Fig 1.
Efficiency of light projection is dependent on the LEDs etendue.
A lens (blue) is used to collimate the light from an LED with multiple emitters (green squares). While light from the central emitter (green rays) is projected onto the objective aperture, light from off-center LEDs does not reach the objective aperture (grey rays). In addition, outer beam angles produce a wider beam, which also cannot be directed into the objective aperture (black rays).
Fig 2.
Filterset choice determines LED choice.
The excitation (blue, thin line), emission (red, thin line) and dichroic filter properties (green, thin line) for the Chroma ET-EGFP/mCherry set (Chroma 59022) provided by Chroma [13] are plotted and overlaid with the spectra of the Oslon SSL 80 470nm LED (Oslon 470, blue, thick line), the Epitex SMBB490-1100-02 490 nm LED (Epitex 490, cyan, thick line) and the Luxeon PC amber 595 nm (Luxeon 595 amber, thick line) provided by Thorlabs [14]. The 470 nm LED spectrum fits well with the first excitation of the filter set, while the 490 nm LED light, although more ideal for the excitation of EGFP is largely blocked by the excitation filter. The phosphor-converted 595 LED spectrum is broad and only parts of the emitted spectrum can be used.
Fig 3.
Comparison of Arc lamp and LED illumination intensities.
The output of several LEDs was compared to the output of a 120W Mercury Vapor Short Arc lamp (Excelitas X-cite 120) by using Chroma fluorescent plastic slides as samples (see Materials and Methods for details). a) Comparison of blue LEDs to Arc lamp excitation (Arc). LEDs tested were Lumileds Luxeon Rebel royal blue (Luxeon 450), Lumileds Luxeon Rebel blue (Luxeon 470), Cree XT-E2 blue (Cree 470), Oslon SSL 80 470 nm (Oslon 470) and Epitex SMBB490-1100-02 490 nm (Epitex 490). The LEDs centered around 470 nm performed better as they better matched the filter set used. b) Comparison of red LEDs to Arc lamp excitation (Arc). LEDs tested were Lumileds Luxeon Rebel lime (Luxeon 565), Lumileds Luxeon Rebel PC amber (Luxeon 595), Cree XT-E2 PC amber (Cree 590) and Oslon SSL 80 590 nm (Oslon 590). The best performing LEDs in the amber range were about 6 fold less efficient than an Arc lamp in this spectral region. Shown are the averages and standard deviations of the mean intensities of at least 10 individual images.
Fig 4.
Koehler illumination results in superior illumination evenness.
a) Schematic of critical (top) and Koehler modes. Images were inspired by [15]. In critical illumination, the LED light (green) is collimated using a collimation lens (blue) and projected onto the objective. The objective focuses the light, resulting in an image of the LED at the sample plane. In Koehler illumination mode, the collimated light gets focused onto the back focal plane (BFP) of the objective using a second lens (right most lens, blue). This way, the objective projects a collimated cone of light onto the sample plane, resulting in better illumination flatness. A set of two more lenses can be used to generate conjugate planes that than can be regulated by diaphragms to adjust the brightness, contrast, resolution and depth of field of the image (Aperture diaphragm) and the illuminated field of view (Field diaphragm) (grey box). In critical illumination with infinite conjugate plate, an Aperture diaphragm can be used to regulate brightness, contrast, resolution and depth of field. b) Brightness and illumination flatness in critical and Koehler illumination modes were measured on a custom-built epifluorescence microscope (see Materials and Methods for details). Representative images are shown and intensities are coded as colors as indicated in the legend and can be compared between critical and Koehler modes but not between LEDs as different exposure times were used. Koehler illumination showed a better field flatness and slightly increase in intensity (≈ 1.3–29.9% depending on image position for the 470 nm LED and about 2.8–12.96% for the 595 nm LED (see also S5 Table)). c) Comparison of fluorescence intensity distributions. At least three images were taken and the fluorescent plastic slide was moved between images. Images were analyzed in Fiji [3] by drawing a diagonal line ROI and intensities along that line were measured using the “plot profile” tool (black arrow). Resulting intensity values were imported into Matlab (Mathworks), the mean was calculated for three images each and means were normalized to their individual maxima and plotted as a line plot depicting the degree of illumination differences over the measured diagonal. Critical illumination with directly emitting LEDs (470 nm LED) lead to uneven illumination, while the evenness with the phosphor-converted LED (595 nm LED) was only slightly worse than with Koehler illumination.
