, and blue light, while three-chip CCDs dedicate a chip to each channel, producing better images.

The ProtoCOL RGB from Cambridge, UK-based Synbiosis, includes a color camera and software functionality to simultaneously separate up to eight classes of colonies by color.2 The ProtoCOL product line also includes the basic ProtoCOL system and the ProtoCOL XR, a high-resolution configuration that allows colony size differentiation and detection of smaller colonies.

For many applications, monochrome systems can effectively handle color-based separations because color samples can have notably different gray levels. "If you end up with a grayscale difference between colored colonies you have the opportunity to threshold them out," explains Paul S. Pover, director of Suffolk, UK-based Perceptive Instruments. "Dark blue or red colonies appear much blacker than white colonies on a Sorcerer" (Perceptive's monochrome system).

Camera resolution is another important consideration, and several manufacturers have added high-resolution models to their product lines. The High Resolution Colony Counter by Loats Associates of Westminster, Md., for example, can be outfitted with a 2,048 x 2,048-pixel camera. Company literature reports a 95% detection efficiency with colonies as small as 88µm using these high-resolution cameras.

Counting and Image Analysis

The early Artek counters used hardware-based edge detection to find and enumerate colonies. The instrument scanned the plate in horizontal lines, determining for each line the position of every colony crossed by that line. This data was then compared to the lines above and below it to determine whether the object was new or had previously been scanned. Though the imaging hardware in BioLogics' AccuCount product line has changed considerably from those early efforts, the system still relies on edge detection because it "is less subjective than threshold-based counting," according to Zervoudis.

Most other automated colony counting systems examine the plate image and then apply certain predefined or user-specified filters and other image analysis techniques to determine whether a particular object will be counted or excluded. A common example is the specification of a minimum difference, or threshold, in gray level with respect to the background. Depending on the system, researchers can apply other filters, such as object circularity, size, and intensity to differentiate colonies from background bubbles, debris, and irregularities.

Image-based counting software commonly enables manual addition and deletion of colonies, recording of plate images, and collection of colony dimension and colony count data. Many packages can also handle additional applications such as dilutions, antibiotic inhibition zones, and spiral plates.

Documentation Does Matter

Many automated colony counters ease compliance with Good Laboratory Practices (GLP) and similar documentation requirements. GLP is a documentation and performance standard for nonclinical research that is submitted to the Food and Drug Administration (FDA). While most commercial colony counters include provisions for GLP compliance, there is an additional rule, 21 CFR 11, which applies to investigators planning to submit electronic data to the FDA.

This rule addresses the FDA's concerns about electronic records and signatures. Since, as Pover points out, "FDA has said they will fast-track submissions in electronic form," there is considerable interest in this rule. Pover adds that demonstrating the data's security is also important for compliance with the rule. For instance, "when someone saves data [using the Sorcerer], they ... reenter a password to verify their identity." Though not all instruments are 21 CFR 11 compliant, Zervoudis notes, concerned users likely have access to a compliant Laboratory Information Management System (LIMS). Thus, the counter's compliance is not necessarily a critical issue.

Upcoming Releases

Two new entries into the colony-counting market promise to broaden the range of functionality available to users. Synbiosis plans to launch the aCOLyte system, which is a small unit that connects via a USB port to a PC and transmits a color image of the plate to be examined. The aCOLyte system will be sold in two different configurations based on the same imaging hardware. The basic aCOLyte Click'n'count configuration will allow users to identify colonies by clicking on the PC image, acting like a light box but eliminating transcription errors and facilitating image archiving. A second version, known as Supercount, will contain a simplified version of the ProtoCOL automated counting software for users who are "not interested in large numbers of frills," according to Bob Town, general sales manager. Supercount will also be available as a software upgrade.

