concepts. Although there are exceptions,understanding the guidelines will enableusers to cope with the exceptions and willprevent filter selection from becomingconfusing and intimidating. (For a list ofcommon filter terminology, see the boxon page 55.)

Know the light sourceTypically, an epifluorescence micro-

scope will have either a mercury or axenon arc lamp. Usually, the lamp typeand its wattage rating are marked on themicroscope’s lamp housing or its powersupply box.

Microscopes for general fluorescenceimaging typicallycome with a mer-cury arc lamp andare marked with“Hg” or “HBO”and/or wattage

F ilter sets, a basic component of flu-orescence microscopy, are extremelyimportant to the success or failure

of experiments. Often researchers spendmonths choosing the correct microscope,camera and image processing equipment,but give short shrift to choosing the correctfilter set. That decision can compromisethe optical performance of the system.

A filter set has three components: anexciter filter, a dichroic mirror and anemission filter. These parts are located ina filter cube, so they can be moved easilyin and out of a microscope. To choosethe best filter set for any given applica-tion, users need to understand a few basic

SUMMARY

Knowing the basics of how to select a filter set for different fluorescentlabels or multiple labels will help researchers get the best data from fluorescence experiments.

By Jennifer Kramer

The Right Filter SetGets the Most out of a Microscope

ratings of 50 or 100 W. The mercury arclamp’s power output spectrum containsenergy peaks at various wavelengths thatcan be exploited in imaging. These peakscan allow the excitation filter to be of anarrower bandwidth, which will reducephotobleaching at less important wave-lengths. For example, an excitation filterthat passes only a narrow slice of the spec-trum around the 546-nm spike in the mer-cury power spectrum allows researchersto get the maximum signal when usingrhodamine or Cy3 markers.

Microscopes used more for ratio imag-ing or imaging in the far red may have axenon arc lamp. They usually are marked“xenon,” “Xe” and “XBO” and havewattage ratings of 75 or 150 W. Xenon’spower curve is very flat through the visi-ble light regions, but it increases signifi-cantly at 750 nm and continues to bestrong through 1000 nm. The relativeevenness of power throughout the visi-ble spectrum makes this light source at-tractive for ratio imaging.

Some microscopes have both types oflamps; in this case, the users must decidewhich light source they plan to use fortheir applications. Also, if the microscopehas no indication of the light source type,

FILTERS FOR MICROSCOPYBIOPHOTONICS

A bandpass filter (1a) excludes the red signal from mousefibroblasts that have the actin filaments labeled with BodipyFL phallacidin (green) and the nuclei with ethidiumhomodimer 1 (red). The same image seen through a long-passemission filter (1b) shows bleedthrough of the red signal.

1a

1b

Reprinted from the January/February 1999 issue of Biophotonics International © Laurin Publishing Co. Inc.

users should get that information fromthe manufacturer.

For the most part, only confocal mi-croscopes use lasers for illumination. Thetwo most common are the argon-ion andkrypton-argon lasers. Various types of he-lium-neon lasers are also used in imag-ing. Researchers who use a laser shouldknow which line of the laser they intendto employ for excitation. In some cases, alaser line excitation filter helps isolate aparticular laser line. In other cases, for-going the excitation filter is useful, par-ticularly for monochromatic sources, be-cause the standard narrowband filtersused to isolate laser lines often signifi-cantly reduce laser transmission.

Detection methodThe detection method is important be-

cause different detectors are sensitive todifferent wavelength regions. Most charge-coupled device (CCD) cameras, for ex-ample, are sensitive to wavelengths oflight much farther red than the humaneye. As a result, a filter set designed foreye viewing alone may allow infrared sig-nal to bleed through to the CCD camera.With the wrong filter set, an image thatlooks great through the eyepieces maylook saturated to a camera. Usually, thisis more of a problem with multiband fil-ters than with single dye filters, and it canbe solved by adding an infrared-blockingfilter in front of the camera.

Single- or multicolor labelingExcitation and emission filters are des-

ignated bandpass, long-pass or short-pass.Although bandpass and long-pass filtersare common types for fluorescence mi-croscopy, the short-pass filter is not.

Before selecting a filter set, researchersshould know whether they will be label-ing samples with only one fluorescentlytagged target or will have multiple targetsand markers.

