Both subjects reported having no sleep disorders and normal age-appropriate 24-hr sleep/wake patterns, as confirmed by quantitative assessments of circadian rest-activity behavior carried out with wrist actigraphy while they lived at home [14, 15]; these results are consistent with a functionally intact retinohypothalmic tract [1, 16] (Figure 2). A normal circadian phase was further confirmed using urinary 6-sulphatoxymelatonin (aMT6s) rhythms in the male subject [3] (Figure 2; also, Supplemental Data available online).

In experiment 1, conducted with the male subject, we aimed to test the spectral sensitivity of the circadian, neuroendocrine, and neurobehavioral axes (Figure 3 and Figure S1). First, we confirmed that he retained a normal melatonin-suppression response to bright-white light exposure [1] on two separate occasions three years apart (see Supplemental Data). We then conducted a 14 day inpatient study to compare the effects of 6.5 hr exposure to 460 nm and 555 nm monochromatic light on circadian phase resetting, melatonin suppression, and enhancement of arousal [17, 18]. In order to compare the relative contribution of the photosensitive retinal ganglion cells (pRGCs) and classical (rod/cone) photoreceptors, we chose two light sources that would differentially stimulate these systems: a monochromatic “blue” light source with a peak emission (λ_max) at 460 nm and hence close to the λ_max of human pRGCs (~480 nm) [11, 19], and a monochromatic light source with a λ_max at 555 nm corresponding to the peak of human photopic vision. Given that this subject exhibited a 24-hr sleep-wake pattern and an entrained aMT6s rhythm, we predicted that the pRGC/melanopsin-driven system would be intact and that the short-wavelength stimulus would elicit full circadian, neuroendocrine, and neurobehavioral responses, whereas the lack of classical photoreception would preclude any response to mid-wavelength 555 nm light.

In a randomized, single-blind design, we exposed the subject to an equal photon density (2.8 × 10¹³ photons/cm²/s) of 555 nm and 460 nm monochromatic light for 6.5 hr, timed to start 1.25 hr before the prestudy bedtime [17, 18]. The subject was seated 90 min prior to and during light exposure, and for 60 min afterward, and was administered a pupil dilator (1 drop per eye, 0.5% cyclopentolate HCl; Cyclogyl, Alcon Laboratories, Texas) and kept in darkness for 15 min prior to lights on (see Supplemental Data). As hypothesized because of the absence of a functional cone response, ocular exposure to 555 nm light had no effect on plasma melatonin, whereas 460 nm light suppressed melatonin by 57% (Figure 3A). Exposure to 460 nm light also caused a −1.2 hr phase delay in the timing of the circadian melatonin rhythm, whereas 555 nm light caused a minimal phase shift (−0.4 hr). In addition, the blue light preferentially increased alpha activity (8–10 Hz) in the waking electroencephalogram (EEG) recordings, indicating a more alert state [18, 20] (Figure 3B), and appeared to decrease subjective sleepiness and improve auditory performance during the latter half of the light exposure (Figure S1), consistent with the short-wavelength sensitivity for the acute effects of light in sighted subjects under similar conditions [17, 18, 21]. It is interesting to note that the blue light did not cause a suppression of delta and theta activity in the waking EEG, as we have previously observed in sighted subjects [18], and it is tempting to suggest that the lack of rod-cone photoreception in this subject may account for the altered EEG response at those particular frequencies, as we recently speculated [22]. Further data are required, however, to confirm this hypothesis. Nevertheless, the short-wavelength near-maximal sensitivity to light at this photon density for a range of responses indicates that this blind subject has a fully functional non-rod, non-cone photoreceptor system mediating the circadian, neuroendocrine, and neurobehavioral effects of light, presumably via intact melanopsin-containing pRGCs.

