Technical Comments

Response to Comment on “The Human K-Complex Represents an Isolated Cortical Down-State”

See allHide authors and affiliations

Science  01 Oct 2010:
Vol. 330, Issue 6000, pp. 35
DOI: 10.1126/science.1182204


Our study confirmed the hypothesis of Amzica and Steriade that the human K-complex (KC) shares neural mechanisms with so-called slow oscillation between periods of intense neuronal firing and silence, but found that the KC can occur independently of this oscillatory activity. We agree with Amzica that the KC often has multiple components but contend that the major component is surface-negative and corresponds to the cortical down-state.

Using mainly intracellular recordings in cats, Amzica and colleagues made monumental contributions to understanding the neural basis of sleep rhythms, proposing that the K-complex (KC) shares cellular and network mechanisms with slow oscillations (16). Our work strongly confirms Amzica and Steriade’s proposal (7). However, contrary to the standard KC definition and more than 70 years of research (810), Amzica and Steriade identified the major component of human KC as surface-positive, and therefore corresponding to a cortical up-state (16). Our results differ from those findings in showing that the major component of the human KC is a surface-negative down-state. We also showed that the KC often occurs outside of ongoing oscillations, in isolation, as have authors in previous human (11) and recent animal work (12).

In his critique (13), Amzica emphasizes that the KC includes surface-positive as well as surface-negative components. We agree, and we discussed the microphysiology of the surface-positive components in our paper, which in concert with the main surface-negative component, lasts for over 800 ms [see figures 2 and 3 in (7)]. However, such components were weak and variable in comparison to the surface-negative wave. This is consistent with scalp electroencephalogram (EEG) recordings, where numerous studies have found that the amplitude of the surface-negative component (often termed the N550) is at least 5 to 10 times as large as the preceding or following surface-positive components. Although, for these reasons, we concentrated on the large surface-negative component, we did not claim in that the KC is exclusively composed of a single component.

In his comment, Amzica gives equal weight to both positive and negative KC components. However, in his previous work discussing the human KC, Amzica emphasized a single component (16), comprising “surface-positive K-complexes (representing excitation in large pools of cortical neurons)” (3). Like numerous other investigators of the human KC, however, we found that the major component of the KC is negative at the scalp [for extensive reviews, see (8, 10)].

To resolve this controversy, we removed the polarity markings from Amzica and Steriade’s human scalp EEG traces (16) and showed them to 12 board-certified electroencephalographers who read clinical EEGs on a routine basis and are highly familiar with the KC. None of the electroencephalographers were aware of the origin of the traces or of the controversy surrounding them. They considered the KCs to be typical but, unanimously and with high certainty, judged the polarity of the recordings to be positive up, the opposite of the polarity with which Amzica and Steriade marked them. The consequence of this error was that they identified the predominant component of the KC as a cortical up-state. Our relatively minor contribution is to show that the (surface-negative) main component of the KC possesses the microphysiological characteristics of the cortical down-state (7). The interpretation of large, slow surface negativities during non-rapid-eye-movement (non-REM) sleep as cortical down-states is completely consistent with the pioneering work of Amzica and Steriade.

Amzica also takes issue with our observation that the KC may occur in isolation, outside of an ongoing rhythmic oscillation. In his previous work, he has asserted that KCs are always rhythmic events with a frequency of ~1 Hz (0.6 to 0.9 Hz) (16). First, it is important to note that if KC-like events were to occur in a repeated and rhythmic manner with a frequency of ~1 Hz, they would be considered a slow oscillation, not a KC, despite the fact that KC occurrence can show temporal aggregation. In fact, the example provided in Amzica’s comment (13) has more features of a slow oscillation than a KC precisely because it is so rhythmic. Indeed, 16 quantitative studies have examined the frequency of KC during stage 2 non-REM sleep and have found that an average of 35 s passes between events [see table 1 in (10)]. We were unable to find significant oscillations preceding our KC even after averaging (7). Importantly, proof that KCs can occur without preceding oscillations is that they can be induced by external stimuli with the same morphology and, as we showed, microphysiology as spontaneous KCs (7). Thus, KCs and slow oscillations are closely related at a mechanistic level but differ precisely in their rhythmicity. Researchers agree that rhythmic sequences of up- and down-states occur; they are called the slow oscillation. The question is whether up- or down-states can only occur as part of an ongoing oscillation. We show that down-states sometimes can occur in relative isolation and that they correspond to what electroencephalographers call the major component of the KC. To disprove this possibility, one would have to show that every KC, including stimulus-dependent KCs, is part of an ongoing oscillation.

Amzica further notes that one of the coauthors of our study (P. Halász) previously proposed a functional significance for the KC that is contrary to what we concluded in (7). It is a scientist’s prerogative to change his or her mind and a duty to do so if the data suggests a better hypothesis. Dr. Halász’s reviews have long suggested that the KC plays a compound role in sleep, noting for example that “KC is a complex multifunctional phenomenon of the sleeping brain involved in information processing and defense against the arousal effect of sensory stimuli” (10).

Finally, Amzica challenges our findings because they were recorded in patients with epilepsy who were taking various medications. He does not suggest how this may have altered our results, and indeed because our basic physiological findings are in close correspondence with his, it is difficult to know how he finds them to be pathological. Nonetheless, we again note that electrodes and recording periods contaminated by epileptiform activity were eliminated from consideration. As discussed in (7), each patient showed typical stage 2 sleep during the recording periods used in this study, and the results were highly consistent across patients despite considerable differences in epileptogenic pathology and medications. Although a confound is possible, Amzica’s comment offers no reason to suspect that it occurred, nor any clue as to how it may have been manifested.

We greatly admire the seminal contributions of Amzica and his colleagues to the fundamental neurophysiology of sleep rhythms. The main importance of our paper was to confirm their prediction that the human KC shares mechanisms and microphysiology with the slow oscillation while clarifying that the major component is surface-negative and equivalent to a down-state.


View Abstract

Navigate This Article