5D Library › 下丘脑垂体肾上腺轴耦合振荡、皮质醇觉醒反应扁平化与倦怠恢复动力学 · English

第二章 这些脉冲是为了什么

一个会饱和的函数作用在变动的输入上

如果这条轴只需要每天送出一定量的激素,一股平稳的细流就够了。它并没有用平稳的细流,理由是受体的反应会饱和。一个会饱和的函数作用在变动的输入上,与它作用在平均输入上并不相同,而差的方向是固定的:函数的平均低于平均的函数。所以同样的日总量以脉冲送达时,平均上更少地占用受体,而在峰值处更多地占用它。

Lightman 与 Conway-Campbell 陈述了这样一个论点:这正是脉冲性的意义所在,而不是它的副作用,而一个持续动态平衡的系统需要这些脉冲才能保持反应性 (Lightman and Conway-Campbell 2010)。实验上的对应物是直接的。Stavreva 等人给细胞以脉冲式与恒定式两种方式施加相同总量的激素,发现脉冲式暴露驱动出成串的基因转录脉冲,而恒定暴露并不能复现它们 (Stavreva et al. 2009)。

证据

已发表。 Lightman 与 Conway-Campbell 论证这条轴的脉冲活动对持续的动态平衡是必要的,而不是附带的 (Lightman and Conway-Campbell 2010)。Stavreva 等人证明,超日激素脉冲产生受体介导的成串基因转录,而相同总量的恒定暴露并不产生 (Stavreva et al. 2009)。

推导

# Chapter 2. p03: Why the hormone arrives in pulses rather than steadily.
# A receptor's occupancy is a saturating function of the hormone around
# it, and a saturating function of a fluctuating input is not the same as
# the function of its average. Delivering the same daily amount in tall
# brief pulses therefore produces a different mean occupancy from
# delivering it as a flat infusion, and the difference is not a detail:
# it is the whole reason a pulsatile axis can carry a large signal
# without keeping its own receptors permanently engaged.
KD, HILL = 1.0, 1.0        # occupancy half-point and Hill exponent
MEAN, DT, HOURS = 1.0, 0.002, 24.0
PATTERNS = (
    # label, pulses a day, duty cycle (share of the day with hormone up)
    ("flat infusion",      1, 1.000),
    ("6 broad pulses",     6, 0.500),
    ("12 pulses",         12, 0.250),
    ("16 pulses",         16, 0.167),
    ("24 sharp pulses",   24, 0.100),
    ("36 sharper pulses", 36, 0.060),
)

def occupancy(c):
    return c ** HILL / (KD ** HILL + c ** HILL)

def profile(n_pulses, duty):
    # square pulses of height set so the daily mean is MEAN
    height = MEAN / duty
    period = HOURS / n_pulses
    on = period * duty
    steps = int(round(HOURS / DT))
    out = []
    for i in range(steps):
        t = (i * DT) % period
        out.append(height if t < on else 0.0)
    return out

if __name__ == "__main__":
    print("   same daily mean of {:.1f} in every pattern".format(MEAN))
    print("   pattern             peak    mean occupancy   peak occupancy")
    rows = []
    for label, n, duty in PATTERNS:
        c = profile(n, duty)
        occ = [occupancy(v) for v in c]
        m = sum(occ) / len(occ)
        rows.append((label, max(c), m, max(occ)))
        print("   {:<20}{:<8.2f}{:<17.4f}{:.4f}"
              .format(label, max(c), m, max(occ)))
    flat = rows[0]
    print("   occupancy of a flat infusion at the same mean: {:.4f}"
          .format(flat[2]))
    print("   pattern             occupancy relative to flat")
    for label, peak, m, pk in rows:
        print("   {:<20}{:.4f}".format(label, m / flat[2]))
    best = min(rows[1:], key=lambda r: r[2])
    print("   the sharpest pattern holds occupancy down to {:.4f} of the"
          .format(best[2] / flat[2]))
    print("   flat value while delivering the same amount of hormone")
    print("   and its peak occupancy is {:.4f} against {:.4f} flat,"
          .format(best[3], flat[3]))
    print("   so the signal is larger and the average load lower")
    print("   peak to mean ratio of occupancy, by pattern:")
    for label, peak, m, pk in rows:
        print("   {:<20}{:.3f}".format(label, pk / m))
    print("   that ratio is the contrast a downstream reader can see,")
    print("   and it rises from {:.3f} to {:.3f} across the table"
          .format(rows[0][3] / rows[0][2], rows[-1][3] / rows[-1][2]))
#    same daily mean of 1.0 in every pattern
#    pattern             peak    mean occupancy   peak occupancy
#    flat infusion       1.00    0.5000           0.5000
#    6 broad pulses      2.00    0.3333           0.6667
#    12 pulses           4.00    0.2000           0.8000
#    16 pulses           5.99    0.1440           0.8569
#    24 sharp pulses     10.00   0.0916           0.9091
#    36 sharper pulses   16.67   0.0566           0.9434
#    occupancy of a flat infusion at the same mean: 0.5000
#    pattern             occupancy relative to flat
#    flat infusion       1.0000
#    6 broad pulses      0.6667
#    12 pulses           0.4000
#    16 pulses           0.2879
#    24 sharp pulses     0.1832
#    36 sharper pulses   0.1132
#    the sharpest pattern holds occupancy down to 0.1132 of the
#    flat value while delivering the same amount of hormone
#    and its peak occupancy is 0.9434 against 0.5000 flat,
#    so the signal is larger and the average load lower
#    peak to mean ratio of occupancy, by pattern:
#    flat infusion       1.000
#    6 broad pulses      2.000
#    12 pulses           4.000
#    16 pulses           5.952
#    24 sharp pulses     9.926
#    36 sharper pulses   16.667
#    that ratio is the contrast a downstream reader can see,
#    and it rises from 1.000 to 16.667 across the table