A new dedicated colony counter for mammalian cells was released at the American Association for Cancer Research 2002 national meeting in April. Oxford, UK-based Oxford Optronix's ProCount prototype was developed in the lab of Boris Vojnovic, Head of the Advanced Technology Development Group at the Gray Cancer Institute in Middlesex, UK. According to Vojnovic, the ProCount breaks with typical colony counters in several ways. First, customized optics are used to ensure that stain retained on walls of the tissue culture flask does not interfere with the field-of-view and that colonies in the corners of the flask can be detected. Second, says Vojnovic, "rather than use conventional lighting, we use a green electroluminescent panel," which provides even illumination and "matches well to [typical] blue- or purple-stained colonies." Finally, the ProCount "seeks ... the geometric center of the colony.... [It] draws a series of radii tangential to the edge of the colony" and analyzes the region at which the radii cross, he explains. This technique makes the ProCount less sensitive to irregular colony shapes and better able to distinguish overlapping irregular colonies, and one study suggests that it will be at least as accurate and consistent as manual cell counts.3

Colony Picking

When research requirements call for industrial-scale bacterial work, scientists can sometimes turn to automated colony pickers. These usually multi-function robots are priced and designed for labs that need high-throughput sample processing. Although their design varies significantly, the basic colony-picking process is the same for each instrument. The plate is first imaged and colonies localized in a manner similar to that described for colony counters. Colonies are then poked by small pins, which are used to inoculate 96- or 384-well plates. Finally, the pins are sterilized, usually by heat or ethanol exposure, and picking begins again.

Many such machines can run unattended for hours, using automated handling and storage of input plates and output trays. Some machines also have optional or standard lid-removal mechanisms, allowing the lids to remain sealed before and after the colony-picking procedure to minimize contamination.

One colony picker is the QPix, a robot manufactured by Genetix in Hampshire, UK. In his position as Library Core Group Assistant Coordinator for the National Institutes of Health Intramural Sequencing Center at Gaithersburg, Md., Jeff McCloskey has worked extensively with this instrument, and estimates that while it would take a skilled person between five and seven minutes to pick colonies and transfer them to a 96-well plate, the machine can do it in about 1.5 minutes with sterilization, plus another three to five minutes for setup time to accomplish the same task. This corresponds reasonably well with the 3,500 colony-per-hour estimated production speed given by Genetix. McCloskey estimates the instrument's inoculation efficiency "is at least 98%," and says that there have been "no problems picking blue/white, big colonies, [or] small colonies."

Cost Considerations

Automated colony pickers are relatively expensive pieces of equipment. McCloskey says the QPix "was an excellent purchase for us." At $108,000 for a basic configuration, however, the successor QPixII model probably warrants a thorough cost/benefit analysis. On the other hand, if a robot can replace the need for at least one technician, it is likely to pay for itself in relatively short order. As an alternative for labs requiring small-scale or occasional colony-picking projects, some companies, including Genetix, offer colony-picking services.

Low-cost alternatives to colony counters also exist. These are generally software packages that scientists couple to inexpensive imaging tools such as flatbed scanners, to make a "roll-your-own" colony-counting system. Several such packages exist, including MACE by Branford, Conn.-based Weiss Associates (www.colonycount.com) and the Gel-Pro Analyzer 4.0 package from Media Cybernetics of Silver Spring, Md. (www.mediacy.com). MACE is designed for colony-counting applications, while Gel-Pro bills itself as a complete imaging solution. Both can receive image input from TWAIN compatible devices, including nearly all commercially available scanners. But, according to Dean Sequera, Media Cybernetics' vice president of marketing and product development, Gel-Pro Analyzer customers can also use video and digital cameras that are typically used in gel documentation and analysis systems to image colony plates. Naturally, the detection limits, accuracy, and utility of these products will vary considerably depending on the device used to capture the initial image.

Jeremy Peirce (jlpeirce@princeton.edu) is a freelance writer in Jersey City, NJ.

1. J. Lederberg, "An instrumentation crisis in biology," in: The Joshua Lederberg Papers, Profiles in Science, National Library of Medicine, profiles.nlm.nih.gov/BB/G/C/V/S/.

2. A. Paladichuk, "Count on ProtoCOL," The Scientist, 14[9]:21, May 1, 2000.

3. P.R. Barber et al., "Automated counting of mammalian cell colonies," Physics in Medicine and Biology, 46:63-76, 2001.

4. J.D. Cortese, "Array of options," The Scientist, 14[11]:26, May 29, 2000.

Supplemental Materials