arrays = [np.hstack([['One']*2, ['Two']*2]) , ['A', 'B', 'C', 'D']]
columns = pd.MultiIndex.from_arrays(arrays)
data =  pd.DataFrame(np.random.randn(5, 4), columns=list('ABCD'))
data.columns = columns
import seaborn as sns
cm = sns.light_palette("green", as_cmap=True)
data.style.background_gradient(cmap=cm, subset=['A'])

data.style.background_gradient(cmap=cm, subset=pd.IndexSlice[:, pd.IndexSlice[:, 'A']])

In [5]: data.loc[pd.IndexSlice[:, pd.IndexSlice[:, 'A']]]
Out[5]:
        One
          A
0 -0.808483
1  0.009371
2  0.977138
3 -0.875554
4 -0.052424

In [6]: data
Out[6]:
        One                 Two
          A         B         C         D
0 -0.808483 -2.280683  0.576145  0.649688
1  0.009371  0.721510  1.013764 -0.157493
2  0.977138  1.441392  1.718618 -0.320826
3 -0.875554 -1.060507  1.457075  0.570195
4 -0.052424 -0.742842 -0.203830 -1.202091

data.style.background_gradient(cmap=cm, subset=[('One','A')])

arrays = [np.hstack([['One']*2, ['Two']*2]) , ['A', 'B', 'C', 'D']]
columns = pd.MultiIndex.from_arrays(arrays)
data =  pd.DataFrame(np.random.randn(5, 4), columns=list('ABCD'))
data.columns = columns
cm = sns.light_palette("green", as_cmap=True)
data.style.background_gradient(cmap=cm, subset=[('One','A'),('Two','C')])