18Hyperpolarization-activated cyclic-nucleotide-gated (HCN) channels control electrical 19 rhythmicity and excitability in the heart and brain, but the function of HCN channels at 20 subcellular level in axons remains poorly understood. Here, we show that the action 21 potential conduction velocity in both myelinated and unmyelinated central axons can 22 bidirectionally be modulated by HCN channel blockers, cyclic adenosine 23 monophosphate (cAMP), and neuromodulators. Recordings from mice cerebellar mossy 24 fiber boutons show that HCN channels ensure reliable high-frequency firing and are 25 strongly modulated by cAMP (EC 50 40 µM; estimated endogenous cAMP concentration 26 13 µM). In accord, immunogold-electron microscopy revealed HCN2 as the dominating 27 subunit in cerebellar mossy fibers. Computational modeling indicated that HCN2 28 channels control conduction velocity primarily via altering the resting membrane 29 potential and was associated with significant metabolic costs. These results suggest that 30 the cAMP-HCN pathway provides neuromodulators an opportunity to finely tune 31 energy consumption and temporal delays across axons in the brain. 32 33 for statistical testing). Since some studies implied that ZD7288 might have some 102 unspecific side effects, such as blocking voltage-dependent Na + channels (Chevaleyre 103 and Castillo, 2002;Wu et al., 2012), we recorded Na + currents from 53 cMFBs and 104 found no change in amplitude or kinetics of voltage-dependent Na + currents after 105 ZD7288 application (Supplemental Fig. 1), suggesting that under our conditions and at 106 a concentration of 30 µM, ZD7288 did not affect the Na + currents. Because of the 107 modulation of HCN channels by intracellular cAMP, we measured conduction velocity 108 during application of 8-bromoadenosine 3′,5′-cyclic monophosphate (8-Br-cAMP; 109 500 µM), a membrane-permeable cAMP-analogue. The conduction velocity increased 110 by 5.9 ± 2.8% in cerebellar mossy fibers (n = 17), by 3.7 ± 1.4% in parallel fibers (n = 111 10), and by 4.6 ± 0.6% in optic nerves (n = 5; see Fig. 1 and legend for statistical 112 testing). These results indicate that HCN channels control the conduction velocity both 113 Schweighofer et al., 2004), which decrease the cAMP concentration. Although we used 129 rather high concentrations of the agonists and off-target effects cannot be excluded (e.g., 130 NE activating dopamine receptors; Sánchez-Soto et al., 2016), these data nevertheless 131 indicate that physiological neuromodulators can both increase or decrease action 132 potential conduction velocity, depending on the type of neuromodulator and receptor. 133