Fig 5.
LED and optomechanical assembly.
a) Luxeon LED on a 25 mm PCB with wires soldered onto them next to a LLG holder (Thorlabs AD3LLG), here used as a heatsink. b) Assembly consisting of a LED glued on top of an LLG holder and screwed into a xy slip plate (Thorlabs SPT1). c) Overview over the parts needed to assemble the LED illuminator structure. The orientation of the parts in the photograph roughly reflects the order of assembly. d) One arm of the illuminator attached to the dichroic filter cube. e) Schematic and f) photograph of the final assembly. Note that the coated-surface of dichroic mirror should face and reflect the longer wavelength source.
Fig 6.
Electronic circuitry for LED drivers and microcontroller interfacing.
a) Electronic wiring of the LEDs to drivers and the Arduino Uno microcontroller for LED-modulation without logic inverter and b) with logic inverter showing the circuitry with the additional components partially drawn with Fritzing [17]. Blue lines are signal connections, red lines are positive connections and black lines are ground connections essential to avoid different potentials between components. A 12V power supply (12V DC) of at least 3A is needed. The logic converter consists of 2N3906 PNP transistors and two 4.7 kohm resistors. c) Complete assembly with two LEDs mounted, two Buckpuck drivers with wiring harness and connectors wired to an Arduino Uno. The posts and post holder on the illuminator assembly are optional. The output port of the dichro cube can be connected to a microscope port adapter.
Fig 7.
Dimming circuitry allows dimming through μManager.
If brightness modulation through μManager is needed, an alternative circuit can be used. Here, a digital to analog converter (DAC) is used to generate voltage levels between 0 to 5 V. As the voltage sensing circuitry in the BuckPuck driver is reversed, 0 V corresponds to LED on and 5V to LED off. The dimming response curve is however not linear and can be found in the BuckPuck data sheet [16]. As the DAC board does not allow triggering through μManager, an Arduino is used instead and its signal is used to switch the DAC signal on and off through a simple 2N3904 NPN transistor. A 330 ohm resistor is used to protect the transistors gate (middle pin) and a 3.3 kohm or higher resistor is used as a pull-up resistor making sure that the LED shuts off when the Arduino signal is off. Blue lines show signal connections and black lines ground connections. Red lines show a 5 V line used for the pull-up resistors. A 12V power supply (12V DC) of at least 3A is needed. The schematic was partially made with Fritzing [17].
Fig 8.
Triggering allows high-speed multi-channel live-imaging of axonal transport.
Embryonic superior cervical ganglion (SCG) neurons were cultured in tri-chambers [4] and neurons in the soma-compartment were infected with PRV 137 expressing gM-EGFP (gM, green) and mRFP-VP26 (VP26, red). 10 hours post infection viral particle trafficking was imaged in axons penetrating into the middle and neuron compartment using non-triggered Arc lamp illumination (Arc lamp–no trigger) or triggered LED illumination (LED–trigger) (for details see Materials and Methods), a) Three consecutive stills of a representative track for each mode with roughly the same length (≈14 sec) and average particle speed of ≈1.5μm/sec are shown. The time between consecutive dual-color images is ~106 ms for LED illumination and 625 ms for Arc lamp illumination. b) The same tracks as in a) shown as a Kymograph. As triggered LED illumination allows much higher frame rates, particle positions with LED illumination almost overlap in both channels while they do not for non-triggered Arc lamp illumination. c) Signal-to-noise ratio (SNR) calculated for each imaging condition. Scale bar indicates 5 μm.