In experiment 2, we investigated the spectral sensitivity of pupil construction in the female subject by using analytical-photobiological action-spectroscopy techniques. On the basis of her 24 hr sleep/wake pattern and our previous studies on rodents [9, 23], we reasoned that she should also possess some pupil reactivity to bright light, despite the clinical reports that she was unresponsive to the brief light exposure from either a penlight or indirect ophthalmoscopic examination. Quantitative pupillometry, employing monochromatic light at a broad range of wavelengths and irradiances (10¹¹–10¹⁶ log photons/cm²/s) with an exposure duration of 10 s, showed that the subject possessed a functioning pupillomotor system responsive to bright light. The pupil-constriction response was spectrally tuned, peaking (λ_max) at 476 nm. Irradiance-response curves showed a high statistical fit of their derived half-saturation constants to a vitamin A opsin-pigment nomogram (R² = 0.89, compared to R² = 0.35 for rod and R² < 0.01 for all three cone classes), suggesting that pupil constriction was being driven by a single photopigment (Figure 4). The spectral maxima of 476 nm corresponds well to the action spectra for pRGCs in both human (483 nm) and nonhuman primates (482 nm) [10, 24], but not the λ_max of human rods (~498 nm) or short, medium, and long-wavelength cones (λ_max ~420, 534, and 563 nm, respectively) [25] (Figure 4). When the pupil-action spectrum was corrected for preretinal lens absorption [26], the peak spectral sensitivity shifted slightly from 476 nm to 480 nm. Consistent with the results from experiment 1, these data show that this subject possesses both an intact retinopretectal projection (pupillary constriction) and a retinohypothalamic projection (circadian entrainment), and that these responses to light are driven exclusively by short-wavelength-sensitive pRGCs in subjects lacking rods and cones and do not require input from the photopic system [24]. Notably, the confirmation of a pupil response following longer-duration exposure than typically used in brief penlight examinations questions the relevance of this technique, given that unreactive pupils are considered clinically to be a sine qua non of profound blindness of retinal origin despite earlier evidence for short-wavelength sensitivity in human pupil responses [27, 28].

The recent finding in primates that the pRGCs project to the dorsal lateral geniculate nucleus (dLGN) [10]—the thalamic relay that provides a direct input to the visual cortex—led us to explore the possibility that these photoreceptors might contribute to an individual's ability to detect or even experience some awareness of light. We therefore tested whether the female subject could report whether a given light stimulus was present in the first or second of two temporal intervals in a two-alternative forced-choice paradigm (2AFC). After some initial hesitancy about being asked to report the presence of visual stimuli of which she was nominally unaware, she was able to correctly identify the interval in which a 481 nm test light appeared (p < 0.001) but failed (p > 0.05) to detect light at longer or shorter wavelengths (420, 460, 500, 515, 540, 560, and 580 nm) (Figure 4). These detection probabilities remained unchanged when corrected for multiple testing (Bonferroni). Furthermore, she reported that the presence of the detectable stimuli (481 nm) elicited in her a percept that she described as “brightness.” Although superficially these responses resemble cortical blindsight in that she was able to detect a stimulus with a rate of success above chance [29], these data represent a markedly different phenomenon because subjects with damage to the primary visual cortex (V1) have no conscious perception of the stimulus presented [29].

Could these responses to light have arisen from a small number of surviving rods and/or cones rather than from the pRGCs? Although visually evoked potentials (VEP), electroretinogram (ERG), and ocular coherence tomography (OCT) analysis cannot preclude the persistence of a residual population of rods and/or cones, there was no functional evidence of any significant rod or cone involvement. Both the λ_max of ~480 nm and the correspondence of the action spectrum to a single opsin- and vitamin A-based photopigment template strongly implicate phototransduction by the pRGC subsystem alone.

The question remains, however, which neuronal pathways and brain structures mediate these “nonvisual” effects of light. Neuroanatomical investigations in rodents show that melanopsin-containing ganglion cells project to a range of retinorecipient nuclei, including major projections to (1) the hypothalamic suprachiasmatic nuclei (SCN), the site of endogenous circadian pacemaker; (2) the intergeniculate leaflet of the thalamus, an area that is closely linked to normal circadian function and conveys photic and nonphotic signals to the SCN; (3) the ventrolateral preoptic area, an area that controls the switch between sleep and wake states; (4) the olivary-pretectal nucleus, implicated in the pupillary constriction response; and (5) the superior colliculus, which mediates visual and auditory sensorimotor reponses [30, 31]. As indicated previously, a subset of melanopsin-containing ganglion cells also project to the dLGN [10, 31] and in primates have a peak spectral sensitivity (λ_max) of 482 nm [10], thereby possibly providing the neuroanatomical substrate in support of the identical short-wavelength sensitivity for the visual awareness response observed in the female subject. Moreover, recent imaging studies in humans are beginning to identify brain regions associated with light-induced improvements in performance and cognition [32–34] and show preferential short-wavelength activation of the thalamus and the anterior insula, structures strongly implicated in arousal and memory function [34].