那张表里每一种型式都带着 1.0 的日均值。平铺输注给出 0.5000 的平均占据与 0.5000 的峰值占据,因为什么都不变。最陡的型式给出的平均占据是平铺值的 0.1132,同时峰值达到 0.9434。峰值与均值之比,从平铺输注的 1.000 升到最陡型式的 16.667。

这两个数字放在一起就是设计本身。受体所承受的平均负荷,也就是驱动第三章那场下调的东西,下降了将近一个数量级;而信号的高度,也就是下游读取者所能察觉的东西,几乎翻倍。脉冲性一次买到两者,别的东西都做不到。

谁听得见一次脉冲

一次脉冲只对快到足以分辨它的东西才有信息。受体的下游坐着速度相差很大的若干过程,这里把每一个都建模为带自己时间常数的一阶积分器。

推导

# Chapter 2. p04: What a pulse pattern says, and to whom. Downstream of
# the receptor sit processes with their own speeds: a gene that responds
# within minutes and a remodelling process that takes days. Each is
# modelled as a first-order integrator of receptor occupancy with its
# own time constant. A fast reader follows the pulses and sees their
# height; a slow reader averages them and sees only the daily total. So
# the same daily amount of hormone, delivered in different patterns,
# says different things to different readers, and that is what it means
# for a pattern to carry information the total does not. Each reader is
# run for sixty days and measured over the final twenty-four hours of
# that, so even the slowest has settled and the swing reported is the
# swing within one day rather than a leftover of the approach to it.
KD, HILL = 1.0, 1.0
MEAN, DT, DAYS = 1.0, 0.002, 60.0
READERS = (("fast gene, 0.2 h", 0.2), ("slow gene, 2 h", 2.0),
           ("tissue, 24 h", 24.0), ("remodelling, 120 h", 120.0))
PATTERNS = (("flat infusion", 1, 1.000), ("12 pulses", 12, 0.250),
            ("24 sharp pulses", 24, 0.100),
            ("36 sharper pulses", 36, 0.060))

def occupancy(c):
    return c ** HILL / (KD ** HILL + c ** HILL)

def drive(n_pulses, duty, t):
    height = MEAN / duty
    period = 24.0 / n_pulses
    return height if (t % period) < period * duty else 0.0

def response(n_pulses, duty, tau):
    y = 0.0
    lo = hi = None
    steps = int(round(DAYS * 24.0 / DT))
    for i in range(steps):
        t = i * DT
        y += DT * (occupancy(drive(n_pulses, duty, t)) - y) / tau
        if t > DAYS * 24.0 - 24.0:
            lo = y if lo is None else min(lo, y)
            hi = y if hi is None else max(hi, y)
    return 0.5 * (lo + hi), hi - lo