A bandpass filter passes wavelengthswithin a certain range and absorbs or re-flects wavelengths outside its range. Thisrange of wavelengths, called the passband,is specified using the full-width half-max-imum designation. Full width half max-imum refers to the points on the cut-onand cutoff edges where the transmissionis one-half of the maximum transmission.The bandpass filter also is defined by the

FILTERS FOR MICROSCOPY

Absorption glass: It behaves as a nor-mal long-pass, bandpass or short-pass filter, except that it absorbslight at wavelengths outside thetransmission region. Absorptionglass filters are very hardy for harshenvironments because they haveno coatings to burn or scratch.However, they can be autofluores-cent, and this can be particularlynoticeable with low-light-levelimaging. Furthermore, these filterstend to have very long cutoff andcut-on profiles, meaning that theytake a long time to move fromtransmission to absorption of light.This can be a disadvantage in imag-ing with dyes with narrow Stokesshifts.

Background: What the area without thesample looks like; for example, thearea in between cells or outside ofthe tissue section.

Bandpass filter: These filters pass agiven set of wavelengths of lightonly, while reflecting or absorbinglight at other wavelengths.

Bleedthrough: This is the term usedwhen, in multiple labeling, morethan one dye is visible through a sin-gle-color filter set. Bleedthrough isalso sometimes used to describe abright background caused by lightleaking from the excitation sourceinto the emission filter.

Center wavelength: This is the center ofthe bandpass region of a filter, butnot necessarily the region with max-imum transmission.

Dichroic mirror: Placed at a 45° anglerelative to the incoming excitationlight, the dichroic mirror performstwo functions: It reflects the excita-tion light to the sample and trans-mits the emission signal to the mi-croscope eyepieces.

Emission (or barrier) filter: This filtergoes between the objective lensand the eyepieces or camera porton the microscope. It screens outexcitation light and any other lightnot related to the sample’s fluores-cence.

Exciter (or excitation) filter: This filtergoes between the light source andthe objective lens in the micro-scope’s illumination. Typically it islocated in a filter wheel, in an exci-tation slider or in the filter cube it-self. The coatings on this filter areoften tough enough to stand up tothe harsh light from the light source.

Filter cube: The filter cube is a filterholder specifically designed for agiven microscope. It holds the ex-citation and emission filters and thedichroic mirror in the correct posi-tions for performing fluorescencemicroscopy.

Full width half maximum: The midpointof transmission as a band of a filterbegins to transmit and as it begins toend transmission. This is sometimesreferred to as the 50 percent trans-mission point, or 50 percent of themaximum transmission. Therefore,if a filter transmits maximally at 80percent, the 50 percent transmissionpoint would occur at 40 percent ofthe absolute transmission.

Interference filter: This type of filtercontains coatings in layers that filterlight by causing reflectance or de-structive interference of specificwavelengths.

Long-pass filter: These filters pass allwavelengths of light from a givenwavelength to the red of that wave-length (or to all longer wavelengthsof the designated point).

Narrowband: A bandpass filter with avery narrow band, typically 1 to 3nm wide. It can be used effectivelyonly with a laser or other high-intensity light source, because thenarrow bandwidth often reducestransmission significantly.

Short-pass filter: These filters pass allwavelengths of light from a givenwavelength to the blue of that wave-length (or to all shorter wavelengthsof the designated point).

Stokes shift: The distance between thepeak absorption and emission of adye, usually in nanometers.

Definitions of filter terminology

FOR MICROSCOPYFILTERS

wavelength that occurs in the center ofthe passband. This wavelength, however,is not necessarily the one of maximumtransmission.

In epifluorescence microscopy, band-pass filters are typically 20 to 35 nm wide,but passbands as wide as 70 nm or as nar-row as 10 nm are also common. Themarkings on bandpass filters typically des-ignate the center wavelength, followed bythe width of the passband — for exam-ple, 530DF30 or 530/30 means that atfull width half maximum, the filter trans-mits wavelengths at 530 ±15 nm, or 515to 545 nm.

A long-pass filter passes all wavelengthsat or longer than a specific wavelength.

For instance, a filter that ismarked 515EFLPor 515LP has halfits maximumtransmission ca-pacity at 515 nm,and all wave-lengths above 515

nm will transmit through the filter. Excitation filters are almost always

bandpass filters. Emission filters can bebandpass or long-pass filters. A long-passfilter set usually contains a bandpass ex-citation filter, a long-pass dichroic mir-ror and a long-pass emission filter. Abandpass filter set contains a bandpassexcitation filter, a long-pass dichroic mir-ror and a bandpass emission filter.

Single-label filtersLabeling only one target or antigen pro-

vides the most flexibility in choosing agood filter set. Either a long-pass or band-pass set will work for these applications.Researchers who are not planning to

expand from single labeling into multiplelabeling may want to purchase a long-pass set because these filters maximizethe amount of signal passing to the de-tector. This reduces the exposure timeneeded by the detector, decreasing thetime the sample is illuminated and, inturn, helping to reduce the amount ofphotobleaching in the sample.

Long-pass filters have some drawbacks,such as bleedthrough. For example, a re-searcher who combines rhodamine la-beling with fluorescein labeling but usesa long-pass fluorescein filter set will likelysee rhodamine signal through those fil-ters. Long-pass filters also tend to have abrighter background than bandpass fil-ters, which have blacker backgrounds.Although some filter sets have a com-pletely black background, it is not un-common for long-pass filters to have agray or slightly colored background. Inthe case of a fluorescein long-pass set, forexample, the background might be palegreen, whereas a rhodamine filter set mayhave a pale orange background.

A single exposure using a multiband filter (2a) of the fluorescentlabels DAPI (blue), Bodipy (green) and MitoTracker (red) has poorercolor balance than an overlay (2b) of three images (2c, d and e)taken with individual filter sets.

2a 2d

2e2b

2c

Researchers who plan to expand intomultiple labeling, or who want more flex-ibility, should purchase a bandpass filterset. Imaging samples that are double-la-beled with rhodamine and fluoresceinand visualized with a bandpass fluoresceinfilter set should produce a minimum ofrhodamine bleedthrough.

Unfortunately, there is a trade-off inthe amount of signal that reaches the de-tector. The bandpass filter cuts off a pieceof the dye’s emission spectrum, reducing

the amount of signal going to the detec-tor. The result is that exposure times aretypically longer and the risk of photo-bleaching increases.

Multiband filtersLabeling more than one antigen or tar-

get gives rise to more considerations inchoosing the correct filter set. Imagingcan be performed with a series of indi-vidual bandpass filters, for viewing eachcolor sequentially without the signalsfrom other labels. However, a multibandfilter set enables imaging of all colors atthe same time. Also, there is the possi-bility of a hybrid of the two: exciting withindividual excitation filters and imagingthrough a multiband dichroic mirror andemission filter.

The simplest option is to select a seriesof individual bandpass filters to see eachcolor by itself, one at a time. This methodrequires a series of multiple exposures ofthe sample to get all the color in oneimage. It has several advantages: Eachcolor’s exposure time can be adjusted sono one color overwhelms the others; dimlabels are more easily seen; and it can beused with a color detector (film, colorCCD) or with a black-and-white detector(CCD or photomultiplier tube).

The disadvantage to using the individ-ual bandpass filter sets is that only onecolor is visible at a time. To see what allthe colors look like at the same time

necessitates computer work (in the case ofa CCD or photomultiplier tube) or wait-ing for film to come back from the de-veloper. Of course, if the eye is the onlydetector, the colors never will be visible si-multaneously. In addition, sometimes aslight shift in the image is observed as fil-ter sets are changed. A variety of smallproblems can cause this shift in regis-tration, most relating to manufacturer’stolerances when making filter cubes.However, registration can be easily

corrected in many image processing soft-ware packages.

Multiband filter sets, which allow a re-searcher to see two or more colors si-multaneously through the eyepieces ofthe microscope, also have become popu-lar in recent years. There are advantages tousing these sorts of filters. If the eye is theonly detector, this is the only way to seeall of the labels at the same time. It can beconvenient to photograph the sampleswith a multiband filter as well, becausegetting the full image requires only onephotograph.

Multiband filters also have some draw-backs. First, to see all wavelength bandssimultaneously, one must have a colordetector. This usually means that the eye,film or a color CCD camera are the onlydetectors that can be used. Second, at highmagnification using an objective lens witha high numerical aperture, such as a water

or oil immersion lens, all of the labelsmight not be in focus at the same timebecause they might not all be in the sameplane of focus for the lens. This is less ofa problem with low-NA dry lenses (usu-ally 403 or lower in magnification).Third, there is less that can be done toprevent one color from overwhelming theothers. Finally, it is important to remem-ber that with most multiband filter sets,the bands are very narrow, and dim flu-orescence might be impossible to see.

In the search for the best of bothworlds, some researchers have tried usinga multiband dichroic mirror and emis-sion filter with a series of individual ex-citation filters in a filter slider or a filterwheel. This way, a computer can changethe excitation filter while imaging throughthe same dichroic and emission filter.The excitation filters can be of the sin-gle- or a multiband variety. Most imageprocessing programs (shareware and

commercially available) contain a sim-ple program for taking a series of pho-tographs with varying exposure timesusing a filter wheel. This combination offilters is often referred to as a Pinkel set,after its inventor, Dan Pinkel of theUniversity of California at San Francisco.