if __name__ == "__main__":
    print("   mean response of each reader, by pattern")
    print("   reader              " + "".join("{:>10}".format(p[0][:9])
                                              for p in PATTERNS))
    means = {}
    for label, tau in READERS:
        row = []
        for pl, n, duty in PATTERNS:
            m, swing = response(n, duty, tau)
            means[(label, pl)] = (m, swing)
            row.append(m)
        print("   {:<20}".format(label)
              + "".join("{:>10.4f}".format(v) for v in row))
    print("   swing of each reader, by pattern")
    print("   reader              " + "".join("{:>10}".format(p[0][:9])
                                              for p in PATTERNS))
    for label, tau in READERS:
        print("   {:<20}".format(label)
              + "".join("{:>10.4f}".format(means[(label, p[0])][1])
                        for p in PATTERNS))
    print("   ratio of the sharpest pattern to the flat infusion:")
    print("   reader              mean ratio   swing ratio")
    for label, tau in READERS:
        a = means[(label, "flat infusion")]
        b = means[(label, "36 sharper pulses")]
        print("   {:<20}{:<13.4f}{}"
              .format(label, b[0] / a[0],
                      "{:.1f}".format(b[1] / a[1]) if a[1] > 1e-5
                      else "flat has none"))
    print("   a flat infusion leaves every reader with no swing at all,")
    print("   so pattern is a channel the daily total cannot use")
    print("   the fast reader's swing does not simply grow with the pulse")
    print("   count; it peaks and then falls, because very brief pulses")
    print("   do not give the integrator time to charge:")
    print("   pulses a day   fast reader swing")
    best = None
    for n in (2, 4, 6, 8, 12, 16, 24, 36, 48, 72):
        duty = min(0.5, 6.0 / n)
        m, swing = response(n, duty, 0.2)
        if best is None or swing > best[1]:
            best = (n, swing)
        print("   {:<15d}{:.4f}".format(n, swing))
    print("   the swing is largest at {} pulses a day, at {:.4f}"
          .format(best[0], best[1]))
    print("   and the observed human axis runs near 16, which is inside")
    print("   that range rather than at either extreme")
#    mean response of each reader, by pattern
#    reader               flat infu 12 pulses 24 sharp  36 sharpe
#    fast gene, 0.2 h        0.5000    0.3678    0.1827    0.0928
#    slow gene, 2 h          0.5000    0.2061    0.0923    0.0570
#    tissue, 24 h            0.5000    0.2000    0.0909    0.0566
#    remodelling, 120 h      0.5000    0.2000    0.0910    0.0566
#    swing of each reader, by pattern
#    reader               flat infu 12 pulses 24 sharp  36 sharpe
#    fast gene, 0.2 h        0.0000    0.7348    0.3575    0.1705
#    slow gene, 2 h          0.0000    0.1478    0.0409    0.0178
#    tissue, 24 h            0.0000    0.0125    0.0034    0.0015
#    remodelling, 120 h      0.0000    0.0025    0.0007    0.0003
#    ratio of the sharpest pattern to the flat infusion:
#    reader              mean ratio   swing ratio
#    fast gene, 0.2 h    0.1856       flat has none
#    slow gene, 2 h      0.1141       flat has none
#    tissue, 24 h        0.1132       flat has none
#    remodelling, 120 h  0.1132       flat has none
#    a flat infusion leaves every reader with no swing at all,
#    so pattern is a channel the daily total cannot use
#    the fast reader's swing does not simply grow with the pulse
#    count; it peaks and then falls, because very brief pulses
#    do not give the integrator time to charge:
#    pulses a day   fast reader swing
#    2              0.6667
#    4              0.6667
#    6              0.6666
#    8              0.6660
#    12             0.6580
#    16             0.6787
#    24             0.5627
#    36             0.3593
#    48             0.2363
#    72             0.1172
#    the swing is largest at 16 pulses a day, at 0.6787
#    and the observed human axis runs near 16, which is inside
#    that range rather than at either extreme

那些慢的读取者几乎只看到均值。对二十四小时与一百二十小时的读取者,均值都落到平铺值的 0.1132,这就是 p03 的占据结果穿过来而已,而它们一天之内的摆幅可以忽略。那个快的读取者不同:它跟随脉冲,一天之内摆幅很大。

但这个快读取者的摆幅并不随脉冲频率单调增加。它在每天两次脉冲时为 0.6667,在 16 次处见顶为 0.6787,而后落到 72 次处的 0.1172,因为太短的脉冲不给积分器充电的时间。人体这条轴运行在每天十六次脉冲附近,落在那个范围之内而不在任何一端。

快读取者的摆幅对每天脉冲次数,图中标出实测频率(p04)。

假说 · 尚未验证

作者的诠释(尚未验证): 这条轴的脉冲频率由它需要驱动的最快那批转录读取者的时间常数所决定,所以糖皮质激素应答基因动力学较慢的物种或组织,应当表现出相应较慢的脉冲频率。什么会证明它错: 如果跨物种的实测脉冲频率与它们糖皮质激素应答转录的实测反应时间无关,或者更能由体重或代谢率预测,那么这一读法就是错的。

小结

在相同的日总量下,最陡的脉冲型式把平均受体占据压到平铺输注的 0.1132,同时把峰值占据从 0.5000 提到 0.9434,而峰值与均值之比从 1.000 升到 16.667。慢的读取者只看到均值;快读取者的摆幅在每天 16 次脉冲处见顶,为 0.6787,而人体这条轴正坐在那里。第三章转向受体本身,以及平均负荷对它做了什么。

参考文献

Lightman, Stafford L., and Becky L. Conway-Campbell. 2010. “The Crucial Role of Pulsatile Activity of the HPA Axis for Continuous Dynamic Equilibration.” Nature Reviews Neuroscience 11 (10): 710 to 718.
Stavreva, Diana A., Malgorzata Wiench, Sam John, et al. 2009. “Ultradian Hormone Stimulation Induces Glucocorticoid-Receptor-Mediated Pulses of Gene Transcription.” Nature Cell Biology 11 (9): 1093 to 1102.

Fan, Y. (2026). HPA-Axis Coupled Oscillations, Cortisol Awakening Response Flattening and Burnout Recovery Kinetics: Glucocorticoid Receptor Downregulation, Ultradian Pulsatility and Allostatic Load Reset [下丘脑垂体肾上腺轴耦合振荡、皮质醇觉醒反应扁平化与倦怠恢复动力学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090252 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).