This method is advantageous in manyways. First, the colors of the specimen arevisible individually or all at once, de-pending on which excitation filter is inplace. Second, it makes possible individ-ual pictures of a sample, allowing for in-dividual exposure times for each color tocreate a more balanced photograph. Third,the computer can be programmed to dothe same series of exposure times at anypoint in the sample, streamlining the datacollection process. Finally, the registra-tion of the individual colors is usuallybetter with a single filter cube.

The biggest disadvantage to this methodis cost. Typically, a researcher will needto purchase a filter wheel and software to

FOR MICROSCOPYFILTERS

(nm)

Mercury Arc

254

200 700600500400300

577546436405365334313

Xenon

200 11001000900800700600500400300

(nm)The xenon arc lamp has a relativelyflat spectrum through the visiblewavelengths that is good for ratioimaging.

The mercury arc lamp has peaks inits spectrum that, with properfiltration, can be used to benefitfluorescence microscopy.

Labophot, Diaphot, Original Holder 18 mm R 18 3 26 mm 18 mm RFXA, Optiphot, Microphot

Modified Holder 20 mm R 18 3 26 mm 22 mm R

Quadfluor Attachment Quadfluor Holder 25 mm R 25.7 3 36 mm 25 mm REclipse Series

IMT-2 IMT-2 Holder 22 mm R 21 3 29 mm 20 mm R

BH-2, BHT, BHSAHBS 3, AHBT 3, BH-2 Holder 18 mm R 18 3 26 mm 18 mm R

AMMT 3

AX, BX, IX series BX Holder 25 mm R 25.7 3 36 mm 25 mm R

Axio Series Axio 3 (3-Position Slider) 25 mm R 25.7 3 36 mm 25 mm R

Axio 2 Cube (On Spinning Turret) 25 mm R 25.7 3 36 mm 25 mm R

Standard, IM-35, 18 mm R 22 mm R 18 mm R3RS Photo III,

Universal

BioMed, DMIL, Diaplan IMT-2 Holder 22 mm R 21329 mm 20 mm RDialux 20, 22,

Diavert, Fluovert, Laborlux, Ploemopak Holder 18 mm R 18 3 26 mm 18 mm RLaborvert, Orthoplan 1, BH-2 Holder 18 mm R 18326 mm 18 mm R

Ortholux 2

DMR DMR Holder 22 mm R 21 3 29 mm 22 mm R

Aristoplan, Orthoplan II 20 mm R 20 3 28 mm 20 mm R

DMIRB DMIRB Holder (XC122) 20 mm R 18 3 26 mm 20 mm R

Leica

control it, as well as several extra excita-tion filters. A less expensive method is toplace the individual excitation filters in aneutral density slider in the microscope,but this cannot be computer-controlled.

Signal-to-noise ratio A final consideration in filter selection

is the signal-to-noise ratio. Some appli-cations, such as a neuron microinjectedwith DiI, a very bright membrane stain,likely will have high signal-to-noise ratio.The sample background will be nearlyblack, and the neuron will glow sobrightly that a researcher might need aneutral density filter to dim the image foreye comfort. In this case, a typical inter-ference filter set will probably work fine.

Other applications have a low signal-to-noise ratio. For instance, a researcherimaging poorly expressed proteins with

green fluorescent protein will want to col-lect every last possible photon. For theseapplications, companies have designed aseries of filters with very black back-grounds, because in many sets, the exci-tation light is blocked to a greater degreethan with a typical interference bandpassset. The bandpasses are typically a bitwider than with other interference filters,to improve transmission. A wider band-pass allows more signal to get throughthe filter to the detector while maintain-ing the blacker background.

These filters also feature very sharp cut-on and cutoff points in transmission. Inthe case of dyes with very small Stokesshifts, this means that the spectra of ex-citation and emission filters can be posi-tioned more closely together, allowingthe filter to capture more of the dye peak.The excitation and emission filters can’t

be positioned too closely together, how-ever, or excitation light will bleed throughto the emission filter, causing a high back-ground.

The nature of fluorescence microscopymeans that researchers need to understandthe fundamentals of filter technology sothat they can optimize their experiments.Proper filter selection can take advantageof dye and illumination selection to im-prove image quality by minimizing straylight while maximizing the transmissionof the desired fluorescence signal. G

Meet the authorJennifer Kramer is manager of microscopy prod-ucts at Omega Optical Inc. in Brattleboro, Vt.,a manufacturer of optical filters for microscopyand other applications. She received a BS inbiology from the University of North Carolinaat Chapel Hill.

FOR MICROSCOPYFILTERS

Manufacturer Microscope Holder Exciter Dichroic Emitter

Nikon

Olympus

Zeiss